Compare commits

...
Author SHA1 Message Date
Austin Pickett 7d8c31186d feat(tui): interactive Plugins Hub overlay for enable/disable
The TUI had no way to toggle plugins — `/plugins` only printed a static
list, and the classic `hermes plugins` picker is curses-based and can't
run inside the Ink UI. Users had to drop to a separate shell and run
`hermes plugins enable/disable`.

Add a PluginsHub overlay modeled on the existing SkillsHub:

- New gateway RPC `plugins.manage` (list + toggle) backed by the same
  disk-discovery + dashboard_set_agent_plugin_enabled primitives the CLI
  and dashboard already use, so all three surfaces agree on state. The
  toggle path also wires the plugin's toolset into platform_toolsets.
- `/plugins` with no arg opens the hub; any subcommand still falls
  through to the text slash worker for CLI parity.
- pluginsHub overlay state threaded through overlayStore / interfaces /
  useInputHandlers (Esc closes) / appOverlays (renders the FloatBox);
  preserved across turn teardown like other user-toggled overlays.
- Hub UI: arrow/number select, Enter/Space toggles live, Tab switches
  user-only vs all (bundled) scope, shows ✓/✗/○ activation glyphs.

plugins.manage added to _LONG_HANDLERS (disk + config I/O).
2026-06-09 00:23:50 -04:00
teknium1 3dcfbbfc49 chore(release): add underthestars-zhy to AUTHOR_MAP
Salvage follow-up for PR #42444 — maps the contributor's commit email
so the changelog generator can attribute the Photon gRPC channel work.
2026-06-08 21:03:58 -07:00
underthestars-zhy 3b983e7791 fix(photon): add home channel env seed and simplify space resolution 2026-06-08 21:03:58 -07:00
underthestars-zhy 0d25cae041 fix(photon): remove reply-to support and fix typing API
Drop `replyTo` from all outbound send paths and update the `/typing`
endpoint to use the documented `typing("start" | "stop")` content
builder. Adds a `stop_typing` method on the adapter to pair with
`send_typing`.
2026-06-08 21:03:58 -07:00
underthestars-zhy e79e44af79 fix(photon): use spectrum-ts reply builder for threaded messages
Replace raw `{ replyTo }` send options with the `spectrumReply` content
builder from spectrum-ts, which is the correct API for threading
replies.
Adds `maybeReplyContent` helper with graceful fallback to normal send
when
the reply target cannot be resolved.
2026-06-08 21:03:58 -07:00
underthestars-zhy fdf48c63c8 fix(photon): wrap text sends with spectrumText helper 2026-06-08 21:03:58 -07:00
underthestars-zhy 0646656884 fix(photon): support E.164 and DM GUID targets for home channel
Allow PHOTON_HOME_CHANNEL to accept a bare E.164 phone number or a
`any;-;+1...` DM chat GUID in addition to a Spectrum space id. Inbound
DM spaces are cached so replies resolve without a second SDK lookup,
and `photon` is added to _PHONE_PLATFORMS so send_message treats E.164
strings as explicit targets rather than falling through to channel-name
resolution.
2026-06-08 21:03:58 -07:00
underthestars-zhy 92179352fb feat(photon): auto-configure allowlist and cron channel on setup
During `hermes photon setup`, allowlist the operator's number and set
their DM as the cron home channel when those env vars are unset. Without
this, the gateway denies the operator's own messages and cron has no
default delivery target. Re-runs never overwrite hand-tuned values.

Also teaches the sidecar's `resolveSpace` to accept a bare E.164 number
as a space identifier, resolving it to the user's DM space so
`PHOTON_HOME_CHANNEL` can be set to a phone number instead of an opaque
space id.
2026-06-08 21:03:58 -07:00
underthestars-zhy e9b26c7c8b style(photon): Colorize iMessage number box in setup output 2026-06-08 21:03:58 -07:00
underthestars-zhy 84e4b4b9a5 fix(photon): use per-user assigned line for agent iMessage number
On shared-number plans, `/lines` has no dedicated entry, so the
`assignedPhoneNumber` field on the user object is the source of truth
for which number to text the agent. Fall back to the line inventory
only when no per-user assignment exists.
2026-06-08 21:03:58 -07:00
underthestars-zhy 314af28e86 feat(photon): download and inline inbound attachments 2026-06-08 21:03:58 -07:00
underthestars-zhy b3aef57f21 refactor(photon): use TYPE_CHECKING for httpx import and fix client ref 2026-06-08 21:03:58 -07:00
underthestars-zhyandClaude Opus 4.8 4e4d27875f feat(photon): gRPC-native iMessage channel (no webhook)
Make Photon iMessage a first-class persistent-connection channel like
Discord/Slack, using the spectrum-ts gRPC stream for both directions.

- Inbound: the sidecar forwards the SDK's app.messages gRPC stream to the
  adapter over a loopback GET /inbound (NDJSON) instead of webhooks. Drops
  the aiohttp webhook server, HMAC signature verification, public URL, and
  PHOTON_WEBHOOK_* config; adapter reconnects with backoff.
- Management plane: device login uses client_id=photon-cli against the
  single dashboard host (Bearer), matching the official photon-hq/cli;
  find-or-create "Hermes Agent" project, enable Spectrum, rotate secret,
  register user (with phone dedup), surface the assigned iMessage line.
- SDK projectId is the project's spectrumProjectId, not the dashboard id;
  runtime creds persist to ~/.hermes/.env like every other channel.
- CLI: 6-step setup, webhook subcommands removed.
- Tests/docs updated for the gRPC flow; sidecar pins spectrum-ts ^1.17.1.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-08 21:03:58 -07:00
teknium1 c3420d91ad chore: add jooray to AUTHOR_MAP for salvaged simplex PR #27978 2026-06-08 21:03:45 -07:00
Juraj Bednar 0c2e81df00 feat(simplex): groups, native attachments, text batching, auto-accept
Salvage of PR #27978 cherry-picked onto current main, resolving conflicts
with main's intervening SimpleX plugin fixes (resp-envelope normalization,
health-monitor reconnect-churn fix, bare-form DM addressing).

What's new:
- Group support via SIMPLEX_GROUP_ALLOWED (comma-separated IDs or '*');
  inbound items surface chat_id=group:<id> + chat_type=group. Disabled by
  default so a bot in a group doesn't process every member's traffic.
- Inbound files/voice via rcvFileDescrReady (immediate /freceive) deferred
  through _pending_file_transfers, replayed on rcvFileComplete. Voice notes
  -> MessageType.VOICE.
- Native outbound media: send_image (PNG/JPEG + inline thumbnail), send_voice
  (msgContent.type=voice), send_video, send_document. All addressed by numeric
  ID via /_send ... json [...].
- MEDIA:<path> tags in agent replies stripped and dispatched as voice/document.
- Text-burst batching (HERMES_SIMPLEX_TEXT_BATCH_DELAY, default 0.8s).
- Auto-accept contact requests (SIMPLEX_AUTO_ACCEPT, default true).
- Group send path uses structured /_send #<id> json form (the bracket
  #[<id>] form is parsed as display-name lookup and silently drops).

plugin.yaml bumped to 1.1.0; docs updated. All inside plugins/platforms/simplex/
- no core edits.

Co-authored-by: Juraj Bednar <juraj@bednar.io>
2026-06-08 21:03:45 -07:00
Ben BarclayandHermes Agent a46462ec65 fix(cli): persist custom --portal-url to .env on dashboard register (#42435)
* fix(cli): persist custom --portal-url to .env on dashboard register

`hermes dashboard register --portal-url <url>` resolved the custom portal
for the registration request but only persisted it to .env when the var was
absent AND non-default. So a user who re-registered against a different
portal (e.g. switching preview deploys) silently kept the stale
HERMES_DASHBOARD_PORTAL_URL, and an explicit request for the production
portal was never written at all.

Track whether a custom portal was *explicitly supplied* (--portal-url flag
or HERMES_DASHBOARD_PORTAL_URL env), separately from the resolved value:

  - explicit custom URL -> always persist (update in place via
    save_env_value, which overwrites the matching key rather than appending
    a duplicate), even when it equals the production default; no-op when it
    already matches.
  - no custom URL supplied -> unchanged conservative behaviour: only write an
    inferred portal when absent and non-default; never alter an existing
    entry unexpectedly.

save_env_value already preserves other lines/comments and dedups in place;
this only changes the decision of *when* to call it.

Adds TestCustomPortalPersistence covering all four cases.

Co-authored-by: Hermes Agent <agent@nousresearch.com>

* feat(cli): persist dashboard public URL from --redirect-uri on register

When the user registers a publicly-exposed dashboard with --redirect-uri
(the full OAuth callback, e.g. https://hermes.example.com/auth/callback),
derive its origin and persist it as HERMES_DASHBOARD_PUBLIC_URL — the env var
the dashboard auth layer actually consumes at serve time.

dashboard_auth/routes._redirect_uri reconstructs the callback as
HERMES_DASHBOARD_PUBLIC_URL + "/auth/callback" (verbatim), and
dashboard_auth/prefix.resolve_public_url reads that var (then config.yaml
dashboard.public_url) to decide the public origin. Previously --redirect-uri
was sent to the portal at registration but never persisted, so the operator
had to set HERMES_DASHBOARD_PUBLIC_URL by hand for the login gate to engage
and the callback to round-trip. We now wire it automatically.

Persist the ORIGIN (scheme://host[:port]), not the full callback path —
persisting the raw redirect would double the path when the runtime appends
/auth/callback. Mirrors the portal-url persistence semantics already in this
PR: always write an explicitly-derived value (updating in place, no
duplicate), no-op when it already matches, never written on a localhost-only
install (no --redirect-uri), and skipped for a non-http(s)/malformed redirect.

Verified end-to-end: cmd_dashboard_register writes the origin to .env, then
resolve_public_url() reads it back and public_url + /auth/callback
reconstructs exactly the originally-supplied --redirect-uri.

Adds TestPublicUrlPersistence (8 cases) incl. origin-derivation, port
preservation, update-in-place, no-op, no-flag, non-http skip, and
both-portal-and-public-url-persisted.

Co-authored-by: Hermes Agent <agent@nousresearch.com>

---------

Co-authored-by: Hermes Agent <agent@nousresearch.com>
2026-06-09 13:56:33 +10:00
helix4u b23184cad4 fix(api-server): bind request session context for tools 2026-06-08 20:52:08 -07:00
Ben Barclay 52ae9d9f02 feat(dashboard): make hermes dashboard register idempotent (#42455)
Re-running `hermes dashboard register` now updates the existing dashboard
record in nous-account-service instead of creating a duplicate.

The stable key is the client_id this install already persisted in
HERMES_DASHBOARD_OAUTH_CLIENT_ID on a prior run:
- No stored client_id -> first registration -> create a fresh client with an
  auto-generated name (unchanged behavior).
- Stored client_id present -> re-send it as `client_id` so the portal updates
  that row in place. Without an explicit --name, the name is omitted so the
  portal-stored name isn't churned to a new random value on every re-run.
- Prints "Updated dashboard" vs "Registered dashboard" based on whether the
  portal echoed back the same client_id. A stale/deleted id safely falls
  through to a fresh create server-side.

Requires the matching nous-account-service change (POST
/api/oauth/self-hosted-client accepting an optional client_id + optional name).

Tests: 7 new TestIdempotentRerun cases (key sent, name preserved/overridden,
Updated message, persisted id, stale-id fall-through, blank-id first-run);
existing create-path tests unchanged (23 pass).
2026-06-09 13:19:35 +10:00
brooklyn! 1e5ff4a577 fix(hermes-ink): disable mouse tracking on raw-mode teardown to stop SGR leak (#42527)
The raw-mode teardown path (rawModeEnabledCount -> 0) disabled
modifyOtherKeys, kitty keyboard, focus reporting, and bracketed paste,
then dropped raw mode and detached the readable listener -- but left DEC
mouse tracking (1000/1002/1003/1006) asserted. With raw mode off and no
reader attached, the terminal falls back to cooked-mode echo, so every
mouse move emits a hover report (DEC 1003) that prints as literal text:
a flood of '35;col;row M' shards over the prompt in a long session.

handleSuspend() already guards against exactly this (it writes
DISABLE_MOUSE_TRACKING before SIGSTOP); the ordinary teardown path
missed the same guard. Add DISABLE_MOUSE_TRACKING to the teardown, and
re-assert tracking on raw-mode re-entry (via the Ink instance's
reassertTerminalModes, which is gated on altScreenActive and idempotent)
so a transient drop->re-add round-trips cleanly instead of silently
leaving the mouse dead.

Adds a regression test driving a real Ink mount: the last raw-mode
consumer detaching must emit DISABLE_MOUSE_TRACKING.

Reported via a community bug report.
2026-06-08 21:31:06 -05:00
Jeffrey Quesnelle 6a8dda171c Merge pull request #42515 from NousResearch/fix/desktop-debug-report-links
fix(desktop): render debug-report paste URLs as real clickable links
2026-06-08 22:19:17 -04:00
emozilla e0f6a35ac6 fix(desktop): render debug-report paste URLs as real clickable links
System messages (slash-command output like /debug, plus the generic
system-message fallback) were rendered as plain text, so the uploaded
paste.rs URLs in a debug report were neither clickable nor easily
copyable.

Route both through LinkifiedText so URLs become real <a> links (open
externally via the desktop bridge, selectable/copyable text). Add an
opt-in explicitOnly mode that matches only explicit http(s):// / www.
URLs, used here so filename-shaped tokens in the report (agent.log,
errors.log, gateway.log) aren't mistaken for bare domains and linkified.
Bare-domain matching is preserved for all other LinkifiedText callers.

Adds regression tests covering explicitOnly (links only real URLs, keeps
.log filenames as text) and the default bare-domain behavior.
2026-06-08 21:35:21 -04:00
teknium1 b5f8996ccc test(cli): exercise real _prompt_text_input for native-Windows confirm deadlock
The existing #33961 tests mock _prompt_text_input away, so they only assert
modal-vs-stdin routing — they cannot observe the actual hang. Add a guard
class that drives the real helper chain with a blocking input() on a win32
daemon thread and asserts the worker never hangs. Fails on the pre-#33961
code (win32 -> _prompt_text_input -> off-main input() -> deadlock), passes
on the modal path. Also covers the scheduling-failure degraded branch
(must clean-cancel to None, never call input()).
2026-06-08 15:53:28 -07:00
firefly 714183530b test(cli): convert stale win32 stdin-fallback tests to the modal contract
The four win32 tests asserted the old deadlocking behavior (win32 -> raw
input()). Rewrite them to the corrected contract: native Windows uses the
modal via the app loop, and stdin is kept only for the safe no-app /
scheduling-failure cases. Consolidate three near-identical daemon-thread
tests into one parametrized (linux/win32) test behind a shared _run_on_daemon
harness, and drop dead code from the old main-thread test.

Refs #33961
2026-06-08 15:53:28 -07:00
firefly ab98818e5b fix(cli): use the confirm modal on native Windows instead of deadlocking input()
Native Windows bypassed the destructive-slash modal and fell back to a raw
input() prompt. When the confirm was triggered from the process_loop daemon
thread (the normal case), that input() deadlocked against prompt_toolkit's
main-thread stdin ownership: bare /reset froze with Ctrl-C swallowed, while
/reset now worked only because it skips the prompt. Route native Windows
through the existing call_soon_threadsafe modal path (the same key-binding
channel that already handles normal typing on Windows); keep the stdin
fallback only for the safe no-app / scheduling-failure cases, and clean-cancel
(None) off the main thread on win32 so a degraded path never re-deadlocks.

Addresses #33961
Refs #30768
2026-06-08 15:53:28 -07:00
firefly d66bac5a1a test(cli): failing regression test for native-Windows confirm deadlock (#33961) 2026-06-08 15:53:28 -07:00
teknium1 300371c3f2 chore: add AUTHOR_MAP entry for ruangraung (PR #42308 salvage) 2026-06-08 15:53:16 -07:00
ruangraung f4531feee8 fix(telegram): improve MarkdownV2 edit fallback and fix _strip_mdv2 bold handling
When edit_message(finalize=True) fails with a MarkdownV2 parse error,
the silent fallback previously sent raw content with escape sequences.
Now it logs the error and strips markdown formatting via _strip_mdv2()
for clean plain-text fallback.

Also fixes _strip_mdv2 to handle standard markdown bold (\*\*text\*\*)
before MarkdownV2 bold (\*text\*), preventing half-stripped asterisks.

Refs: #41955, #41732
2026-06-08 15:53:16 -07:00
ruangraung 6d2732e786 fix(gateway): apply MarkdownV2 formatting on progress message edits
When a platform adapter sets REQUIRES_EDIT_FINALIZE=True (e.g.
TelegramAdapter), tool progress edits now pass finalize=True so
format_message() is applied before sending to the platform.

Previously, the initial send() formatted the message correctly via
MarkdownV2, but subsequent edit_message() calls skipped formatting
(finalize=False), causing raw markdown (e.g. triple backticks for
bash code blocks) to render as plain text on Telegram.

Refs: #41955, #41732
2026-06-08 15:53:16 -07:00
teknium1 aa424e51ac refactor(doctor): fold custom-provider vendor-slug check into one predicate
Collapse the bare-"custom" allowlist entry and the custom:<name> guard into
a single provider_accepts_vendor_slug predicate so the slug-warning suppression
reads as one rule instead of two scattered conditions. No behavior change.
2026-06-08 15:53:09 -07:00
helix4u 732ababa1a fix(doctor): allow vendor slugs for named custom providers 2026-06-08 15:53:09 -07:00
GodsBoy 421226e404 fix(gateway): stop terminal progress from posting the full command to messaging chats
#41215 rendered a terminal tool call as a native ```bash fenced block on
markdown platforms (Telegram, WhatsApp, Slack, and others), showing the full
command with no truncation, in both all/new and verbose modes. That posted
complete shell commands (heredocs, internal paths, destructive commands) into
the chat before the final answer, visible to everyone in it.

This restores the prior behavior: terminal progress shows the short, truncated
preview line that every other tool already uses, capped at tool_preview_length.
The supports_code_blocks capability flag is left in place for future use.
CLI/TUI rendering is a separate path and was unaffected.

Adds a regression test asserting terminal progress renders as a truncated
preview, not a fenced bash block, even on a markdown-capable gateway.

Fixes #41955
2026-06-08 15:53:00 -07:00
Ray Sun 37561c214b fix(photon): use allowlisted device client_id + validate token before save
Photon now allowlists registered device clients on the device-code
endpoint; the old client_id "hermes-agent" is rejected with
400 invalid_client, breaking the entire login flow. Switch to Photon's
published "photon-cli" device client and send the standard scope.

Also validate the device-flow token against /api/auth/get-session and
/api/projects/ before persisting it, and extract token candidates from
every response shape Photon has used (access_token, accessToken,
data.*, set-auth-token header) so a token that authenticates the
session lookup but is rejected by the project API fails loudly at
login instead of 404ing downstream.

Verified live: request_device_code() now returns 200 + a valid
user_code where "hermes-agent" returned 400 invalid_client.

Salvaged from #34467 by @yanxue06.
2026-06-08 15:52:33 -07:00
51 changed files with 6473 additions and 1792 deletions
@@ -77,6 +77,7 @@ import type { HermesGateway } from '@/hermes'
import { useResizeObserver } from '@/hooks/use-resize-observer'
import { useI18n } from '@/i18n'
import { DATA_IMAGE_URL_RE } from '@/lib/embedded-images'
import { LinkifiedText } from '@/lib/external-link'
import { triggerHaptic } from '@/lib/haptics'
import { GitBranchIcon, Loader2Icon, Volume2Icon, VolumeXIcon } from '@/lib/icons'
import { extractPreviewTargets } from '@/lib/preview-targets'
@@ -919,7 +920,7 @@ const SystemMessage: FC = () => {
>
<span className="font-mono text-muted-foreground/55">{slashStatus.groups.command}</span>
<span className="mx-1.5 text-muted-foreground/35">·</span>
<span className="whitespace-pre-wrap">{slashStatus.groups.output.trim()}</span>
<LinkifiedText className="whitespace-pre-wrap" explicitOnly pretty={false} text={slashStatus.groups.output.trim()} />
</MessagePrimitive.Root>
)
}
@@ -930,7 +931,7 @@ const SystemMessage: FC = () => {
data-role="system"
data-slot="aui_system-message-root"
>
<span className="whitespace-pre-wrap">{text}</span>
<LinkifiedText className="whitespace-pre-wrap" explicitOnly pretty={false} text={text} />
</MessagePrimitive.Root>
)
}
@@ -165,4 +165,31 @@ describe('external link helpers', () => {
'https://expedia.com/things-to-do/puerto-rico-el-yunque-rainforest-adventure'
)
})
it('explicitOnly skips bare filename/domain tokens and only links explicit URLs', () => {
installDesktopBridge()
render(
<LinkifiedText
explicitOnly
pretty={false}
text={'Report https://paste.rs/abc\nagent.log https://paste.rs/def\nerrors.log'}
/>
)
const links = screen.getAllByRole('link')
expect(links.map(a => a.getAttribute('href'))).toEqual(['https://paste.rs/abc', 'https://paste.rs/def'])
// Bare filename-shaped tokens stay as plain text, not links.
expect(screen.queryByText(content => content.includes('agent.log'))).toBeTruthy()
expect(links.some(a => (a.textContent ?? '').includes('.log'))).toBe(false)
})
it('without explicitOnly, bare filename tokens are still linkified (default behavior)', () => {
installDesktopBridge()
render(<LinkifiedText pretty={false} text="open agent.log please" />)
const link = screen.getByRole('link', { name: 'agent.log' })
expect(link.getAttribute('href')).toBe('https://agent.log')
})
})
+9 -2
View File
@@ -12,6 +12,12 @@ const titleSubs = new Map<string, Set<(value: string) => void>>()
const URL_RE =
/(?:https?:\/\/|www\.)[^\s<>"'`]+[^\s<>"'`.,;:!?)]|[a-z0-9](?:[a-z0-9-]*\.)+[a-z]{2,}(?:\/[^\s<>"'`.,;:!?)]*)?/gi
// Explicit-scheme / www. URLs only — no bare-domain matching. Used where the
// surrounding text is full of filename-shaped tokens (e.g. `agent.log`,
// `errors.log` in a /debug report) that the bare-domain branch of URL_RE would
// otherwise mistake for domains and linkify.
const EXPLICIT_URL_RE = /(?:https?:\/\/|www\.)[^\s<>"'`]+[^\s<>"'`.,;:!?)]/gi
const DOMAIN_RE = /^(?:www\.)?[a-z0-9](?:[a-z0-9-]*\.)+[a-z]{2,}(?::\d+)?(?:[/?#][^\s]*)?$/i
const SKIP_PROTO_RE = /^(?:file|data|mailto|javascript|blob|chrome|about|hermes):/i
const LOCAL_HOST_RE = /^(?:localhost|127\.0\.0\.1|0\.0\.0\.0|\[::1\])(?::\d+)?$/i
@@ -261,13 +267,14 @@ interface LinkifiedTextProps {
className?: string
text: string
pretty?: boolean
explicitOnly?: boolean
}
export function LinkifiedText({ className, pretty = true, text }: LinkifiedTextProps) {
export function LinkifiedText({ className, explicitOnly = false, pretty = true, text }: LinkifiedTextProps) {
const nodes: ReactNode[] = []
let cursor = 0
for (const match of text.matchAll(URL_RE)) {
for (const match of text.matchAll(explicitOnly ? EXPLICIT_URL_RE : URL_RE)) {
const raw = match[0]
const url = normalizeExternalUrl(raw)
const index = match.index ?? 0
+33 -35
View File
@@ -6218,27 +6218,20 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
choices visible and lets the normal Enter key binding submit the typed
or highlighted choice.
**Platform note (Windows dead-lock issue #30768):**
The queue-based modal relies on prompt_toolkit key bindings receiving
keyboard events and calling ``_submit_slash_confirm_response``. On
Windows (PowerShell / Windows Terminal) the prompt_toolkit input
channel can become unresponsive when the modal is entered from the
``process_loop`` daemon thread, causing a dead-lock: the user sees the
confirmation panel but keystrokes never reach the key bindings and the
``response_queue.get()`` blocks until the 120-second timeout expires.
**Platform note (Windows issue #33961):**
Earlier code bypassed the modal on ``sys.platform == "win32"`` and fell
back to a raw ``input()`` prompt. When the confirm was triggered from the
``process_loop`` daemon thread (the normal case) that ``input()`` ran off
the main thread and deadlocked against prompt_toolkit's stdin ownership —
the user saw a frozen cursor and Ctrl-C was swallowed (bare ``/reset``
froze; ``/reset now`` worked only because it skips the prompt entirely).
To avoid this, we fall back to ``_prompt_text_input`` (a simple
``input()``-based prompt) when any of these conditions hold:
* ``sys.platform == "win32"`` native Windows console (ConPTY /
win32_input) does not support the modal reliably.
* ``self._app`` is not set unit tests / non-interactive contexts.
On non-Windows platforms the modal itself is still safe from the
``process_loop`` daemon thread as long as the main-thread event loop
owns the prompt_toolkit buffer mutations. When we are off the main
thread, schedule the modal snapshot / restore work on ``self._app.loop``
via ``call_soon_threadsafe`` and keep the queue-based response path.
Native Windows now uses the same path as Linux/macOS: the modal is set up
on ``self._app.loop`` via ``call_soon_threadsafe`` and answered by the
normal prompt_toolkit key bindings (the same input channel that already
handles ordinary typing on Windows). The raw ``input()`` fallback is kept
only for the genuinely safe cases: no running app (unit tests /
non-interactive), no resolvable event loop, or a scheduling failure.
"""
import threading
import time as _time
@@ -6251,23 +6244,26 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
if not getattr(self, "_app", None):
return self._prompt_text_input("Choice [1/2/3]: ")
# On Windows the prompt_toolkit input channel can deadlock when the
# modal is entered from the process_loop daemon thread — keystrokes
# never reach the key bindings, so response_queue.get() blocks for
# the full timeout (issue #30768). Fall back to the simpler
# stdin-based prompt which works reliably on Windows.
if sys.platform == "win32":
return self._prompt_text_input("Choice [1/2/3]: ")
try:
app_loop = self._app.loop
except Exception:
app_loop = None
in_main_thread = threading.current_thread() is threading.main_thread()
if not in_main_thread and app_loop is None:
def _stdin_fallback() -> str | None:
# On native Windows a raw input() from a non-main thread deadlocks
# against prompt_toolkit's stdin ownership (#33961). With an app
# running we cannot safely prompt off the main thread, so cancel
# cleanly (None) rather than hang the terminal.
if sys.platform == "win32" and not in_main_thread:
self._invalidate()
return None
return self._prompt_text_input("Choice [1/2/3]: ")
if not in_main_thread and app_loop is None:
return _stdin_fallback()
response_queue = queue.Queue()
def _setup_modal() -> None:
@@ -6307,7 +6303,7 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
return ready.wait(timeout=5)
if not _run_on_app_loop(_setup_modal):
return self._prompt_text_input("Choice [1/2/3]: ")
return _stdin_fallback()
_last_countdown_refresh = _time.monotonic()
try:
@@ -8223,9 +8219,10 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
print(" ⚠️ MCP reload timed out (30s). Some servers may not have reconnected.")
# Inline-skip tokens that bypass the destructive-slash confirmation modal.
# Matches the escape-hatch pattern users on broken modal platforms
# (currently native Windows PowerShell — issue #30768) need to self-serve
# without having to flip approvals.destructive_slash_confirm in config.
# A general escape hatch for non-interactive use (scripting/automation) and
# for the degraded path where the modal can't be marshaled onto the app loop
# — lets users self-serve without flipping approvals.destructive_slash_confirm
# in config. (Native Windows now drives the modal normally — see #33961.)
_DESTRUCTIVE_SKIP_TOKENS = frozenset({"now", "--yes", "-y"})
@classmethod
@@ -8283,8 +8280,9 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
Inline-skip: if ``cmd_original`` contains ``now``, ``--yes``, or
``-y`` as an argument (e.g. ``/reset now``, ``/new --yes My title``),
the modal is bypassed and ``"once"`` is returned immediately. This is
an escape hatch for platforms where the prompt_toolkit modal hangs
(issue #30768 — native Windows PowerShell). Callers are responsible
an escape hatch for non-interactive use and for the degraded path where
the modal can't be marshaled onto the app loop (native Windows itself now
drives the modal normally see #33961). Callers are responsible
for stripping the skip tokens from any remaining argument parsing
(see :meth:`_split_destructive_skip`).
+39 -28
View File
@@ -3510,35 +3510,46 @@ class APIServerAdapter(BasePlatformAdapter):
loop = asyncio.get_running_loop()
def _run():
agent = self._create_agent(
ephemeral_system_prompt=ephemeral_system_prompt,
session_id=session_id,
stream_delta_callback=stream_delta_callback,
tool_progress_callback=tool_progress_callback,
tool_start_callback=tool_start_callback,
tool_complete_callback=tool_complete_callback,
gateway_session_key=gateway_session_key,
from gateway.session_context import clear_session_vars, set_session_vars
tokens = set_session_vars(
platform="api_server",
chat_id=session_id or "",
session_key=gateway_session_key or session_id or "",
session_id=session_id or "",
)
if agent_ref is not None:
agent_ref[0] = agent
effective_task_id = session_id or str(uuid.uuid4())
result = agent.run_conversation(
user_message=user_message,
conversation_history=conversation_history,
task_id=effective_task_id,
)
usage = {
"input_tokens": getattr(agent, "session_prompt_tokens", 0) or 0,
"output_tokens": getattr(agent, "session_completion_tokens", 0) or 0,
"total_tokens": getattr(agent, "session_total_tokens", 0) or 0,
}
# Include the effective session ID in the result so callers
# (e.g. X-Hermes-Session-Id header) can track compression-
# triggered session rotations. (#16938)
_eff_sid = getattr(agent, "session_id", session_id)
if isinstance(_eff_sid, str) and _eff_sid:
result["session_id"] = _eff_sid
return result, usage
try:
agent = self._create_agent(
ephemeral_system_prompt=ephemeral_system_prompt,
session_id=session_id,
stream_delta_callback=stream_delta_callback,
tool_progress_callback=tool_progress_callback,
tool_start_callback=tool_start_callback,
tool_complete_callback=tool_complete_callback,
gateway_session_key=gateway_session_key,
)
if agent_ref is not None:
agent_ref[0] = agent
effective_task_id = session_id or str(uuid.uuid4())
result = agent.run_conversation(
user_message=user_message,
conversation_history=conversation_history,
task_id=effective_task_id,
)
usage = {
"input_tokens": getattr(agent, "session_prompt_tokens", 0) or 0,
"output_tokens": getattr(agent, "session_completion_tokens", 0) or 0,
"total_tokens": getattr(agent, "session_total_tokens", 0) or 0,
}
# Include the effective session ID in the result so callers
# (e.g. X-Hermes-Session-Id header) can track compression-
# triggered session rotations. (#16938)
_eff_sid = getattr(agent, "session_id", session_id)
if isinstance(_eff_sid, str) and _eff_sid:
result["session_id"] = _eff_sid
return result, usage
finally:
clear_session_vars(tokens)
return await loop.run_in_executor(None, _run)
+4 -2
View File
@@ -1795,9 +1795,11 @@ class BasePlatformAdapter(ABC):
# Whether this platform renders triple-backtick fenced code blocks (i.e.
# ``format_message`` translates/preserves markdown fences into a real code
# block). Drives presentation choices like rendering a ``terminal`` tool
# call's command as a ```bash block instead of a flat preview line.
# block). Capability flag for markdown-aware presentation choices.
# Default False (plain-text platforms); markdown-rendering adapters set True.
# Note: tool-progress deliberately does NOT use this to render a terminal
# command as a ```bash block — that exposed full commands in chat. Progress
# shows a short truncated preview only (see gateway/run.py progress_callback).
supports_code_blocks: bool = False
def __init__(self, config: PlatformConfig, platform: Platform):
+10 -2
View File
@@ -181,6 +181,8 @@ def _strip_mdv2(text: str) -> str:
"""
# Remove escape backslashes before special characters
cleaned = re.sub(r'\\([_*\[\]()~`>#\+\-=|{}.!\\])', r'\1', text)
# Remove standard markdown bold (**text** → text) BEFORE MarkdownV2 bold
cleaned = re.sub(r'\*\*([^*]+)\*\*', r'\1', cleaned)
# Remove MarkdownV2 bold markers that format_message converted from **bold**
cleaned = re.sub(r'\*([^*]+)\*', r'\1', cleaned)
# Remove MarkdownV2 italic markers that format_message converted from *italic*
@@ -2208,11 +2210,17 @@ class TelegramAdapter(BasePlatformAdapter):
# "Message is not modified" is a no-op, not an error
if "not modified" in str(fmt_err).lower():
return SendResult(success=True, message_id=message_id)
# Fallback: retry without markdown formatting
# Fallback: strip MarkdownV2 escapes and retry as clean plain text
logger.warning(
"[%s] MarkdownV2 edit failed, falling back to plain text: %s",
self.name,
fmt_err,
)
_plain = _strip_mdv2(content) if content else content
await self._bot.edit_message_text(
chat_id=int(chat_id),
message_id=int(message_id),
text=content,
text=_plain,
)
return SendResult(success=True, message_id=message_id)
except Exception as e:
+4 -26
View File
@@ -12971,32 +12971,10 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
# Build progress message with primary argument preview
from agent.display import get_tool_emoji
emoji = get_tool_emoji(tool_name, default="⚙️")
# Markdown-capable platforms render a terminal command as a native
# ```bash fenced block (full command, no quotes, no label, no
# truncation) instead of the noisy `terminal: "cmd…"` line. Gated
# on the adapter's ``supports_code_blocks`` capability so every
# markdown-rendering platform (and plugin adapters that opt in) gets
# it, while plain-text platforms keep the compact line.
_bash_block = None
try:
_progress_adapter = self.adapters.get(source.platform)
except Exception:
_progress_adapter = None
if (
getattr(_progress_adapter, "supports_code_blocks", False)
and tool_name == "terminal"
and isinstance(args, dict)
and isinstance(args.get("command"), str)
and args["command"].strip()
):
_bash_block = f"```bash\n{args['command'].rstrip()}\n```"
# Verbose mode: show detailed arguments, respects tool_preview_length
if progress_mode == "verbose":
if _bash_block is not None:
msg = _bash_block
elif args:
if args:
from agent.display import get_tool_preview_max_len
_pl = get_tool_preview_max_len()
args_str = json.dumps(args, ensure_ascii=False, default=str)
@@ -13016,9 +12994,7 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
# "all" / "new" modes: short preview, respects tool_preview_length
# config (defaults to 40 chars when unset to keep gateway messages
# compact — unlike CLI spinners, these persist as permanent messages).
if _bash_block is not None:
msg = _bash_block
elif preview:
if preview:
from agent.display import get_tool_preview_max_len
_pl = get_tool_preview_max_len()
_cap = _pl if _pl > 0 else 40
@@ -13130,6 +13106,8 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
"message_id": message_id,
"content": content,
}
if getattr(adapter, "REQUIRES_EDIT_FINALIZE", False):
kwargs["finalize"] = True
if _edit_accepts_metadata:
kwargs["metadata"] = _progress_metadata
return await adapter.edit_message(**kwargs)
+3
View File
@@ -106,6 +106,7 @@ def set_session_vars(
user_id: str = "",
user_name: str = "",
session_key: str = "",
session_id: str = "",
message_id: str = "",
cwd: str = "",
) -> list:
@@ -127,6 +128,7 @@ def set_session_vars(
_SESSION_USER_ID.set(user_id),
_SESSION_USER_NAME.set(user_name),
_SESSION_KEY.set(session_key),
_SESSION_ID.set(session_id),
_SESSION_MESSAGE_ID.set(message_id),
]
try:
@@ -157,6 +159,7 @@ def clear_session_vars(tokens: list) -> None:
_SESSION_USER_ID,
_SESSION_USER_NAME,
_SESSION_KEY,
_SESSION_ID,
_SESSION_MESSAGE_ID,
):
var.set("")
+139 -12
View File
@@ -94,19 +94,36 @@ def _register_self_hosted_client(
*,
access_token: str,
portal_base_url: str,
name: str,
name: Optional[str],
custom_redirect_uri: Optional[str],
existing_client_id: Optional[str] = None,
timeout: float = 15.0,
) -> dict:
"""POST to the portal's self-hosted-client endpoint and return the JSON body.
When ``existing_client_id`` is provided (the client_id this install
persisted on a prior run), it is sent so the portal updates that existing
dashboard record in place instead of minting a duplicate — this is what
makes re-running ``hermes dashboard register`` idempotent. The portal
falls back to creating a fresh client if the id no longer resolves to a row
in the caller's org (stale/deleted), so passing it is always safe.
``name`` may be ``None`` on the idempotent update path (re-run without an
explicit ``--name``): omitting it tells the portal to keep the name it
already stored rather than overwriting it. It is required on the create
path; the caller guarantees a value there.
Raises RuntimeError with a user-facing message on any non-2xx response or
transport failure.
"""
url = f"{portal_base_url.rstrip('/')}/api/oauth/self-hosted-client"
body: dict[str, str] = {"name": name}
body: dict[str, str] = {}
if name:
body["name"] = name
if custom_redirect_uri:
body["custom_redirect_uri"] = custom_redirect_uri
if existing_client_id:
body["client_id"] = existing_client_id
data = json.dumps(body).encode("utf-8")
req = urllib.request.Request(
@@ -165,16 +182,20 @@ def _print_post_register_hint(
portal_base_url: str,
custom_redirect_uri: Optional[str],
wrote_portal_url: bool,
public_url: str = "",
) -> None:
"""Print the success summary + the gate-engagement caveat."""
from hermes_cli.config import get_env_path
env_path = get_env_path()
_cid = client_id
print()
print(f" Wrote to {env_path}:")
print(f" HERMES_DASHBOARD_OAUTH_CLIENT_ID={client_id}")
print(" HERMES_DASHBOARD_OAUTH_CLIENT_ID=" + str(_cid))
if wrote_portal_url:
print(f" HERMES_DASHBOARD_PORTAL_URL={portal_base_url}")
print(" HERMES_DASHBOARD_PORTAL_URL=" + str(portal_base_url))
if public_url:
print(" HERMES_DASHBOARD_PUBLIC_URL=" + str(public_url))
print()
print(
" Heads up — Nous login only *engages* on a non-loopback bind. A plain\n"
@@ -240,12 +261,48 @@ def cmd_dashboard_register(args) -> None:
# Portal override: explicit --portal-url flag wins, else the
# HERMES_DASHBOARD_PORTAL_URL env var, else the stored login's portal.
#
# We track whether a custom URL was *explicitly supplied* (flag or env)
# separately from the resolved value. An explicit custom URL is an
# intentional choice the user wants to persist (and update in place if it
# already exists in .env); a portal merely inferred from the stored login
# keeps the older, more conservative write-only-if-absent behaviour so we
# don't clutter .env for the common production case.
portal_override = getattr(args, "portal_url", None) or os.environ.get(
"HERMES_DASHBOARD_PORTAL_URL"
)
custom_portal_supplied = bool(
isinstance(portal_override, str) and portal_override.strip()
)
portal_base_url = _resolve_portal_base_url(portal_override)
name = getattr(args, "name", None) or _generate_dashboard_name()
# Idempotency: if this install already registered a dashboard, we hold its
# client_id locally (HERMES_DASHBOARD_OAUTH_CLIENT_ID). Re-send it so the
# portal UPDATES that existing record instead of creating a duplicate. No
# stored client_id -> this is a first registration -> create a fresh one
# (the original behavior). This mirrors the portal's rule: no client id =
# new dashboard; client id present = the stable key of the row to modify.
existing_client_id = None
try:
existing_client_id = get_env_value("HERMES_DASHBOARD_OAUTH_CLIENT_ID")
except Exception:
existing_client_id = None
if isinstance(existing_client_id, str):
existing_client_id = existing_client_id.strip() or None
else:
existing_client_id = None
explicit_name = getattr(args, "name", None)
# Auto-generate a random name ONLY for a first registration. On a re-run
# (we hold a client_id) without an explicit --name, keep the name the
# portal already stored rather than churning it to a new random value
# every time — so leave `name` unset and let the portal preserve it.
if explicit_name:
name = explicit_name
elif existing_client_id:
name = None
else:
name = _generate_dashboard_name()
custom_redirect_uri = getattr(args, "redirect_uri", None)
# 2. Register with the portal.
@@ -255,20 +312,26 @@ def cmd_dashboard_register(args) -> None:
portal_base_url=portal_base_url,
name=name,
custom_redirect_uri=custom_redirect_uri,
existing_client_id=existing_client_id,
)
except RuntimeError as exc:
print(f"✗ Registration failed: {exc}")
sys.exit(1)
client_id = str(result["client_id"])
registered_name = str(result.get("name") or name)
registered_name = str(result.get("name") or name or "")
print(f'✓ Registered dashboard "{registered_name}"')
# Distinguish create vs update for the user: the portal echoes back the
# same client_id we sent when it updated in place.
updated_existing = bool(
existing_client_id and client_id == existing_client_id
)
if updated_existing:
print(f'✓ Updated dashboard "{registered_name}"')
else:
print(f'✓ Registered dashboard "{registered_name}"')
# 3. Write env vars idempotently. Always set the client_id. Only set the
# portal URL when it isn't already configured (env or config) AND differs
# from the production default, so we don't clutter .env for the common case
# but DO persist a non-default portal (e.g. a preview deploy used in dev).
# 3. Write env vars idempotently. Always set the client_id.
try:
save_env_value("HERMES_DASHBOARD_OAUTH_CLIENT_ID", client_id)
except Exception as exc:
@@ -276,6 +339,18 @@ def cmd_dashboard_register(args) -> None:
print(f" Set it manually: HERMES_DASHBOARD_OAUTH_CLIENT_ID={client_id}")
sys.exit(1)
# Persist the portal URL. Two cases:
# a) The user explicitly supplied a custom portal (--portal-url flag or
# HERMES_DASHBOARD_PORTAL_URL env). That's an intentional choice we
# always persist so it survives across sessions — overwriting any
# existing entry in place (save_env_value updates a matching key
# rather than appending a duplicate). This is true even when it equals
# the production default: the user asked for it explicitly.
# b) No custom portal was supplied. Keep the older conservative behaviour:
# only write a portal inferred from the stored login when it isn't
# already configured AND differs from the production default, so we
# don't clutter .env for the common production case and don't alter an
# existing entry unexpectedly.
wrote_portal_url = False
default_portal = "https://portal.nousresearch.com"
existing_portal = None
@@ -283,7 +358,15 @@ def cmd_dashboard_register(args) -> None:
existing_portal = get_env_value("HERMES_DASHBOARD_PORTAL_URL")
except Exception:
existing_portal = None
if not existing_portal and portal_base_url.rstrip("/") != default_portal:
if custom_portal_supplied:
should_write_portal = existing_portal != portal_base_url
else:
should_write_portal = (
not existing_portal and portal_base_url.rstrip("/") != default_portal
)
if should_write_portal:
try:
save_env_value("HERMES_DASHBOARD_PORTAL_URL", portal_base_url)
wrote_portal_url = True
@@ -291,10 +374,54 @@ def cmd_dashboard_register(args) -> None:
# Non-fatal: the client_id is the load-bearing value.
pass
# Persist the dashboard public URL derived from the OAuth redirect URI.
#
# --redirect-uri is the full public HTTPS callback the user registered with
# the portal, e.g. https://hermes.example.com/auth/callback. At serve time
# the dashboard auth layer (dashboard_auth/routes._redirect_uri) reconstructs
# that same callback by taking HERMES_DASHBOARD_PUBLIC_URL and appending
# "/auth/callback" verbatim. So the value the runtime actually consumes is
# the ORIGIN (scheme://host[:port]), not the full callback path — persisting
# the raw redirect URI would double up the path. We derive the origin from
# the supplied redirect URI and persist it as HERMES_DASHBOARD_PUBLIC_URL so
# the operator doesn't have to re-supply it and the public-URL override is
# actually wired (the gate engages and the callback round-trips correctly).
#
# Like the portal URL, an explicitly supplied value is always written
# (updating an existing entry in place rather than appending a duplicate),
# a no-op when it already matches, and never written on a localhost-only
# install (no --redirect-uri).
wrote_public_url = False
public_url = ""
if custom_redirect_uri:
try:
from urllib.parse import urlparse
parsed = urlparse(custom_redirect_uri)
if parsed.scheme in ("http", "https") and parsed.netloc:
public_url = f"{parsed.scheme}://{parsed.netloc}"
except Exception:
public_url = ""
if public_url:
existing_public_url = None
try:
existing_public_url = get_env_value("HERMES_DASHBOARD_PUBLIC_URL")
except Exception:
existing_public_url = None
if existing_public_url != public_url:
try:
save_env_value("HERMES_DASHBOARD_PUBLIC_URL", public_url)
wrote_public_url = True
except Exception:
# Non-fatal: the client_id is the load-bearing value.
pass
# 4. Hint.
_print_post_register_hint(
client_id=client_id,
portal_base_url=portal_base_url,
custom_redirect_uri=custom_redirect_uri,
wrote_portal_url=wrote_portal_url,
public_url=public_url if wrote_public_url else "",
)
+12 -4
View File
@@ -738,11 +738,14 @@ def run_doctor(args):
issues,
)
# Warn if model is set to a provider-prefixed name on a provider that doesn't use them
# Warn if model is set to a provider-prefixed name on a provider that doesn't use them.
# Vendor/model slugs are valid on aggregator-style providers and on any custom
# provider — bare "custom" or a named "custom:<name>" that fronts an OpenAI-compatible
# aggregator (e.g. custom:hpc-ai serving deepseek/deepseek-v4-flash) requires the prefix.
provider_for_policy = runtime_provider or catalog_provider
provider_policy_id = str(provider_for_policy or "").strip().lower()
providers_accepting_vendor_slugs = {
"openrouter",
"custom",
"auto",
"kilocode",
"opencode-zen",
@@ -750,11 +753,16 @@ def run_doctor(args):
"lmstudio",
"nous",
}
provider_accepts_vendor_slug = (
provider_policy_id in providers_accepting_vendor_slugs
or provider_policy_id == "custom"
or provider_policy_id.startswith("custom:")
)
if (
default_model
and "/" in default_model
and provider_for_policy
and provider_for_policy not in providers_accepting_vendor_slugs
and provider_policy_id
and not provider_accepts_vendor_slug
):
check_warn(
f"model.default '{default_model}' uses a vendor/model slug but provider is '{provider_raw}'",
+4 -1
View File
@@ -492,7 +492,10 @@ def get_label(provider_id: str) -> str:
def is_aggregator(provider: str) -> bool:
"""Return True when the provider is a multi-model aggregator."""
pdef = get_provider(provider)
provider_norm = normalize_provider(provider or "")
if provider_norm.startswith("custom:"):
return True
pdef = get_provider(provider_norm)
return pdef.is_aggregator if pdef else False
+94 -81
View File
@@ -1,123 +1,136 @@
# Photon iMessage platform plugin
This plugin connects Hermes Agent to iMessage (and WhatsApp Business +
future Spectrum interfaces) through [Photon][photon] — a managed
service that handles the iMessage line allocation, delivery, and
abuse-prevention layer so users don't have to run their own Mac
relay.
This plugin connects Hermes Agent to iMessage (and other Spectrum
interfaces) through [Photon][photon] — a managed service that handles
iMessage line allocation, delivery, and abuse-prevention so users don't
have to run their own Mac relay.
The free tier uses Photon's shared iMessage line pool (`type: shared`)
and is the path we recommend for everyone who doesn't already pay for a
dedicated number.
The free tier uses Photon's shared iMessage line pool and is the path we
recommend for everyone who doesn't already pay for a dedicated number.
## Architecture
Like Discord and Slack, Photon is a **persistent-connection** channel — no
public URL, no webhook, no signing secret. The `spectrum-ts` SDK holds a
long-lived **gRPC stream** to Photon for both directions. Because the SDK is
TypeScript-only, Hermes runs it inside a small supervised Node sidecar and
talks to it over loopback.
```
┌─────────────────────────┐ HMAC-signed POSTs ┌──────────────────┐
│ Photon Spectrum cloud │ ──────────────────────► │ Hermes Agent │
(iMessage line owner) │ │ (Python)
└─────────────────────────┘ JSON over loopback │
◄────────────────────── │ PhotonAdapter │
│ │ + aiohttp recv │
│ spectrum-ts
│ SDK (Node) │ spawns + super- │
vises ▼
┌─────────────────────────┐ ├──────────────────┤
│ Node sidecar │ ◄──── X-Hermes- ─ │ Node sidecar │
│ (plugins/.../sidecar) │ Sidecar-Token │ child process │
└─────────────────────────┘ └──────────────────┘
gRPC (spectrum-ts)
┌─────────────────────────┐ ◄───────────────► ┌──────────────────────┐
Photon Spectrum cloud app.messages │ Node sidecar
│ (iMessage line owner) │ space.send() (plugins/…/sidecar)
└─────────────────────────┘ ─────────────────────
GET /inbound (NDJSON) │ ▲ POST /send
inbound events /typing
┌──────────────────────┐
PhotonAdapter
│ (Python, in gateway) │
└──────────────────────┘
```
Inbound traffic is webhook-only — Hermes runs an aiohttp listener
that verifies `X-Spectrum-Signature` and dedupes on `message.id`.
Outbound traffic goes through a tiny Node sidecar that runs the
`spectrum-ts` SDK. Photon does not currently expose an HTTP
send-message endpoint; their own docs say:
> Pass `space.id` to `Space.send(...)` from a separate `spectrum-ts`
> SDK instance to reply. **No public HTTP send endpoint exists today.**
> — https://photon.codes/docs/webhooks/events
When Photon ships an HTTP send endpoint, `_sidecar_send` is the one
function that swaps and the sidecar disappears. The rest of the
plugin stays the same.
- **Inbound**: the sidecar consumes the SDK's `app.messages` gRPC stream,
normalizes each message, and streams it to the adapter over a loopback
`GET /inbound` (NDJSON). The adapter dedupes on `messageId` and dispatches
a `MessageEvent` to the gateway. It reconnects automatically if the stream
drops; the sidecar owns the gRPC reconnect to Photon.
- **Outbound**: `send` / `send_typing` are loopback POSTs to the sidecar,
authenticated with a shared `X-Hermes-Sidecar-Token`.
## First-time setup
```bash
# 1. One-shot setup: device login (opens browser) + project + user + sidecar deps
# One-shot setup: device login (opens browser) + project + user + sidecar deps
hermes photon setup --phone +15551234567
# 2. Expose your webhook URL to the public internet
# (cloudflared, ngrok, your gateway's public hostname, etc.)
# Then register it with Photon:
hermes photon webhook register https://your-host.example.com/photon/webhook
# 3. Save the signing secret it prints to ~/.hermes/.env
# as PHOTON_WEBHOOK_SECRET=...
# Photon only returns it ONCE.
# 4. Start the gateway
# Start the gateway
hermes gateway start --platform photon
```
`hermes photon setup` runs the RFC 8628 device-code login as its first
step — it opens `https://app.photon.codes/` for approval, then
provisions the Spectrum project + iMessage line. There is no separate
`login` command; like every other Hermes channel, onboarding goes
through one setup surface. Re-running `setup` reuses an existing token
and project, so it's safe to run again to finish a partial setup.
`hermes photon setup` does, in order:
1. **Device login** (RFC 8628, `client_id=photon-cli`) — opens
`https://app.photon.codes/` for approval and stores the bearer token.
2. **Find or create** the `Hermes Agent` project on the Photon dashboard.
3. **Enable Spectrum**, read the project's `spectrumProjectId`, rotate the
project secret, and persist both.
4. **Register your phone number** as a Spectrum user (idempotent — skipped if
a user with that number already exists).
5. **Print the assigned iMessage line** — the number you text to reach your
agent.
6. **Install the sidecar deps** (`spectrum-ts`).
There is no separate `login` command; like every other Hermes channel,
onboarding goes through one setup surface. Re-running `setup` reuses an
existing token/project, so it's safe to run again to finish a partial setup.
Run `hermes photon status` to see what's configured.
## Credentials
Stored in `~/.hermes/auth.json` under `credential_pool`:
Runtime SDK credentials live in `~/.hermes/.env` (the same place every other
channel keeps its token), and the adapter reads them from the environment:
```bash
PHOTON_PROJECT_ID=<spectrumProjectId> # the SDK's projectId
PHOTON_PROJECT_SECRET=<projectSecret>
```
Management metadata lives in `~/.hermes/auth.json` under `credential_pool`:
```jsonc
{
"credential_pool": {
"photon": [
{ "access_token": "<dashboard-bearer>", "issued_at": ... }
{ "access_token": "<device-bearer>", "issued_at": ... }
],
"photon_project": [
{ "project_id": "...", "project_secret": "...", "name": "Hermes Agent" }
{
"dashboard_project_id": "<dashboard id>",
"spectrum_project_id": "<spectrumProjectId>",
"project_secret": "<projectSecret>",
"name": "Hermes Agent"
}
]
}
}
```
The per-URL webhook signing secret is treated like an API key and
lives in `~/.hermes/.env` as `PHOTON_WEBHOOK_SECRET`.
> **Note on ids.** A Photon project has two identifiers: the dashboard `id`
> (used for management API calls) and the `spectrumProjectId` (what the SDK
> authenticates with). `PHOTON_PROJECT_ID` is the **spectrum** id.
## Configuration knobs
All env vars are documented in `plugin.yaml`. The most important are:
All env vars are documented in `plugin.yaml`. The most important:
| Env var | Default | Meaning |
|--------------------------|--------------------|-----------------------------------------|
| `PHOTON_PROJECT_ID` | from auth.json | Spectrum project ID |
| `PHOTON_PROJECT_SECRET` | from auth.json | Spectrum project secret (HTTP Basic) |
| `PHOTON_WEBHOOK_SECRET` | (unset) | Signing secret returned at register |
| `PHOTON_WEBHOOK_PORT` | 8788 | Local port for the aiohttp listener |
| `PHOTON_WEBHOOK_PATH` | /photon/webhook | Path under which the listener mounts |
| `PHOTON_SIDECAR_PORT` | 8789 | Loopback port for sidecar control |
| `PHOTON_HOME_CHANNEL` | (unset) | Default space ID for cron delivery |
| `PHOTON_ALLOWED_USERS` | (unset) | Comma-separated E.164 allowlist |
| Env var | Default | Meaning |
|---------------------------|----------------------------|--------------------------------------|
| `PHOTON_PROJECT_ID` | from .env / auth.json | Spectrum project id (SDK `projectId`)|
| `PHOTON_PROJECT_SECRET` | from .env / auth.json | Project secret |
| `PHOTON_SIDECAR_PORT` | 8789 | Loopback port for the sidecar |
| `PHOTON_SIDECAR_AUTOSTART`| true | Spawn the sidecar on connect |
| `PHOTON_DASHBOARD_HOST` | https://app.photon.codes | Dashboard API host |
| `PHOTON_HOME_CHANNEL` | your number (set by setup) | Default space for cron delivery — a space id, or a bare E.164 number (resolved to a DM) |
| `PHOTON_ALLOWED_USERS` | your number (set by setup) | Comma-separated E.164 allowlist |
| `PHOTON_REQUIRE_MENTION` | false | Gate group chats on a wake word |
| `PHOTON_MAX_INLINE_ATTACHMENT_BYTES` | 20 MB | Max inbound attachment size the sidecar reads & inlines |
## Limitations (current Photon API)
## Attachments & limitations
- **Inbound attachments are metadata only.** Inbound webhooks include the
filename + MIME type but no download URL. The plugin surfaces a
text marker (`[Photon attachment received: …]`) so the agent knows
something arrived, but cannot read the bytes. Photon's docs note
an attachment retrieval endpoint is on the roadmap.
- **Outbound attachments are supported.** Images, voice notes, video,
and documents are sent via `space.send(attachment(...))` /
- **Inbound attachments are downloaded.** The sidecar reads the bytes
(`content.read()`) and base64-inlines them on the NDJSON event; the adapter
caches them to the shared media cache and populates `media_urls` /
`media_types`, so the agent sees the real image/file (vision included) —
parity with the BlueBubbles iMessage channel. Attachments larger than
`PHOTON_MAX_INLINE_ATTACHMENT_BYTES` (default 20 MB), or any byte read that
fails, fall back to a text marker (`[Photon attachment received: …]`) so the
agent still knows something arrived.
- **Outbound attachments are supported.** Images, voice notes, video, and
documents are sent via `space.send(attachment(...))` /
`space.send(voice(...))` through the sidecar's `/send-attachment`
endpoint. A caption is delivered as a separate text bubble after the
media.
- **Reactions, message effects, polls** — not exposed yet; the
`spectrum-ts` SDK supports them, and the sidecar is the natural
place to add them when the agent has reason to use them.
endpoint; a caption is delivered as a separate text bubble after the media.
- **Reactions, message effects, polls** — supported by `spectrum-ts` but not
yet exposed; the sidecar is the natural place to add them.
[photon]: https://photon.codes/
+286 -245
View File
@@ -1,32 +1,30 @@
"""
Photon Spectrum (iMessage) platform adapter for Hermes Agent.
Both directions of traffic flow through a small supervised Node sidecar
(see ``sidecar/index.mjs``) that runs the ``spectrum-ts`` SDK the SDK is
TypeScript-only and there is no public HTTP message API, so a sidecar is
unavoidable.
Inbound:
Photon delivers signed JSON ``POST``s to a URL we register. The
adapter spins up an aiohttp server on ``PHOTON_WEBHOOK_PORT``,
verifies ``X-Spectrum-Signature`` (HMAC-SHA256 of
``v0:{timestamp}:{body}`` keyed by the per-URL signing secret),
rejects deliveries with a timestamp drift > 5 minutes, dedupes on
``message.id``, and dispatches a normalized ``MessageEvent`` to the
gateway runner via ``BasePlatformAdapter.handle_message``.
The SDK's ``app.messages`` is a long-lived **gRPC** stream. The sidecar
serializes each message to a normalized JSON event and streams it to this
adapter over a loopback ``GET /inbound`` (NDJSON). A background task here
consumes that stream, dedupes on ``messageId``, and dispatches a
``MessageEvent`` to the gateway via ``BasePlatformAdapter.handle_message``.
No webhook, no public URL, no signing secret.
Outbound:
Photon does not currently expose a public HTTP send-message
endpoint, so the adapter spawns a small Node sidecar (see
``sidecar/index.mjs``) that runs the ``spectrum-ts`` SDK. Each
``send`` / ``send_typing`` / attachment call from Hermes is a
loopback POST to the sidecar with a shared bearer token. Outbound
media (images, voice notes, video, documents) goes through
spectrum-ts' ``attachment()`` / ``voice()`` content builders.
When Photon ships an HTTP send endpoint we can collapse the sidecar
into ``_send_via_http`` and drop the Node dependency entirely.
``send`` / ``send_typing`` are loopback POSTs to the sidecar's control
endpoints, authenticated with a shared bearer token. Outbound media
(images, voice notes, video, documents) goes through spectrum-ts'
``attachment()`` / ``voice()`` content builders via the sidecar's
``/send-attachment`` endpoint.
"""
from __future__ import annotations
import asyncio
import hashlib
import hmac
import base64
import json
import logging
import os
@@ -39,21 +37,21 @@ import sys
import time
from datetime import datetime, timezone
from pathlib import Path
from typing import Any, Dict, Optional
from typing import TYPE_CHECKING, Any, Dict, List, Optional
try:
if TYPE_CHECKING:
# Type checkers see ``httpx`` as the always-imported module, so every use
# site type-checks cleanly. The runtime fallback below keeps the optional
# dependency truly optional (each use site is guarded by HTTPX_AVAILABLE).
import httpx
HTTPX_AVAILABLE = True
except ImportError: # pragma: no cover - httpx is already a Hermes dep
HTTPX_AVAILABLE = False
httpx = None # type: ignore[assignment]
try:
from aiohttp import web
AIOHTTP_AVAILABLE = True
except ImportError:
AIOHTTP_AVAILABLE = False
web = None # type: ignore[assignment]
else:
try:
import httpx
HTTPX_AVAILABLE = True
except ImportError: # pragma: no cover - httpx is already a Hermes dep
HTTPX_AVAILABLE = False
httpx = None
from gateway.config import Platform, PlatformConfig
from gateway.platforms.base import (
@@ -63,21 +61,13 @@ from gateway.platforms.base import (
SendResult,
)
from .auth import (
DEFAULT_SPECTRUM_HOST,
load_project_credentials,
_spectrum_host,
)
from .auth import load_project_credentials
logger = logging.getLogger(__name__)
# ---------------------------------------------------------------------------
# Constants
_DEFAULT_WEBHOOK_PORT = 8788
_DEFAULT_WEBHOOK_PATH = "/photon/webhook"
_DEFAULT_WEBHOOK_BIND = "0.0.0.0"
_DEFAULT_SIDECAR_PORT = 8789
_DEFAULT_SIDECAR_BIND = "127.0.0.1"
@@ -86,11 +76,8 @@ _DEFAULT_SIDECAR_BIND = "127.0.0.1"
# size to ~16 KB. Keep a conservative cap that matches BlueBubbles.
_MAX_MESSAGE_LENGTH = 8000
# Spec says reject deliveries older than ~5 minutes for replay protection.
_TIMESTAMP_DRIFT_SECONDS = 300
# Dedup parameters — keep at least 1k IDs for ~48h per Photon's
# at-least-once guidance.
# Dedup parameters — the gRPC stream is at-least-once, and a sidecar
# reconnect can replay, so keep at least 1k ids for ~48h.
_DEDUP_MAX_SIZE = 4000
_DEDUP_WINDOW_SECONDS = 48 * 3600
@@ -118,7 +105,7 @@ def _coerce_port(value: Any, default: int) -> int:
def check_requirements() -> bool:
"""Return True when both Python deps and the Node sidecar are available."""
if not HTTPX_AVAILABLE or not AIOHTTP_AVAILABLE:
if not HTTPX_AVAILABLE:
return False
if not shutil.which(os.getenv("PHOTON_NODE_BIN") or "node"):
return False
@@ -146,61 +133,33 @@ def is_connected(cfg: PlatformConfig) -> bool:
def _env_enablement() -> Optional[dict]:
"""Seed PlatformConfig.extra from env so env-only setups appear in status."""
"""Seed PlatformConfig.extra from env so env-only setups appear in status.
The special ``home_channel`` key is handled by the core plugin hook and
becomes a proper ``HomeChannel`` on ``PlatformConfig``.
"""
project_id, project_secret = load_project_credentials()
if not (project_id and project_secret):
return None
return {
"project_id": project_id,
"project_secret": project_secret,
"webhook_port": _coerce_port(os.getenv("PHOTON_WEBHOOK_PORT"), _DEFAULT_WEBHOOK_PORT),
"webhook_path": os.getenv("PHOTON_WEBHOOK_PATH") or _DEFAULT_WEBHOOK_PATH,
}
# ---------------------------------------------------------------------------
# Signature verification
def verify_signature(
*,
body: bytes,
timestamp_header: str,
signature_header: str,
signing_secret: str,
now: Optional[float] = None,
drift: int = _TIMESTAMP_DRIFT_SECONDS,
) -> bool:
"""Constant-time verify a Photon webhook signature.
Returns True iff the timestamp is within ``drift`` of *now* AND
``signature_header == "v0=" + hmac_sha256(secret, "v0:{ts}:{body}")``.
Exposed at module scope so tests can exercise it without an adapter
instance.
"""
if not timestamp_header or not signature_header or not signing_secret:
return False
try:
ts = int(timestamp_header)
except ValueError:
return False
if abs((now or time.time()) - ts) > drift:
return False
if not signature_header.startswith("v0="):
return False
expected = hmac.new(
signing_secret.encode("utf-8"),
f"v0:{ts}:".encode("utf-8") + body,
hashlib.sha256,
).hexdigest()
return hmac.compare_digest(expected, signature_header[3:])
seed = {"project_id": project_id, "project_secret": project_secret}
home = os.getenv("PHOTON_HOME_CHANNEL", "").strip()
if home:
seed["home_channel"] = {
"chat_id": home,
"name": os.getenv("PHOTON_HOME_CHANNEL_NAME", "Home"),
}
return seed
# ---------------------------------------------------------------------------
# Adapter
class PhotonAdapter(BasePlatformAdapter):
"""Inbound: signed webhook on aiohttp. Outbound: Node sidecar via loopback HTTP."""
"""Bidirectional bridge to Photon Spectrum via the Node spectrum-ts sidecar.
Inbound: consume the sidecar's ``/inbound`` gRPC stream.
Outbound: loopback POSTs to the sidecar's control channel.
"""
MAX_MESSAGE_LENGTH = _MAX_MESSAGE_LENGTH
@@ -209,6 +168,8 @@ class PhotonAdapter(BasePlatformAdapter):
extra = config.extra or {}
# Project credentials (env wins, then config.extra, then auth.json).
# ``project_id`` here is the project's spectrumProjectId — the value
# the spectrum-ts SDK authenticates with.
stored_id, stored_sec = load_project_credentials()
self._project_id: str = (
os.getenv("PHOTON_PROJECT_ID")
@@ -223,27 +184,6 @@ class PhotonAdapter(BasePlatformAdapter):
or ""
)
# Webhook receiver
self._webhook_port = _coerce_port(
extra.get("webhook_port") or os.getenv("PHOTON_WEBHOOK_PORT"),
_DEFAULT_WEBHOOK_PORT,
)
self._webhook_path = (
extra.get("webhook_path")
or os.getenv("PHOTON_WEBHOOK_PATH")
or _DEFAULT_WEBHOOK_PATH
)
self._webhook_bind = (
extra.get("webhook_bind")
or os.getenv("PHOTON_WEBHOOK_BIND")
or _DEFAULT_WEBHOOK_BIND
)
self._webhook_secret: str = (
os.getenv("PHOTON_WEBHOOK_SECRET")
or extra.get("webhook_secret")
or ""
)
# Sidecar
self._sidecar_port = _coerce_port(
extra.get("sidecar_port") or os.getenv("PHOTON_SIDECAR_PORT"),
@@ -259,12 +199,13 @@ class PhotonAdapter(BasePlatformAdapter):
self._node_bin = os.getenv("PHOTON_NODE_BIN") or shutil.which("node") or "node"
# Runtime state
self._runner: Optional["web.AppRunner"] = None
self._sidecar_proc: Optional[subprocess.Popen] = None
self._sidecar_supervisor_task: Optional[asyncio.Task] = None
self._inbound_task: Optional[asyncio.Task] = None
self._inbound_running = False
self._http_client: Optional["httpx.AsyncClient"] = None
# Lightweight in-memory dedup. Photon's at-least-once guarantee
# means we WILL see the same message.id more than once.
# Lightweight in-memory dedup. The gRPC stream is at-least-once, so we
# may see the same messageId more than once (e.g. after a reconnect).
self._seen_messages: Dict[str, float] = {}
# Group-chat mention gating (parity with BlueBubbles). When enabled,
@@ -345,13 +286,6 @@ class PhotonAdapter(BasePlatformAdapter):
# -- Connection lifecycle ---------------------------------------------
async def connect(self) -> bool:
if not AIOHTTP_AVAILABLE:
self._set_fatal_error(
"MISSING_DEP",
"aiohttp not installed. Run: pip install aiohttp",
retryable=False,
)
return False
if not HTTPX_AVAILABLE:
self._set_fatal_error(
"MISSING_DEP", "httpx not installed", retryable=False
@@ -366,19 +300,11 @@ class PhotonAdapter(BasePlatformAdapter):
)
return False
# Start the aiohttp receiver first; without it the sidecar would
# be able to forward inbound traffic to a closed port.
try:
await self._start_webhook_server()
except OSError as e:
self._set_fatal_error(
"PORT_IN_USE",
f"webhook port {self._webhook_port} unavailable: {e}",
retryable=True,
)
return False
client = httpx.AsyncClient(timeout=30.0)
self._http_client = client
# Spin up the Node sidecar (required for outbound).
# The sidecar holds the gRPC stream for BOTH directions, so it is
# required now (not just for outbound).
if self._autostart_sidecar:
try:
await self._start_sidecar()
@@ -388,23 +314,39 @@ class PhotonAdapter(BasePlatformAdapter):
f"failed to start Photon sidecar: {e}",
retryable=True,
)
await self._stop_webhook_server()
await client.aclose()
self._http_client = None
return False
else:
logger.info("[photon] sidecar autostart disabled — outbound will fail")
logger.warning(
"[photon] sidecar autostart disabled — inbound + outbound will fail"
)
# Start consuming the inbound gRPC stream from the sidecar.
self._inbound_running = True
self._inbound_task = asyncio.get_event_loop().create_task(
self._inbound_loop()
)
self._http_client = httpx.AsyncClient(timeout=30.0)
self._mark_connected()
logger.info(
"[photon] connected — webhook at %s:%d%s, sidecar on %s:%d",
self._webhook_bind, self._webhook_port, self._webhook_path,
"[photon] connected — sidecar on %s:%d, streaming inbound over gRPC",
self._sidecar_bind, self._sidecar_port,
)
return True
async def disconnect(self) -> None:
self._inbound_running = False
if self._inbound_task is not None:
self._inbound_task.cancel()
try:
await self._inbound_task
except asyncio.CancelledError:
pass
except Exception:
pass
self._inbound_task = None
await self._stop_sidecar()
await self._stop_webhook_server()
if self._http_client is not None:
try:
await self._http_client.aclose()
@@ -413,68 +355,61 @@ class PhotonAdapter(BasePlatformAdapter):
self._http_client = None
self._mark_disconnected()
# -- Webhook server ----------------------------------------------------
# -- Inbound stream consumer ------------------------------------------
async def _start_webhook_server(self) -> None:
app = web.Application()
app.router.add_post(self._webhook_path, self._handle_webhook)
app.router.add_get("/healthz", lambda _: web.Response(text="ok"))
self._runner = web.AppRunner(app)
await self._runner.setup()
site = web.TCPSite(self._runner, self._webhook_bind, self._webhook_port)
await site.start()
async def _inbound_loop(self) -> None:
"""Consume the sidecar's ``/inbound`` NDJSON stream, with reconnect.
async def _stop_webhook_server(self) -> None:
if self._runner is not None:
The sidecar owns the gRPC reconnect/heartbeat to Photon; this loop
only has to re-open the loopback HTTP stream if it drops (e.g. the
sidecar restarts).
"""
client = self._http_client
if client is None:
return
url = f"http://{self._sidecar_bind}:{self._sidecar_port}/inbound"
headers = {"X-Hermes-Sidecar-Token": self._sidecar_token}
backoff = 1.0
while self._inbound_running:
try:
await self._runner.cleanup()
except Exception:
pass
self._runner = None
async def _handle_webhook(self, request: "web.Request") -> "web.Response":
body = await request.read()
if self._webhook_secret:
ts = request.headers.get("X-Spectrum-Timestamp", "")
sig = request.headers.get("X-Spectrum-Signature", "")
if not verify_signature(
body=body,
timestamp_header=ts,
signature_header=sig,
signing_secret=self._webhook_secret,
):
logger.warning("[photon] rejected webhook with bad signature")
return web.Response(status=401, text="invalid signature")
else:
logger.warning(
"[photon] PHOTON_WEBHOOK_SECRET unset — accepting unsigned "
"deliveries. Set the per-URL signing secret returned by "
"register-webhook to enable verification."
)
async with client.stream(
"GET", url, headers=headers, timeout=None,
) as resp:
if resp.status_code != 200:
raise RuntimeError(f"/inbound returned {resp.status_code}")
backoff = 1.0 # reset on a successful connect
async for line in resp.aiter_lines():
if not self._inbound_running:
break
line = line.strip()
if not line:
continue # heartbeat
await self._on_inbound_line(line)
except asyncio.CancelledError:
raise
except Exception as e:
if not self._inbound_running:
break
logger.warning(
"[photon] inbound stream dropped (%s); reconnecting in %.1fs",
e, backoff,
)
await asyncio.sleep(backoff)
backoff = min(backoff * 2, 30.0)
async def _on_inbound_line(self, line: str) -> None:
try:
payload = json.loads(body or b"{}")
event = json.loads(line)
except json.JSONDecodeError:
return web.Response(status=400, text="invalid json")
if payload.get("event") != "messages":
# Photon currently emits only `messages`; any future event
# types are ack'd 200 so they don't retry.
return web.Response(text="ok")
msg = payload.get("message") or {}
msg_id = msg.get("id")
if not msg_id:
return web.Response(status=400, text="missing message.id")
if self._is_duplicate(msg_id):
return web.Response(text="ok (dup)")
logger.debug("[photon] skipping non-JSON inbound line")
return
msg_id = event.get("messageId")
if msg_id and self._is_duplicate(msg_id):
return
try:
await self._dispatch_inbound(payload)
await self._dispatch_inbound(event)
except Exception:
logger.exception("[photon] inbound dispatch failed")
# 200 anyway — we own the dedup; failing here would cause
# Photon to retry the same id.
return web.Response(text="ok")
def _is_duplicate(self, msg_id: str) -> bool:
now = time.time()
@@ -488,44 +423,77 @@ class PhotonAdapter(BasePlatformAdapter):
self._seen_messages[msg_id] = now
return False
async def _dispatch_inbound(self, payload: Dict[str, Any]) -> None:
msg = payload.get("message") or {}
space = msg.get("space") or payload.get("space") or {}
sender = msg.get("sender") or {}
content = msg.get("content") or {}
async def _dispatch_inbound(self, event: Dict[str, Any]) -> None:
"""Normalize a sidecar inbound event and dispatch it to the gateway.
Event shape (from ``sidecar/index.mjs``)::
{
"messageId": "...",
"platform": "iMessage",
"space": {"id": "...", "type": "dm"|"group", "phone": "+E164"},
"sender": {"id": "+E164"},
"content": {"type": "text", "text": "..."}
| {"type": "attachment", "id", "name", "mimeType",
"size", "data"?, "encoding"?},
"timestamp": "2026-05-14T19:06:32.000Z"
Attachment content carries the bytes inline as base64 ``data`` (with
``encoding == "base64"``) when the sidecar could read them within its
size cap; otherwise only metadata is present and we surface a marker.
}
"""
space = event.get("space") or {}
sender = event.get("sender") or {}
content = event.get("content") or {}
space_id = space.get("id") or ""
sender_id = sender.get("id") or ""
if not space_id:
logger.warning("[photon] inbound missing space.id")
return
# Space type — Photon documents iMessage DM ids as `any;-;+E164`
# and group ids as `any;+;<chat-guid>`. Use that as the
# heuristic; everything else is treated as DM.
chat_type = "group" if ";+;" in space_id else "dm"
# iMessage spaces carry their type directly — no id string-sniffing.
chat_type = "group" if space.get("type") == "group" else "dm"
sender_id = sender.get("id") or space.get("phone") or space_id
# Timestamp — ISO 8601 from the platform.
ts_str = msg.get("timestamp") or ""
ts_str = event.get("timestamp") or ""
try:
timestamp = datetime.fromisoformat(ts_str.replace("Z", "+00:00"))
timestamp = (
datetime.fromisoformat(ts_str.replace("Z", "+00:00"))
if ts_str
else datetime.now(tz=timezone.utc)
)
except ValueError:
timestamp = datetime.now(tz=timezone.utc)
# Content normalization. Spectrum is a discriminated union;
# text vs attachment metadata. Attachments are metadata-only
# today (no download URL) — log + carry the name so the agent
# at least knows something was sent.
if content.get("type") == "text":
# Media attachments (local cached paths) handed to the agent via the
# gateway's image-routing path, exactly like the BlueBubbles channel.
media_urls: List[str] = []
media_types: List[str] = []
ctype = content.get("type")
if ctype == "text":
text = content.get("text") or ""
mtype = MessageType.TEXT
elif content.get("type") == "attachment":
elif ctype == "attachment":
name = content.get("name") or "(unnamed)"
mime = content.get("mimeType") or ""
text = f"[Photon attachment received: {name} ({mime}) — no download URL yet]"
mtype = _attachment_message_type(mime)
cached = _cache_inbound_attachment(content, name, mime)
if cached:
media_urls.append(cached)
media_types.append(mime or "application/octet-stream")
# The real bytes are attached, so the agent sees the media
# itself — a short marker is enough text, and it keeps group
# mention-gating consistent with plain messages.
text = "(attachment)"
else:
# No bytes (over the sidecar cap, a failed read, or a caching
# failure) — fall back to a metadata marker so the agent still
# knows something arrived.
text = f"[Photon attachment received: {name} ({mime})]"
else:
text = f"[Photon content type not handled: {content.get('type')}]"
text = f"[Photon content type not handled: {ctype}]"
mtype = MessageType.TEXT
# Group-mention gating (parity with BlueBubbles). In group chats with
@@ -545,18 +513,20 @@ class PhotonAdapter(BasePlatformAdapter):
chat_id=space_id,
chat_name=space_id,
chat_type=chat_type,
user_id=sender_id or space_id,
user_id=sender_id,
user_name=sender_id or None,
)
event = MessageEvent(
message_event = MessageEvent(
text=text,
message_type=mtype,
source=source,
message_id=msg.get("id"),
raw_message=payload,
message_id=event.get("messageId"),
raw_message=event,
timestamp=timestamp,
media_urls=media_urls,
media_types=media_types,
)
await self.handle_message(event)
await self.handle_message(message_event)
# -- Sidecar lifecycle -------------------------------------------------
@@ -670,7 +640,7 @@ class PhotonAdapter(BasePlatformAdapter):
reply_to: Optional[str] = None,
metadata: Optional[Dict[str, Any]] = None,
) -> SendResult:
return await self._sidecar_send(chat_id, content, reply_to=reply_to)
return await self._sidecar_send(chat_id, content)
# -- Outbound media (parity with the BlueBubbles iMessage channel) -----
#
@@ -696,7 +666,7 @@ class PhotonAdapter(BasePlatformAdapter):
# Couldn't fetch the URL — fall back to sending it as text.
return await super().send_image(chat_id, image_url, caption, reply_to)
return await self._sidecar_send_attachment(
chat_id, local_path, caption=caption, reply_to=reply_to,
chat_id, local_path, caption=caption,
)
async def send_image_file(
@@ -709,7 +679,7 @@ class PhotonAdapter(BasePlatformAdapter):
**kwargs,
) -> SendResult:
return await self._sidecar_send_attachment(
chat_id, image_path, caption=caption, reply_to=reply_to,
chat_id, image_path, caption=caption,
)
async def send_voice(
@@ -722,7 +692,7 @@ class PhotonAdapter(BasePlatformAdapter):
**kwargs,
) -> SendResult:
return await self._sidecar_send_attachment(
chat_id, audio_path, caption=caption, reply_to=reply_to, kind="voice",
chat_id, audio_path, caption=caption, kind="voice",
)
async def send_video(
@@ -735,7 +705,7 @@ class PhotonAdapter(BasePlatformAdapter):
**kwargs,
) -> SendResult:
return await self._sidecar_send_attachment(
chat_id, video_path, caption=caption, reply_to=reply_to,
chat_id, video_path, caption=caption,
)
async def send_document(
@@ -749,7 +719,7 @@ class PhotonAdapter(BasePlatformAdapter):
**kwargs,
) -> SendResult:
return await self._sidecar_send_attachment(
chat_id, file_path, name=file_name, caption=caption, reply_to=reply_to,
chat_id, file_path, name=file_name, caption=caption,
)
async def send_animation(
@@ -767,23 +737,29 @@ class PhotonAdapter(BasePlatformAdapter):
async def send_typing(self, chat_id: str, metadata=None) -> None:
try:
await self._sidecar_call("/typing", {"spaceId": chat_id})
await self._sidecar_call(
"/typing", {"spaceId": chat_id, "state": "start"}
)
except Exception as e:
logger.debug("[photon] send_typing failed: %s", e)
async def stop_typing(self, chat_id: str) -> None:
try:
await self._sidecar_call(
"/typing", {"spaceId": chat_id, "state": "stop"}
)
except Exception as e:
logger.debug("[photon] stop_typing failed: %s", e)
async def get_chat_info(self, chat_id: str) -> Dict[str, Any]:
"""Return whatever we know about a Spectrum space id.
Photon's `space.id` is opaque (`any;-;+E164` for DMs,
`any;+;<guid>` for groups). We surface that shape directly so
the gateway has something to show in session pickers / logs.
Photon's ``space.id`` is opaque; the inbound event also carries the
DM/group type, but here we only have the id, so infer conservatively.
"""
chat_type = "group" if ";+;" in chat_id else "dm"
return {"name": chat_id, "type": chat_type, "id": chat_id}
return {"name": chat_id, "type": "dm", "id": chat_id}
async def _sidecar_send(
self, space_id: str, text: str, *, reply_to: Optional[str] = None,
) -> SendResult:
async def _sidecar_send(self, space_id: str, text: str) -> SendResult:
if len(text) > self.MAX_MESSAGE_LENGTH:
logger.warning(
"[photon] truncating outbound from %d to %d chars",
@@ -791,8 +767,6 @@ class PhotonAdapter(BasePlatformAdapter):
)
text = text[: self.MAX_MESSAGE_LENGTH]
body: Dict[str, Any] = {"spaceId": space_id, "text": text}
if reply_to:
body["replyTo"] = reply_to
try:
data = await self._sidecar_call("/send", body)
except Exception as e:
@@ -807,7 +781,6 @@ class PhotonAdapter(BasePlatformAdapter):
name: Optional[str] = None,
mime_type: Optional[str] = None,
caption: Optional[str] = None,
reply_to: Optional[str] = None,
kind: str = "attachment",
) -> SendResult:
"""POST a local file to the sidecar's ``/send-attachment`` endpoint.
@@ -842,8 +815,6 @@ class PhotonAdapter(BasePlatformAdapter):
body["mimeType"] = mime_type
if caption:
body["caption"] = caption
if reply_to:
body["replyTo"] = reply_to
try:
data = await self._sidecar_call("/send-attachment", body)
except Exception as e:
@@ -887,11 +858,81 @@ def _attachment_message_type(mime: str) -> MessageType:
return MessageType.DOCUMENT
# MIME → file-extension maps for caching inbound attachment bytes. These mirror
# the BlueBubbles iMessage channel so both adapters name cached media the same.
_IMAGE_EXT_BY_MIME = {
"image/jpeg": ".jpg",
"image/png": ".png",
"image/gif": ".gif",
"image/webp": ".webp",
"image/heic": ".jpg",
"image/heif": ".jpg",
"image/tiff": ".jpg",
}
_AUDIO_EXT_BY_MIME = {
"audio/mp3": ".mp3",
"audio/mpeg": ".mp3",
"audio/ogg": ".ogg",
"audio/wav": ".wav",
"audio/x-caf": ".mp3",
"audio/mp4": ".m4a",
"audio/aac": ".m4a",
}
def _cache_inbound_attachment(
content: Dict[str, Any], name: str, mime: str
) -> Optional[str]:
"""Decode a base64-inlined inbound attachment and cache it locally.
The sidecar inlines the attachment bytes as ``content["data"]`` (base64).
We decode them and route to the shared media cache by MIME type, returning
the cached absolute path so the caller can populate ``media_urls`` (which
the gateway then hands to the model). Returns ``None`` when there are no
bytes (over the sidecar's inline cap or a failed read) or when caching
fails, so the caller can fall back to a text marker.
"""
data_b64 = content.get("data")
if not data_b64:
return None
try:
raw = base64.b64decode(data_b64)
except (ValueError, TypeError) as exc:
logger.warning("[photon] failed to decode inbound attachment bytes: %s", exc)
return None
from gateway.platforms.base import (
cache_audio_from_bytes,
cache_document_from_bytes,
cache_image_from_bytes,
)
mime = (mime or "").lower()
# Prefer the real extension from the filename; fall back to the MIME map.
suffix = Path(name).suffix if name else ""
try:
if mime.startswith("image/"):
ext = suffix or _IMAGE_EXT_BY_MIME.get(mime, ".jpg")
try:
return cache_image_from_bytes(raw, ext)
except ValueError:
# Bytes don't look like a supported image (e.g. HEIC magic) —
# still deliver them as a document rather than dropping them.
return cache_document_from_bytes(raw, name)
if mime.startswith("audio/"):
ext = suffix or _AUDIO_EXT_BY_MIME.get(mime, ".mp3")
return cache_audio_from_bytes(raw, ext)
# Video, application/*, and everything else → document cache.
return cache_document_from_bytes(raw, name)
except Exception as exc:
logger.warning("[photon] failed to cache inbound attachment %s: %s", name, exc)
return None
# ---------------------------------------------------------------------------
# Standalone (out-of-process) send for cron deliveries when the gateway
# is not co-resident. Spins up an ephemeral sidecar call by spawning
# the existing sidecar binary one-shot; if a live sidecar is already
# listening on the configured port we reuse it.
# is not co-resident. Reuses a live sidecar already listening on the
# configured port (cron processes cannot spawn the sidecar themselves).
async def _standalone_send(
pconfig: PlatformConfig,
@@ -980,7 +1021,7 @@ def register(ctx) -> None:
ctx.register_platform(
name="photon",
label="Photon iMessage",
label="iMessage via Photon",
adapter_factory=lambda cfg: PhotonAdapter(cfg),
check_fn=check_requirements,
validate_config=validate_config,
@@ -1011,7 +1052,7 @@ def register(ctx) -> None:
"Treat replies like regular text messages — short, friendly, no "
"markdown rendering. Recipient identifiers are E.164 phone "
"numbers; never expose them in responses unless the user asked. "
"Attachments arrive as metadata only (no download URL yet)."
"Attachments arrive as metadata only."
),
)
File diff suppressed because it is too large Load Diff
+146 -131
View File
@@ -7,25 +7,26 @@ Subcommands:
setup full first-time setup (device login + project + user + sidecar)
status show login + project + sidecar dep state
install-sidecar npm install inside plugins/platforms/photon/sidecar/
webhook register register the local webhook URL with Photon
webhook list list registered webhooks
webhook delete delete a webhook by id
The device-code login runs automatically as the first step of ``setup``;
there is no standalone ``login`` verb (matching how every other Hermes
gateway channel onboards through a single setup surface).
Photon uses the spectrum-ts gRPC stream for inbound there is no webhook
to register, so there are no webhook subcommands.
"""
from __future__ import annotations
import argparse
import getpass
import json
import os
import shutil
import subprocess
import sys
from pathlib import Path
from hermes_cli.colors import Colors, color
from . import auth as photon_auth
_SIDECAR_DIR = Path(__file__).parent / "sidecar"
@@ -38,9 +39,14 @@ def register_cli(parser: argparse.ArgumentParser) -> None:
"""Wire up `hermes photon ...` subcommands."""
subs = parser.add_subparsers(dest="photon_command", required=False)
p_setup = subs.add_parser("setup", help="First-time setup (device login + project + user + sidecar)")
p_setup.add_argument("--project-name", default=None, help="Project name (default: 'Hermes Agent')")
p_setup.add_argument("--phone", default=None, help="Your E.164 phone number (e.g. +15551234567)")
p_setup = subs.add_parser(
"setup",
help="First-time setup (device login + project + user + sidecar)",
)
p_setup.add_argument("--project-name", default=None,
help="Project name (default: 'Hermes Agent')")
p_setup.add_argument("--phone", default=None,
help="Your E.164 phone number (e.g. +15551234567)")
p_setup.add_argument("--first-name", default=None)
p_setup.add_argument("--last-name", default=None)
p_setup.add_argument("--email", default=None)
@@ -52,14 +58,6 @@ def register_cli(parser: argparse.ArgumentParser) -> None:
subs.add_parser("status", help="Show login + project + sidecar dep state")
subs.add_parser("install-sidecar", help="Run npm install inside the sidecar directory")
p_hook = subs.add_parser("webhook", help="Manage Photon webhook registrations")
hook_subs = p_hook.add_subparsers(dest="photon_webhook_command", required=True)
p_hook_reg = hook_subs.add_parser("register", help="Register a webhook URL")
p_hook_reg.add_argument("url", help="Publicly reachable URL Photon should POST to")
hook_subs.add_parser("list", help="List registered webhooks for the current project")
p_hook_del = hook_subs.add_parser("delete", help="Delete a webhook by id")
p_hook_del.add_argument("webhook_id")
parser.set_defaults(func=dispatch)
@@ -77,8 +75,6 @@ def dispatch(args: argparse.Namespace) -> int:
return _cmd_status(args)
if sub == "install-sidecar":
return _cmd_install_sidecar(args)
if sub == "webhook":
return _cmd_webhook(args)
print(f"unknown subcommand: {sub}", file=sys.stderr)
return 2
@@ -122,7 +118,7 @@ def _cmd_setup(args: argparse.Namespace) -> int:
# 1. Login (skip if we already have a token).
token = photon_auth.load_photon_token()
if not token:
print("[1/4] No Photon token found — running device login...")
print("[1/5] No Photon token found — running device login...")
rc = _run_device_login(args)
if rc != 0:
return rc
@@ -131,85 +127,163 @@ def _cmd_setup(args: argparse.Namespace) -> int:
print("login completed but token was not stored", file=sys.stderr)
return 1
else:
print("[1/4] Reusing existing Photon token")
print("[1/5] Reusing existing Photon token")
# 2. Create (or surface existing) project.
existing_id, existing_secret = photon_auth.load_project_credentials()
project_id: str
project_secret: str
if existing_id and existing_secret:
project_id, project_secret = existing_id, existing_secret
# `project_id` is a Photon-assigned UUID, not a secret — but we
# keep the print terse to avoid CodeQL flow noise.
print("[2/4] Reusing existing Photon project")
else:
name = args.project_name or "Hermes Agent"
print(f"[2/4] Creating Photon project '{name}' (spectrum=true, imessage)...")
try:
data = photon_auth.create_project(token, name=name)
except Exception as e:
print(f"create-project failed: {e}", file=sys.stderr)
return 1
project_id = data.get("spectrumProjectId") or data.get("id") or ""
project_secret = data.get("projectSecret") or ""
if not project_id or not project_secret:
print(
"create-project did not return spectrumProjectId + "
"projectSecret. Re-run after enabling Spectrum on the "
"project, or open https://app.photon.codes/ to fetch the "
"secret manually.",
file=sys.stderr,
)
return 1
photon_auth.store_project_credentials(project_id, project_secret, name=name)
print(" ✓ project provisioned (run `hermes photon status` to see the id)")
# 2. Find or create the "Hermes Agent" project.
name = args.project_name or photon_auth.DEFAULT_PROJECT_NAME
dashboard_id = photon_auth.load_dashboard_project_id()
try:
if dashboard_id:
print("[2/5] Reusing configured Photon project")
else:
existing = photon_auth.find_project_by_name(token, name)
if existing and existing.get("id"):
dashboard_id = existing["id"]
print(f"[2/5] Found existing project '{name}'")
else:
print(f"[2/5] Creating Photon project '{name}'...")
created = photon_auth.create_project(token, name=name)
dashboard_id = created.get("id")
print("project created")
except Exception as e:
print(f"project setup failed: {e}", file=sys.stderr)
return 1
if not dashboard_id:
print("could not resolve a Photon project id", file=sys.stderr)
return 1
# 3. Create a Spectrum user for the operator.
# 3. Enable Spectrum, fetch the spectrum project id, rotate the secret,
# and persist both (runtime creds -> ~/.hermes/.env, ids -> auth.json).
try:
print("[3/5] Enabling Spectrum and provisioning credentials...")
proj = photon_auth.ensure_spectrum_enabled(token, dashboard_id)
spectrum_id = proj.get("spectrumProjectId")
if not spectrum_id:
print("spectrum provisioning failed: no spectrum project id", file=sys.stderr)
return 1
spectrum_id = str(spectrum_id)
secret = photon_auth.regenerate_project_secret(token, dashboard_id)
photon_auth.store_project_credentials(
spectrum_project_id=spectrum_id,
project_secret=secret,
dashboard_project_id=dashboard_id,
name=name,
)
# spectrum_id is an opaque non-secret id; safe to show.
print(f" ✓ Spectrum enabled (project id {spectrum_id}) — secret saved")
except Exception as e:
print(f"spectrum provisioning failed: {e}", file=sys.stderr)
return 1
# 4. Register the operator's phone number as a Spectrum user (idempotent).
phone = args.phone or _prompt(
"Your iMessage phone number (E.164, e.g. +15551234567): "
color(
"[4/5] Your iMessage phone number (E.164, e.g. +15551234567): ",
Colors.CYAN,
)
)
agent_number = None
if not phone:
print("[3/4] Skipped user creation (no phone given). Re-run with --phone later.")
print(" Skipped user registration (no phone given). Re-run with --phone later.")
else:
print("[3/4] Creating shared Spectrum user...")
# Name/email are optional and never prompted for — pass --first-name /
# --email if you want them sent to the dashboard.
first_name = args.first_name
email = args.email
try:
photon_auth.create_user(
project_id, project_secret,
user, created = photon_auth.register_user_if_absent(
token, dashboard_id,
phone_number=phone,
first_name=args.first_name,
first_name=first_name,
last_name=args.last_name,
email=args.email,
email=email,
)
except Exception as e:
print(f"create-user failed: {e}", file=sys.stderr)
except ValueError as e:
print(f" invalid phone number: {e}", file=sys.stderr)
return 1
print(" ✓ user created — check `hermes photon status` or the dashboard for the assigned iMessage line")
except Exception as e:
print(f" user registration failed: {e}", file=sys.stderr)
return 1
print(" ✓ phone registered" if created else " ✓ phone already registered")
# The number to text the agent is the user's assigned iMessage line
# (the dashboard's "TEXTS ON" column). On shared-number plans there is
# no dedicated entry in /lines, so this per-user field is the source of
# truth — and we already have it from the (reused) user object.
agent_number = photon_auth.user_assigned_line(user)
# Allowlist the operator and make their DM the cron home channel —
# otherwise the gateway denies their own inbound messages
# ("Unauthorized user") and has no default space for cron delivery.
_autoconfigure_access(phone)
# 4. Sidecar deps.
if args.skip_sidecar_install:
print("[4/4] Skipping sidecar npm install (--skip-sidecar-install)")
# 5. Surface the agent's iMessage number (the number to text the agent).
if not agent_number:
# No per-user assignment — fall back to a dedicated line if the project
# has one provisioned in its line inventory.
try:
line = photon_auth.get_imessage_line(token, dashboard_id)
if line:
agent_number = line.get("phoneNumber")
except Exception as e:
print(f" (could not fetch the assigned line: {e})", file=sys.stderr)
if agent_number:
print()
print(color("┌─ Your agent's iMessage number ───────────────────────────────", Colors.GREEN))
print(
color("│ 📱 ", Colors.GREEN)
+ color(str(agent_number), Colors.GREEN, Colors.BOLD)
)
print(color("│ Text this number from your phone to talk to your agent.", Colors.GREEN))
print(color("└──────────────────────────────────────────────────────────────", Colors.GREEN))
else:
print("[4/4] Installing Node sidecar deps (spectrum-ts)...")
print(" No iMessage line assigned yet — check the Photon dashboard.")
# 6. Sidecar deps (spectrum-ts).
if args.skip_sidecar_install:
print("[5/5] Skipping sidecar npm install (--skip-sidecar-install)")
else:
print("[5/5] Installing Node sidecar deps (spectrum-ts)...")
rc = _install_sidecar()
if rc != 0:
return rc
print()
print("✓ Photon setup complete.")
print(" Next: register a webhook URL Photon can reach:")
print(" hermes photon webhook register https://YOUR-PUBLIC-URL/photon/webhook")
print(" Then start the gateway:")
print(" hermes gateway start --platform photon")
print(" Start the gateway: hermes gateway start --platform photon")
return 0
def _autoconfigure_access(phone: str) -> None:
"""Allowlist the operator and set their DM as the cron home channel.
Writes ``PHOTON_ALLOWED_USERS`` (so the gateway authorizes the operator's
own inbound messages instead of denying them) and ``PHOTON_HOME_CHANNEL``
(the default space for cron delivery) to the operator's E.164 number. Each
is only filled when unset, so a hand-tuned allowlist / home channel is
never clobbered on a re-run.
"""
try:
from hermes_cli.config import get_env_value, save_env_value
except ImportError:
return
for key, label in (
("PHOTON_ALLOWED_USERS", "allowlisted your number"),
("PHOTON_HOME_CHANNEL", "set your DM as the cron home channel"),
):
try:
if get_env_value(key):
print(f" {key} already set — leaving it as-is.")
continue
save_env_value(key, phone)
print(f"{label} ({key})")
except Exception as e:
print(f" could not set {key}: {e}", file=sys.stderr)
def _cmd_status(_args: argparse.Namespace) -> int:
# Defer the whole table to auth.print_credential_summary — its emit
# Defer the credential rows to auth.print_credential_summary — its emit
# callback is the only sink that sees credential-derived strings, so
# cli.py keeps zero taint flow according to CodeQL.
photon_auth.print_credential_summary(print)
# The two non-credential rows live here so the helper stays purely
# about credentials.
node_bin = os.getenv("PHOTON_NODE_BIN") or shutil.which("node")
sidecar_installed = (_SIDECAR_DIR / "node_modules").exists()
print(f" node binary : {node_bin or '✗ missing (install Node 18+)'}")
@@ -218,8 +292,7 @@ def _cmd_status(_args: argparse.Namespace) -> int:
def _cmd_install_sidecar(_args: argparse.Namespace) -> int:
rc = _install_sidecar()
return rc
return _install_sidecar()
def _install_sidecar() -> int:
@@ -242,64 +315,6 @@ def _install_sidecar() -> int:
return proc.returncode
def _cmd_webhook(args: argparse.Namespace) -> int:
sub = getattr(args, "photon_webhook_command", None)
project_id, project_secret = photon_auth.load_project_credentials()
if not (project_id and project_secret):
print(
"no Photon project configured — run `hermes photon setup` first",
file=sys.stderr,
)
return 1
if sub == "register":
try:
data = photon_auth.register_webhook(
project_id, project_secret, webhook_url=args.url
)
except Exception as e:
print(f"register failed: {e}", file=sys.stderr)
return 1
# The helper does all the formatting + writing; cli.py never
# touches the signing-secret value, the path it was written
# to, or even the redacted-response dict. on_summary is a
# plain printer callback.
ok = photon_auth.persist_webhook_signing_secret(data, on_summary=print)
if not ok:
print(
"‼ Photon returned no signing secret in the response, "
"or the file write failed. Inspect your home directory "
"permissions and re-run; do not retry without first "
"deleting the orphaned webhook from the Photon dashboard.",
file=sys.stderr,
)
return 1
return 0
if sub == "list":
try:
data = photon_auth.list_webhooks(project_id, project_secret)
except Exception as e:
print(f"list failed: {e}", file=sys.stderr)
return 1
print(json.dumps(data, indent=2))
return 0
if sub == "delete":
try:
photon_auth.delete_webhook(
project_id, project_secret, webhook_id=args.webhook_id
)
except Exception as e:
print(f"delete failed: {e}", file=sys.stderr)
return 1
print(f"deleted webhook {args.webhook_id}")
return 0
print(f"unknown webhook subcommand: {sub}", file=sys.stderr)
return 2
# ---------------------------------------------------------------------------
# Gateway-setup entry point
#
+21 -40
View File
@@ -1,52 +1,37 @@
name: photon-platform
label: Photon iMessage
label: iMessage via Photon
kind: platform
version: 0.1.0
version: 0.2.0
description: >
Photon Spectrum gateway adapter for Hermes Agent.
Connects to iMessage (and other Spectrum interfaces) through Photon's
managed Spectrum platform. Inbound messages arrive as signed webhooks
on a local aiohttp server; outbound messages are sent via a small
supervised Node sidecar that runs the `spectrum-ts` SDK (Photon does
not currently expose a public HTTP send endpoint).
managed Spectrum platform. Both directions run over the `spectrum-ts`
SDK's long-lived gRPC stream via a small supervised Node sidecar —
inbound messages arrive on the SDK's `app.messages` stream (no webhook,
no public URL, no signing secret), and outbound messages are sent over
the same sidecar.
The plugin ships with a `hermes photon` CLI for the one-time login
+ project + user setup, persists Spectrum credentials to
``~/.hermes/auth.json`` under ``credential_pool.photon`` (token) and
``credential_pool.photon_project`` (project id + secret), and exposes
Photon's free shared-line model so users can get started without a
paid plan.
The plugin ships with a `hermes photon` CLI for the one-time device
login + project + user setup. Runtime credentials are written to
``~/.hermes/.env`` (``PHOTON_PROJECT_ID`` = the Spectrum project id,
``PHOTON_PROJECT_SECRET``) like every other channel, with management
metadata (device token, dashboard project id) in ``~/.hermes/auth.json``.
Photon's free shared-line model lets users get started without a paid plan.
author: NousResearch
requires_env:
- name: PHOTON_PROJECT_ID
description: "Spectrum project ID (set by `hermes photon setup`)"
prompt: "Photon Spectrum project ID"
description: "Spectrum project id (the project's spectrumProjectId; set by `hermes photon setup`)"
prompt: "Photon Spectrum project id"
url: "https://app.photon.codes/"
password: false
- name: PHOTON_PROJECT_SECRET
description: "Spectrum project secret (set by `hermes photon setup`)"
prompt: "Photon Spectrum project secret"
description: "Project secret paired with the Spectrum project id (set by `hermes photon setup`)"
prompt: "Photon project secret"
url: "https://app.photon.codes/"
password: true
optional_env:
- name: PHOTON_WEBHOOK_SECRET
description: "Per-URL HMAC-SHA256 signing secret returned at webhook registration"
prompt: "Photon webhook signing secret"
password: true
- name: PHOTON_WEBHOOK_PORT
description: "Local port the webhook receiver listens on (default 8788)"
prompt: "Webhook receiver port"
password: false
- name: PHOTON_WEBHOOK_PATH
description: "Path the webhook receiver listens on (default /photon/webhook)"
prompt: "Webhook receiver path"
password: false
- name: PHOTON_WEBHOOK_BIND
description: "Bind address for the webhook receiver (default 0.0.0.0)"
prompt: "Webhook bind address"
password: false
- name: PHOTON_SIDECAR_PORT
description: "Loopback port for the Node sidecar control channel (default 8789)"
description: "Loopback port for the Node sidecar control + inbound channel (default 8789)"
prompt: "Sidecar control port"
password: false
- name: PHOTON_SIDECAR_AUTOSTART
@@ -57,12 +42,8 @@ optional_env:
description: "Path to the node binary (default: shutil.which('node'))"
prompt: "Node executable path"
password: false
- name: PHOTON_API_HOST
description: "Spectrum management API host (default https://spectrum.photon.codes)"
prompt: "Spectrum API host"
password: false
- name: PHOTON_DASHBOARD_HOST
description: "Dashboard API host (default https://app.photon.codes)"
description: "Photon Dashboard API host (default https://app.photon.codes)"
prompt: "Dashboard host"
password: false
- name: PHOTON_ALLOWED_USERS
@@ -82,8 +63,8 @@ optional_env:
prompt: "Group mention patterns"
password: false
- name: PHOTON_HOME_CHANNEL
description: "Default Spectrum space ID for cron / notification delivery"
prompt: "Home space ID"
description: "Default Photon target for cron / notification delivery: Spectrum space id, DM GUID, or bare E.164 phone number"
prompt: "Home Photon target"
password: false
- name: PHOTON_HOME_CHANNEL_NAME
description: "Human label for the home channel"
+292 -71
View File
@@ -1,40 +1,46 @@
// Hermes Agent — Photon Spectrum sidecar
//
// Spawned by `plugins/platforms/photon/adapter.py` to bridge outbound
// messaging to Photon's Spectrum platform. Inbound messages go directly
// from Photon's webhook to Hermes' Python aiohttp receiver — this
// sidecar handles ONLY outbound calls (which require the spectrum-ts
// SDK because Photon has no public HTTP send endpoint today).
// Spawned by `plugins/platforms/photon/adapter.py` to bridge BOTH directions
// of messaging to Photon's Spectrum platform via the `spectrum-ts` SDK (the
// SDK is TypeScript-only, so a Node sidecar is unavoidable — there is no
// Python SDK and no public HTTP message API).
//
// Protocol:
// - The sidecar listens on http://127.0.0.1:${PORT} (loopback only)
// - Each request must include `X-Hermes-Sidecar-Token: ${TOKEN}`
// - POST /healthz -> {"ok": true}
// - POST /send -> {"ok": true, "messageId": "..."}
// body: {"spaceId": "...", "text": "...", "replyTo": "..." | null}
// - POST /send-attachment -> {"ok": true, "messageId": "..."}
// Inbound (gRPC -> Hermes): the SDK's `app.messages` async iterator is a
// long-lived gRPC stream. We serialize each `[space, message]` to a
// normalized JSON event and stream it to the Python adapter over a
// loopback `GET /inbound` (NDJSON). We pause pulling from the stream while
// no consumer is attached so a backlog isn't pulled-and-lost before the
// gateway connects.
// Outbound (Hermes -> gRPC): `/send` drives `space.send(...)`; `/typing`
// sends the documented `typing("start" | "stop")` content builder.
//
// Protocol (all requests require `X-Hermes-Sidecar-Token: ${TOKEN}`):
// - GET /inbound -> 200 NDJSON stream; one JSON event per line, blank
// lines are heartbeats. One consumer at a time.
// - POST /healthz -> {"ok": true}
// - POST /send -> {"ok": true, "messageId": "..."}
// body: {"spaceId": "...", "text": "..."}
// - POST /send-attachment -> {"ok": true, "messageId": "..."}
// body: {"spaceId": "...", "path": "...", "name": "..." | null,
// "mimeType": "..." | null, "caption": "..." | null,
// "kind": "attachment" | "voice", "replyTo": "..." | null}
// - POST /typing -> {"ok": true}
// body: {"spaceId": "..."}
// - POST /shutdown -> {"ok": true}; then process exits
// "kind": "attachment" | "voice"}
// - POST /typing -> {"ok": true}
// body: {"spaceId": "...", "state": "start" | "stop"}
// - POST /shutdown -> {"ok": true}; then process exits
//
// On SIGINT/SIGTERM the sidecar calls `app.stop()` (3s graceful) before
// exiting. Errors are logged to stderr; Python supervises restart.
// exiting. Logs go to stderr; Python supervises restart.
//
// Env vars (all required):
// PHOTON_PROJECT_ID
// Env vars (required):
// PHOTON_PROJECT_ID (== the project's spectrumProjectId)
// PHOTON_PROJECT_SECRET
// PHOTON_SIDECAR_PORT
// PHOTON_SIDECAR_TOKEN
//
// Optional:
// PHOTON_SIDECAR_BIND (default 127.0.0.1)
// PHOTON_API_HOST (passed through to spectrum-ts if its config
// honours it)
// PHOTON_SIDECAR_BIND (default 127.0.0.1)
import http from "node:http";
import { once } from "node:events";
const projectId = process.env.PHOTON_PROJECT_ID;
const projectSecret = process.env.PHOTON_PROJECT_SECRET;
@@ -42,6 +48,17 @@ const port = parseInt(process.env.PHOTON_SIDECAR_PORT || "8789", 10);
const bind = process.env.PHOTON_SIDECAR_BIND || "127.0.0.1";
const sharedToken = process.env.PHOTON_SIDECAR_TOKEN;
// Inbound attachments are read into memory and base64-inlined on the NDJSON
// event so the Python adapter can cache the real bytes (and the agent can see
// the image). Cap the size we inline — above it we forward metadata only and
// the adapter surfaces a text marker, so one large video can't balloon a
// single NDJSON line. Override via PHOTON_MAX_INLINE_ATTACHMENT_BYTES.
const MAX_INLINE_ATTACHMENT_BYTES =
Number(process.env.PHOTON_MAX_INLINE_ATTACHMENT_BYTES) || 20 * 1024 * 1024;
const DM_CHAT_GUID_RE = /^any;-;(\+\d{6,})$/;
const E164_RE = /^\+\d{6,}$/;
const MAX_KNOWN_SPACES = 2048;
if (!projectId || !projectSecret || !sharedToken) {
console.error(
"photon-sidecar: PHOTON_PROJECT_ID, PHOTON_PROJECT_SECRET and " +
@@ -52,9 +69,15 @@ if (!projectId || !projectSecret || !sharedToken) {
// Lazy-load spectrum-ts so a missing install fails with a clear message
// instead of a cryptic module-resolution error during import.
let Spectrum, imessage, attachment, voice;
let Spectrum, imessage, attachment, voice, spectrumText, spectrumTyping;
try {
({ Spectrum, attachment, voice } = await import("spectrum-ts"));
({
Spectrum,
attachment,
voice,
text: spectrumText,
typing: spectrumTyping,
} = await import("spectrum-ts"));
({ imessage } = await import("spectrum-ts/providers/imessage"));
} catch (e) {
console.error(
@@ -71,17 +94,168 @@ const app = await Spectrum({
providers: [imessage.config()],
});
// Drain the inbound stream — Photon's webhook is the canonical inbound
// path, but we still consume `app.messages` so spectrum-ts' internal
// reconnect/heartbeat logic keeps running. Each event is logged at
// debug level; everything else is a no-op here.
// ---------------------------------------------------------------------------
// Inbound: forward `app.messages` (gRPC stream) to the Python consumer.
// At most one Python consumer is attached at a time (the gateway adapter).
let consumerRes = null;
let consumerWaiters = [];
const knownSpaces = new Map();
function rememberKnownSpace(id, space) {
if (!id || typeof id !== "string" || !space) return;
if (knownSpaces.has(id)) knownSpaces.delete(id);
knownSpaces.set(id, space);
if (knownSpaces.size > MAX_KNOWN_SPACES) {
const oldest = knownSpaces.keys().next().value;
if (oldest) knownSpaces.delete(oldest);
}
}
function phoneTargetFromSpaceId(spaceId) {
if (typeof spaceId !== "string") return null;
if (E164_RE.test(spaceId)) return spaceId;
const dmGuid = spaceId.match(DM_CHAT_GUID_RE);
return dmGuid ? dmGuid[1] : null;
}
function rememberInboundSpace(space, message) {
const msgSpace = message?.space || {};
const ids = [space?.id, msgSpace.id];
for (const id of ids) {
rememberKnownSpace(id, space);
const phone = phoneTargetFromSpaceId(id);
if (phone) rememberKnownSpace(phone, space);
}
}
function waitForConsumer() {
if (consumerRes) return Promise.resolve();
return new Promise((resolve) => consumerWaiters.push(resolve));
}
function setConsumer(res) {
consumerRes = res;
const waiters = consumerWaiters;
consumerWaiters = [];
for (const resolve of waiters) resolve();
}
function clearConsumer(res) {
if (consumerRes === res) consumerRes = null;
}
// Write one NDJSON line to the active consumer. Blocks until a consumer is
// connected; if the write fails (consumer vanished mid-flight) we wait for a
// new consumer and retry, so a message is never silently dropped here.
async function deliver(line) {
for (;;) {
await waitForConsumer();
const res = consumerRes;
if (!res) continue;
try {
const flushed = res.write(line + "\n");
if (!flushed) await once(res, "drain");
return;
} catch {
clearConsumer(res);
}
}
}
async function normalizeContent(content) {
if (!content || typeof content !== "object") {
return { type: "unknown" };
}
if (content.type === "text") {
return { type: "text", text: content.text || "" };
}
if (content.type === "attachment") {
const meta = {
type: "attachment",
id: content.id ?? null,
name: content.name ?? null,
mimeType: content.mimeType ?? null,
size: typeof content.size === "number" ? content.size : null,
};
// Read the bytes eagerly and base64-inline them as `data` so the Python
// adapter can cache the real file (the agent then sees the image itself).
// The spectrum-ts attachment object may not outlive this stream
// iteration, so a lazy/on-demand fetch isn't safe. Over-cap attachments
// (when size is known up front) are forwarded as metadata only and the
// adapter falls back to a text marker. A read failure must never break
// the inbound loop — we just drop `data` and forward metadata.
if (meta.size !== null && meta.size > MAX_INLINE_ATTACHMENT_BYTES) {
console.error(
`photon-sidecar: attachment ${meta.name ?? meta.id} (${meta.size} bytes) ` +
`exceeds inline cap ${MAX_INLINE_ATTACHMENT_BYTES}; forwarding metadata only`
);
return meta;
}
if (typeof content.read === "function") {
try {
const buf = await content.read();
// Guard the case where size was unknown but the bytes turn out to be
// over the cap.
if (buf && buf.length > MAX_INLINE_ATTACHMENT_BYTES) {
console.error(
`photon-sidecar: attachment ${meta.name ?? meta.id} (${buf.length} bytes) ` +
`exceeds inline cap after read; forwarding metadata only`
);
return meta;
}
meta.data = Buffer.from(buf).toString("base64");
meta.encoding = "base64";
} catch (e) {
console.error(
"photon-sidecar: failed to read attachment bytes " +
"(forwarding metadata only): " +
(e && e.stack ? e.stack : String(e))
);
}
}
return meta;
}
return { type: content.type || "unknown" };
}
async function normalizeEvent(space, message) {
try {
const msgSpace = message.space || {};
const ts = message.timestamp;
return {
messageId: message.id ?? null,
platform: message.platform || space.__platform || "iMessage",
space: {
id: space.id ?? msgSpace.id ?? null,
// iMessage spaces carry `type` ("dm"|"group") and `phone` directly.
type: space.type ?? msgSpace.type ?? "dm",
phone: space.phone ?? msgSpace.phone ?? null,
},
sender: { id: message.sender ? message.sender.id : null },
content: await normalizeContent(message.content),
timestamp:
ts instanceof Date ? ts.toISOString() : ts ? String(ts) : null,
};
} catch (e) {
console.error(
"photon-sidecar: failed to normalize inbound message: " + String(e)
);
return null;
}
}
(async () => {
try {
for await (const [, message] of app.messages) {
console.error(
`photon-sidecar: drained inbound from ${message.platform} ` +
`space=${message.space?.id}`
);
for await (const [space, message] of app.messages) {
// Only forward inbound messages (ignore our own outbound echoes).
if (message && message.direction && message.direction !== "inbound") {
continue;
}
rememberInboundSpace(space, message);
const event = await normalizeEvent(space, message);
if (!event) continue;
await deliver(JSON.stringify(event));
}
} catch (e) {
console.error(
@@ -91,6 +265,9 @@ const app = await Spectrum({
}
})();
// ---------------------------------------------------------------------------
// HTTP control + inbound server (loopback only).
async function readBody(req) {
const chunks = [];
for await (const chunk of req) chunks.push(chunk);
@@ -130,27 +307,73 @@ function ok(res, data) {
res.end(JSON.stringify({ ok: true, ...data }));
}
async function resolveSpace(spaceId) {
// spectrum-ts exposes the same Space methods via `app.space(spaceId)` /
// narrowed helpers; we fall back through a few accessor shapes to
// tolerate small SDK API drift.
if (typeof app.space === "function") {
return await app.space(spaceId);
}
if (app.spaces && typeof app.spaces.get === "function") {
return await app.spaces.get(spaceId);
}
// Last resort — the platform-narrowed helper.
if (imessage) {
const im = imessage(app);
if (typeof im.space === "function") {
try {
return await im.space({ id: spaceId });
} catch {
/* fall through */
}
function handleInbound(req, res) {
res.statusCode = 200;
res.setHeader("Content-Type", "application/x-ndjson");
res.setHeader("Cache-Control", "no-store");
res.setHeader("Connection", "keep-alive");
// One consumer at a time — a fresh connection (e.g. after a reconnect)
// supersedes the previous one.
if (consumerRes && consumerRes !== res) {
try {
consumerRes.end();
} catch {
/* ignore */
}
}
setConsumer(res);
// Heartbeat keeps the socket warm through idle periods and lets the Python
// side detect a dead pipe promptly.
const heartbeat = setInterval(() => {
try {
res.write("\n");
} catch {
/* ignore */
}
}, 25000);
const cleanup = () => {
clearInterval(heartbeat);
clearConsumer(res);
};
req.on("close", cleanup);
req.on("aborted", cleanup);
res.on("error", cleanup);
}
async function resolveSpace(spaceId) {
const cached = knownSpaces.get(spaceId);
if (cached) return cached;
const phoneTarget = phoneTargetFromSpaceId(spaceId);
// A bare E.164 phone number addresses a DM. Resolve the user, then the (DM)
// space — `imessage(app).user(phone)` -> `im.space(user)` — so callers can
// pass just "+1..." (e.g. PHOTON_HOME_CHANNEL for cron delivery) instead of
// an opaque inbound space id. Photon also represents DM chat ids as
// `any;-;+1...`; normalize those through the same path so replies to inbound
// DMs still resolve after Python stores the inbound `space.id`.
if (phoneTarget && imessage) {
try {
const im = imessage(app);
const user = await im.user(phoneTarget);
const space = await im.space(user);
rememberKnownSpace(spaceId, space);
rememberKnownSpace(phoneTarget, space);
rememberKnownSpace(space?.id, space);
return space;
} catch (e) {
console.error(
"photon-sidecar: phone->DM resolution failed: " +
(e && e.stack ? e.stack : String(e))
);
}
}
// No cache hit and not a phone/DM target. spectrum-ts exposes no API to
// rehydrate an arbitrary opaque space id: a Space is only obtained from the
// inbound `[space, message]` stream (cached above in `knownSpaces`) or
// reconstructed for a DM from its phone number. So a group space whose cache
// entry was lost — e.g. after a sidecar restart with no fresh inbound message
// in that group — cannot be resolved here; a new inbound message in the group
// re-warms the cache. DMs are unaffected (reconstructed from the phone).
throw new Error(`unable to resolve space id ${spaceId}`);
}
@@ -158,6 +381,10 @@ const server = http.createServer(async (req, res) => {
if (req.headers["x-hermes-sidecar-token"] !== sharedToken) {
return unauthorized(res);
}
// Long-lived inbound NDJSON stream.
if (req.method === "GET" && req.url === "/inbound") {
return handleInbound(req, res);
}
if (req.method !== "POST") {
res.statusCode = 405;
return res.end();
@@ -173,18 +400,16 @@ const server = http.createServer(async (req, res) => {
}
const body = await readBody(req);
if (req.url === "/send") {
const { spaceId, text, replyTo } = body || {};
const { spaceId, text } = body || {};
if (!spaceId || typeof text !== "string") {
return badRequest(res, "spaceId and text are required");
}
const space = await resolveSpace(spaceId);
const result = replyTo
? await space.send(text, { replyTo })
: await space.send(text);
return ok(res, { messageId: result?.id || result?.messageId || null });
const result = await space.send(spectrumText(text));
return ok(res, { messageId: result?.id || null });
}
if (req.url === "/send-attachment") {
const { spaceId, path, name, mimeType, caption, kind, replyTo } =
const { spaceId, path, name, mimeType, caption, kind } =
body || {};
if (!spaceId || typeof path !== "string" || !path) {
return badRequest(res, "spaceId and path are required");
@@ -202,16 +427,13 @@ const server = http.createServer(async (req, res) => {
? voice(path, Object.keys(opts).length ? opts : undefined)
: attachment(path, Object.keys(opts).length ? opts : undefined);
const sendOpts = replyTo ? { replyTo } : undefined;
const result = sendOpts
? await space.send(builder, sendOpts)
: await space.send(builder);
const result = await space.send(builder);
// iMessage delivers the caption as a separate bubble; send it
// after the media so the attachment renders first.
if (caption && typeof caption === "string") {
try {
await space.send(caption);
await space.send(spectrumText(caption));
} catch (e) {
console.error(
"photon-sidecar: attachment sent but caption failed: " +
@@ -219,17 +441,16 @@ const server = http.createServer(async (req, res) => {
);
}
}
return ok(res, { messageId: result?.id || result?.messageId || null });
return ok(res, { messageId: result?.id || null });
}
if (req.url === "/typing") {
const { spaceId } = body || {};
const { spaceId, state = "start" } = body || {};
if (!spaceId) return badRequest(res, "spaceId is required");
const space = await resolveSpace(spaceId);
if (typeof space.typing === "function") {
await space.typing();
} else if (typeof space.setTyping === "function") {
await space.setTyping(true);
if (state !== "start" && state !== "stop") {
return badRequest(res, "state must be start or stop");
}
const space = await resolveSpace(spaceId);
await space.send(spectrumTyping(state));
return ok(res, {});
}
res.statusCode = 404;
File diff suppressed because it is too large Load Diff
@@ -1,7 +1,7 @@
{
"name": "@hermes-agent/photon-sidecar",
"private": true,
"version": "0.1.0",
"version": "0.2.0",
"description": "Spectrum-ts bridge for the Hermes Agent Photon platform plugin.",
"type": "module",
"main": "index.mjs",
@@ -12,6 +12,6 @@
"node": ">=18.17"
},
"dependencies": {
"spectrum-ts": "^0.1.0"
"spectrum-ts": "^1.17.1"
}
}
File diff suppressed because it is too large Load Diff
+21 -2
View File
@@ -1,7 +1,7 @@
name: simplex-platform
label: SimpleX Chat
kind: platform
version: 1.0.0
version: 1.1.0
description: >
SimpleX Chat gateway adapter for Hermes Agent.
Connects to a local simplex-chat daemon via WebSocket and relays
@@ -9,7 +9,7 @@ description: >
SimpleX is decentralised and assigns no persistent user IDs —
every contact is an opaque internal ID generated at connection
time, making it one of the most private messengers available.
author: Mibayy
author: Mibayy, jooray
# ``requires_env`` and ``optional_env`` entries are surfaced in the
# ``hermes config`` UI via the platform-plugin env var injector in
# ``hermes_cli/config.py``.
@@ -27,6 +27,18 @@ optional_env:
description: "Allow any contact to talk to the bot (dev only — disables allowlist)"
prompt: "Allow all contacts? (true/false)"
password: false
- name: SIMPLEX_AUTO_ACCEPT
description: "Auto-accept incoming contact requests (default: true)"
prompt: "Auto-accept contact requests? (true/false)"
password: false
- name: SIMPLEX_GROUP_ALLOWED
description: >-
Comma-separated SimpleX group IDs the bot should participate in, or
'*' to allow any group. Omit to ignore group messages entirely
(safer default — a bot in a group otherwise processes every
member's traffic).
prompt: "Allowed group IDs (comma-separated, or '*' for any)"
password: false
- name: SIMPLEX_HOME_CHANNEL
description: "Default contact/group ID for cron / notification delivery"
prompt: "Home channel contact/group ID (or empty)"
@@ -35,3 +47,10 @@ optional_env:
description: "Human label for the home channel (defaults to the ID)"
prompt: "Home channel display name (or empty)"
password: false
- name: HERMES_SIMPLEX_TEXT_BATCH_DELAY
description: >-
Quiet-period seconds (default: 0.8) used to concatenate rapid-fire
inbound text messages into a single MessageEvent — same pattern as
Telegram's text batching.
prompt: "Text batch flush delay in seconds (default 0.8)"
password: false
+4
View File
@@ -45,6 +45,8 @@ ACP_REGISTRY_MANIFEST = REPO_ROOT / "acp_registry" / "agent.json"
# Auto-extracted from noreply emails + manual overrides
AUTHOR_MAP = {
"zhuhaoyu0909@icloud.com": "underthestars-zhy",
"raysun12142006@gmail.com": "yanxue06",
"alberto.regalado@ymail.com": "ARegalado1",
"alchemistchaos@protonmail.com": "AlchemistChaos", # co-author only
"gilad@smiti.ai": "giladbau",
@@ -971,6 +973,7 @@ AUTHOR_MAP = {
"draixagent@gmail.com": "draix",
"martin.alca@gmail.com": "draix",
"junminliu@gmail.com": "JimLiu",
"juraj@bednar.io": "jooray",
"jarvischer@gmail.com": "maxchernin",
"levantam.98.2324@gmail.com": "LVT382009",
"zhurongcheng@rcrai.com": "heykb",
@@ -1241,6 +1244,7 @@ AUTHOR_MAP = {
"charliekerfoot@gmail.com": "CharlieKerfoot", # PR #18951
# Debug share upload-time redaction (May 2026)
"dhuysamen@gmail.com": "GodsBoy", # PR #19318
"github@nadyahermes.anonaddy.com": "ruangraung", # PR #42308
"mrcoferland@gmail.com": "mrcoferland", # PR #19023
"chenlinfeng@ruije.com.cn": "noOne-list", # PR #19050
"briansu@Mac-mini.attlocal.net": "likejudy", # PR #19052
+306 -193
View File
@@ -1,29 +1,31 @@
"""Regression tests for issue #30768 and #32383.
"""Regression tests for #30768, #32383, and #33961.
``_prompt_text_input_modal`` uses a queue-based modal that relies on
prompt_toolkit key bindings receiving keyboard events. On Windows the
prompt_toolkit input channel can deadlock when the modal is entered from
the ``process_loop`` daemon thread. The fix falls back to the simpler
``_prompt_text_input`` (stdin-based) prompt on Windows.
``_prompt_text_input_modal`` answers destructive-slash confirmations through a
queue-based modal driven by prompt_toolkit key bindings. When invoked from the
``process_loop`` daemon thread it sets the modal up on the app's event loop via
``call_soon_threadsafe``, so it is safe on every platform including native
Windows (#33961), where the earlier ``sys.platform == "win32"`` → raw ``input()``
fallback deadlocked the daemon thread against prompt_toolkit's stdin ownership.
These tests verify:
1. Windows detection triggers the stdin fallback
2. Non-Windows daemon threads still use the modal via the app loop
3. macOS/Linux main-thread path still uses the modal (no regression)
4. No-app path still uses the stdin fallback (existing behavior)
5. Empty choices returns None (existing behavior)
1. Daemon-thread confirm uses the modal via the app loop on Linux AND native
Windows (#33961) — never the raw stdin fallback, never a hang.
2. Main-thread confirm with a running app uses the modal.
3. The raw stdin fallback is kept ONLY for the safe cases: no running app, and
(on win32, off-thread) a scheduling failure degrades to a clean cancel.
4. Empty choices returns None.
"""
import queue
import sys
import threading
import time
from unittest.mock import MagicMock, patch
import pytest
def _make_cli():
"""Minimal HermesCLI shell exposing prompt/modal helpers."""
"""Minimal HermesCLI shell exposing the prompt/modal helpers."""
import cli as cli_mod
obj = object.__new__(cli_mod.HermesCLI)
@@ -37,9 +39,6 @@ def _make_cli():
return obj
# ---------------------------------------------------------------------------
# Sample choices used across tests
# ---------------------------------------------------------------------------
_SAMPLE_CHOICES = [
("once", "Approve Once", "proceed this time only"),
("always", "Always Approve", "proceed and silence this prompt permanently"),
@@ -47,119 +46,106 @@ _SAMPLE_CHOICES = [
]
class TestModalWindowsFallback:
"""Windows dead-lock regression tests for _prompt_text_input_modal."""
def _answer_modal_when_open(cli, response, stop=None):
"""Push ``response`` onto the modal's response_queue once it opens.
def test_windows_falls_back_to_stdin(self):
"""On Windows, _prompt_text_input_modal should use _prompt_text_input."""
Gives up after ~2s, or early when ``stop`` is set (the modal will never open,
e.g. a scheduling failure) so degraded-path tests don't wait the full budget.
"""
for _ in range(100):
if stop is not None and stop.is_set():
return
state = cli._slash_confirm_state
if state and "response_queue" in state:
state["response_queue"].put(response)
return
time.sleep(0.02)
def _run_on_daemon(call, cli, *, platform, response, schedule=None):
"""Invoke ``call`` on a daemon thread — as the process_loop does — answering
the modal with ``response`` once it opens.
Returns ``{result, stdin_called, capture, restore}``. ``schedule`` overrides
the ``call_soon_threadsafe`` side effect (default: run the callback inline);
pass a raiser to simulate a scheduling failure. Fails if the worker hangs,
which is the deadlock canary for #33961.
"""
outcome = {"capture": [], "restore": [], "result": None, "stdin_called": False}
done = threading.Event()
def _worker():
try:
with patch.object(sys, "platform", platform), \
patch.object(cli._app.loop, "call_soon_threadsafe", side_effect=schedule or (lambda cb: cb())), \
patch.object(cli, "_prompt_text_input") as mock_stdin, \
patch.object(cli, "_invalidate"), \
patch.object(cli, "_capture_modal_input_snapshot", side_effect=lambda: outcome["capture"].append(1)), \
patch.object(cli, "_restore_modal_input_snapshot", side_effect=lambda: outcome["restore"].append(1)):
outcome["result"] = call()
outcome["stdin_called"] = mock_stdin.called
finally:
done.set()
worker = threading.Thread(target=_worker, daemon=True)
answerer = threading.Thread(target=_answer_modal_when_open, args=(cli, response, done), daemon=True)
answerer.start()
worker.start()
worker.join(timeout=2.0)
answerer.join(timeout=2.0)
assert not worker.is_alive(), "daemon thread hung — modal deadlocked"
return outcome
class TestModal:
"""Behaviour of _prompt_text_input_modal across platforms and threads."""
@pytest.mark.parametrize("platform", ["linux", "win32"])
def test_daemon_thread_uses_modal_via_app_loop(self, platform):
"""Off the process_loop daemon thread, the confirm uses the modal via
call_soon_threadsafe on every platform including native Windows, where
the old win32 early-return deadlocked on raw input() (#33961)."""
cli = _make_cli()
with patch.object(sys, "platform", "win32"), \
patch.object(cli, "_prompt_text_input", return_value="1") as mock_stdin:
result = cli._prompt_text_input_modal(
title="⚠️ /new — destroys conversation state",
outcome = _run_on_daemon(
lambda: cli._prompt_text_input_modal(
title="⚠️ /reset",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
)
timeout=5,
),
cli,
platform=platform,
response="once",
)
assert outcome["stdin_called"] is False, "must use the modal, not raw input()"
assert outcome["result"] == "once"
assert outcome["capture"] == [1]
assert outcome["restore"] == [1]
assert cli._slash_confirm_state is None
# The stdin-based fallback was used, not the modal queue path.
mock_stdin.assert_called_once_with("Choice [1/2/3]: ")
assert result == "1"
def test_non_main_thread_uses_modal_via_app_loop(self):
"""Off the main thread on Linux, keep the modal path via app-loop setup."""
def test_main_thread_with_app_uses_modal(self):
"""On the main thread with a running app, the queue-based modal is used."""
cli = _make_cli()
result_holder = {}
setup_calls = []
teardown_calls = []
def _call_soon_threadsafe(callback):
callback()
def run_on_daemon():
with patch.object(sys, "platform", "linux"), \
patch.object(cli._app.loop, "call_soon_threadsafe", side_effect=_call_soon_threadsafe), \
patch.object(cli, "_prompt_text_input") as mock_stdin, \
patch.object(cli, "_capture_modal_input_snapshot", side_effect=lambda: setup_calls.append("capture")), \
patch.object(cli, "_restore_modal_input_snapshot", side_effect=lambda: teardown_calls.append("restore")):
result_holder["result"] = cli._prompt_text_input_modal(
title="⚠️ /reset",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
timeout=5,
)
result_holder["stdin_called"] = mock_stdin.called
def _submit_after_delay():
time.sleep(0.2)
state = cli._slash_confirm_state
if state and "response_queue" in state:
state["response_queue"].put("once")
submitter = threading.Thread(target=_submit_after_delay, daemon=True)
t = threading.Thread(target=run_on_daemon, daemon=True)
submitter.start()
t.start()
t.join(timeout=2.0)
submitter.join(timeout=2.0)
assert not t.is_alive(), "daemon thread hung — modal deadlocked"
assert result_holder["stdin_called"] is False
assert result_holder["result"] == "once"
assert setup_calls == ["capture"]
assert teardown_calls == ["restore"]
def test_main_thread_non_windows_uses_modal(self):
"""On macOS/Linux main thread, the queue-based modal is still used."""
cli = _make_cli()
# We need to simulate the modal receiving a response. We'll patch
# the response_queue to immediately return a value.
with patch.object(sys, "platform", "darwin"), \
patch.object(cli, "_capture_modal_input_snapshot"), \
patch.object(cli, "_restore_modal_input_snapshot"), \
patch.object(cli, "_invalidate"):
# Start the modal in a way that it will receive a response
# immediately via the queue.
original_queue = queue.Queue
original_time = time.monotonic
patch.object(cli, "_invalidate"), \
patch.object(cli, "_prompt_text_input") as mock_stdin:
answerer = threading.Thread(target=_answer_modal_when_open, args=(cli, "once"), daemon=True)
answerer.start()
result = cli._prompt_text_input_modal(
title="⚠️ /new",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
timeout=5,
)
answerer.join(timeout=2.0)
def _fake_modal_flow(*args, **kwargs):
"""Simulate the modal flow: set state, put response, return."""
# We'll directly test that the modal path is entered by
# checking that _slash_confirm_state was set.
pass
# Since we can't easily mock the internal queue, let's test
# that the modal path is entered by checking that
# _prompt_text_input was NOT called.
with patch.object(cli, "_prompt_text_input") as mock_stdin:
# Set up a response that will be put into the queue
# after the modal starts waiting.
def _submit_after_delay():
time.sleep(0.2)
state = cli._slash_confirm_state
if state and "response_queue" in state:
state["response_queue"].put("once")
submitter = threading.Thread(target=_submit_after_delay, daemon=True)
submitter.start()
result = cli._prompt_text_input_modal(
title="⚠️ /new",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
timeout=5,
)
submitter.join(timeout=2.0)
# The stdin fallback should NOT have been called.
mock_stdin.assert_not_called()
# The result should be "once" from the simulated modal response.
assert result == "once"
mock_stdin.assert_not_called()
assert result == "once"
def test_no_app_falls_back_to_stdin(self):
"""Without a prompt_toolkit app, always use stdin fallback."""
"""Without a running app (oneshot / non-interactive), use the stdin prompt."""
cli = _make_cli()
cli._app = None
@@ -173,78 +159,102 @@ class TestModalWindowsFallback:
mock_stdin.assert_called_once_with("Choice [1/2/3]: ")
assert result == "3"
def test_empty_choices_returns_none(self):
"""Empty choices list should return None without prompting."""
cli = _make_cli()
with patch.object(cli, "_prompt_text_input") as mock_stdin:
result = cli._prompt_text_input_modal(
title="Test",
detail="Test",
choices=[],
)
mock_stdin.assert_not_called()
assert result is None
def test_windows_fallback_does_not_set_modal_state(self):
"""Verify Windows fallback doesn't leave _slash_confirm_state set."""
def test_windows_no_app_falls_back_to_stdin(self):
"""win32 without a running app keeps stdin — the only case where the raw
prompt is safe on Windows, since no app owns the console to deadlock."""
cli = _make_cli()
cli._app = None
with patch.object(sys, "platform", "win32"), \
patch.object(cli, "_prompt_text_input", return_value="1"):
cli._prompt_text_input_modal(
title="⚠️ /reset",
patch.object(cli, "_prompt_text_input", return_value="1") as mock_stdin:
result = cli._prompt_text_input_modal(
title="⚠️ /new — destroys conversation state",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
)
mock_stdin.assert_called_once_with("Choice [1/2/3]: ")
assert result == "1"
def test_windows_scheduling_failure_clean_cancels(self):
"""win32 off the main thread: if marshaling onto the app loop fails, cancel
cleanly (None) rather than fall to raw input() (which deadlocks on native
Windows) or hang. Asserts the _stdin_fallback guard (#33961)."""
cli = _make_cli()
def _raise(_cb):
raise RuntimeError("loop closed")
outcome = _run_on_daemon(
lambda: cli._prompt_text_input_modal(
title="⚠️ /reset",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
timeout=5,
),
cli,
platform="win32",
response="once",
schedule=_raise,
)
assert outcome["stdin_called"] is False, "win32 off-thread must NOT call raw input()"
assert outcome["result"] is None
assert cli._slash_confirm_state is None
def test_non_main_thread_modal_clears_state(self):
"""Verify daemon-thread modal teardown does not leave state behind."""
@pytest.mark.parametrize(
"platform, expect_stdin, expect_result",
[("win32", False, None), ("linux", True, "1")],
)
def test_daemon_thread_no_app_loop_uses_fallback(self, platform, expect_stdin, expect_result):
"""Off the daemon thread with no resolvable app loop (``self._app.loop``
is None / raises), the modal can never be scheduled, so the method short-
circuits at the app_loop-is-None site (cli.py ~7260) a distinct path
from a call_soon_threadsafe failure. win32 clean-cancels (None) instead of
deadlocking on raw input(); other platforms keep the stdin prompt."""
cli = _make_cli()
errors = []
cli._app.loop = None # forces app_loop is None, off the main thread
def _call_soon_threadsafe(callback):
callback()
outcome = {"result": None, "stdin_called": False}
done = threading.Event()
def run_on_daemon():
def _worker():
try:
with patch.object(sys, "platform", "linux"), \
patch.object(cli._app.loop, "call_soon_threadsafe", side_effect=_call_soon_threadsafe):
def _submit_after_delay():
time.sleep(0.2)
state = cli._slash_confirm_state
if state and "response_queue" in state:
state["response_queue"].put("cancel")
submitter = threading.Thread(target=_submit_after_delay, daemon=True)
submitter.start()
cli._prompt_text_input_modal(
title="⚠️ /new",
with patch.object(sys, "platform", platform), \
patch.object(cli, "_prompt_text_input", return_value="1") as mock_stdin, \
patch.object(cli, "_invalidate"):
outcome["result"] = cli._prompt_text_input_modal(
title="⚠️ /reset",
detail="This starts a fresh session.",
choices=_SAMPLE_CHOICES,
timeout=5,
)
submitter.join(timeout=2.0)
if cli._slash_confirm_state is not None:
errors.append("_slash_confirm_state should be None")
except Exception as exc:
errors.append(str(exc))
outcome["stdin_called"] = mock_stdin.called
finally:
done.set()
t = threading.Thread(target=run_on_daemon, daemon=True)
t.start()
t.join(timeout=2.0)
assert not errors, f"unexpected errors: {errors}"
worker = threading.Thread(target=_worker, daemon=True)
worker.start()
worker.join(timeout=2.0)
assert not worker.is_alive(), "daemon thread hung — modal deadlocked"
assert outcome["stdin_called"] is expect_stdin
assert outcome["result"] == expect_result
assert cli._slash_confirm_state is None
def test_empty_choices_returns_none(self):
"""Empty choices returns None without prompting."""
cli = _make_cli()
with patch.object(cli, "_prompt_text_input") as mock_stdin:
result = cli._prompt_text_input_modal(title="Test", detail="Test", choices=[])
mock_stdin.assert_not_called()
assert result is None
class TestConfirmDestructiveSlashWindows:
"""Integration-level tests for _confirm_destructive_slash on Windows."""
"""End-to-end _confirm_destructive_slash on the native-Windows daemon thread."""
def test_confirm_destructive_slash_bypasses_modal_on_windows(self):
"""_confirm_destructive_slash should work on Windows via stdin fallback."""
def _make_interactive_cli(self):
cli = _make_cli()
cli.model = "test-model"
cli._agent_running = False
@@ -255,37 +265,140 @@ class TestConfirmDestructiveSlashWindows:
cli._pending_tool_info = {}
cli._tool_start_time = 0.0
cli._last_scrollback_tool = ""
return cli
with patch.object(sys, "platform", "win32"), \
patch.object(cli, "_prompt_text_input", return_value="1"), \
patch("cli.load_cli_config", return_value={"approvals": {"destructive_slash_confirm": True}}):
result = cli._confirm_destructive_slash(
"new",
"This starts a fresh session.\nThe current conversation history will be discarded.",
@pytest.mark.parametrize(
"response, expected",
[("once", "once"), ("cancel", None)],
)
def test_confirm_destructive_slash_uses_modal_on_windows(self, response, expected):
"""On native Windows, the bare /new confirm drives the modal (not stdin)
and returns the chosen outcome the bug #33961 froze this path."""
cli = self._make_interactive_cli()
with patch("cli.load_cli_config", return_value={"approvals": {"destructive_slash_confirm": True}}):
outcome = _run_on_daemon(
lambda: cli._confirm_destructive_slash(
"new",
"This starts a fresh session.\nThe current conversation history will be discarded.",
),
cli,
platform="win32",
response=response,
)
assert result == "once"
assert outcome["stdin_called"] is False
assert outcome["result"] == expected
def test_confirm_destructive_slash_cancelled_on_windows(self):
"""Cancellation via stdin fallback works on Windows."""
class TestNativeWindowsNoRawInputDeadlock:
"""Anti-regression guard exercising the REAL ``_prompt_text_input``.
Every other test here mocks ``_prompt_text_input`` away, so they only
assert *routing* (modal vs. stdin) they cannot observe the actual hang
that #33961 was. The historical regression was precisely that
``_prompt_text_input_modal`` delegated to the *real* ``_prompt_text_input``
on native Windows, which on a non-main thread runs a bare ``input()`` that
blocks forever against prompt_toolkit's stdin ownership.
These tests let the real ``_prompt_text_input`` run with a blocking
``input()`` and assert the worker thread never hangs. They fail on the
pre-#33961 code (win32 → ``_prompt_text_input`` → off-main ``input()``)
and pass once the modal path / clean-cancel fallback is in place.
"""
def test_win32_daemon_thread_never_blocks_on_real_input(self):
"""A blocking input() must NOT hang the daemon thread on win32.
Drives the genuine helper chain (no mock of ``_prompt_text_input``)
with ``builtins.input`` patched to block forever. The confirm must
resolve via the app-loop modal (answered on a background thread, as
the real key bindings would) and never sit in ``input()``. On the
pre-#33961 code the win32 early-return routed to the real
``_prompt_text_input`` off-main ``input()`` permanent hang.
"""
cli = _make_cli()
cli.model = "test-model"
cli._agent_running = False
cli._spinner_text = ""
cli._should_exit = False
cli._command_running = False
cli.session_id = "test-session"
cli._pending_tool_info = {}
cli._tool_start_time = 0.0
cli._last_scrollback_tool = ""
cli._app.loop.call_soon_threadsafe = lambda cb: cb()
with patch.object(sys, "platform", "win32"), \
patch.object(cli, "_prompt_text_input", return_value="3"), \
patch("cli.load_cli_config", return_value={"approvals": {"destructive_slash_confirm": True}}):
result = cli._confirm_destructive_slash(
"reset",
"This starts a fresh session.\nThe current conversation history will be discarded.",
)
def _blocking_input(prompt=""): # stands in for "no line ever arrives"
time.sleep(30)
return "1"
# Choice "3" normalizes to "cancel", which returns None.
assert result is None
outcome = {}
done = threading.Event()
def _worker():
try:
with patch.object(sys, "platform", "win32"), \
patch("builtins.input", side_effect=_blocking_input), \
patch.object(cli, "_capture_modal_input_snapshot"), \
patch.object(cli, "_restore_modal_input_snapshot"), \
patch.object(cli, "_invalidate"):
outcome["result"] = cli._prompt_text_input_modal(
title="/new",
detail="destroys conversation state",
choices=_SAMPLE_CHOICES,
timeout=3,
)
finally:
done.set()
worker = threading.Thread(target=_worker, daemon=True)
answerer = threading.Thread(
target=_answer_modal_when_open, args=(cli, "cancel", done), daemon=True
)
answerer.start()
worker.start()
worker.join(timeout=5.0)
answerer.join(timeout=5.0)
assert not worker.is_alive(), (
"daemon thread hung in real input() — native-Windows confirm "
"deadlock regressed (#33961)"
)
# cancel → None; the point is it RETURNED rather than blocking forever.
assert outcome.get("result") in (None, "cancel")
def test_win32_scheduling_failure_cleanly_cancels_no_input(self):
"""If the modal can't be marshaled onto the app loop on native Windows
(scheduling failure) the off-main-thread path must cancel cleanly
NOT fall through to a blocking raw ``input()``.
This is the degraded branch the pre-#33961 code handled with
``return self._prompt_text_input(...)`` (which deadlocks); the fix
returns ``None`` instead.
"""
cli = _make_cli()
def _raise(cb): # call_soon_threadsafe scheduling failure
raise RuntimeError("event loop closed")
cli._app.loop.call_soon_threadsafe = _raise
input_called = {"n": 0}
def _tracking_input(prompt=""):
input_called["n"] += 1
time.sleep(30)
return "1"
outcome = {}
def _worker():
with patch.object(sys, "platform", "win32"), \
patch("builtins.input", side_effect=_tracking_input), \
patch.object(cli, "_invalidate"):
outcome["result"] = cli._prompt_text_input_modal(
title="/new",
detail="destroys conversation state",
choices=_SAMPLE_CHOICES,
timeout=3,
)
worker = threading.Thread(target=_worker, daemon=True)
worker.start()
worker.join(timeout=5.0)
assert not worker.is_alive(), (
"daemon thread hung — win32 scheduling-failure fallback used raw "
"input() instead of cleanly cancelling (#33961)"
)
assert input_called["n"] == 0, "win32 off-thread fallback must not call input()"
assert outcome.get("result") is None
+120
View File
@@ -1264,3 +1264,123 @@ async def test_verbose_mode_respects_explicit_tool_preview_length(monkeypatch, t
assert VerboseAgent.LONG_CODE not in all_content
# But should still contain the truncated portion with "..."
assert "..." in all_content
class CodeBlockProgressAdapter(ProgressCaptureAdapter):
"""A markdown-capable progress adapter (declares supports_code_blocks)."""
supports_code_blocks = True
class TerminalCommandAgent:
"""Emits a terminal tool.started with a real, multi-line command arg."""
CMD = (
"set -euo pipefail\n"
"printf 'node: '; node --version\n"
"npm install -g hyperframes@latest"
)
def __init__(self, **kwargs):
self.tool_progress_callback = kwargs.get("tool_progress_callback")
self.tools = []
def run_conversation(self, message, conversation_history=None, task_id=None):
self.tool_progress_callback(
"tool.started", "terminal", self.CMD, {"command": self.CMD}
)
# Let the async progress task drain the queue and send before returning.
time.sleep(0.35)
return {"final_response": "done", "messages": [], "api_calls": 1}
@pytest.mark.asyncio
async def test_terminal_progress_is_truncated_preview_not_bash_block(monkeypatch, tmp_path):
"""Regression for #41215: terminal progress must render as a short truncated
preview, never the full command in a fenced ```bash block, even on a
markdown-capable (supports_code_blocks) gateway."""
monkeypatch.setenv("HERMES_TOOL_PROGRESS_MODE", "all")
fake_dotenv = types.ModuleType("dotenv")
fake_dotenv.load_dotenv = lambda *args, **kwargs: None
monkeypatch.setitem(sys.modules, "dotenv", fake_dotenv)
fake_run_agent = types.ModuleType("run_agent")
fake_run_agent.AIAgent = TerminalCommandAgent
monkeypatch.setitem(sys.modules, "run_agent", fake_run_agent)
import tools.terminal_tool # noqa: F401 - register terminal emoji
adapter = CodeBlockProgressAdapter(platform=Platform.TELEGRAM)
runner = _make_runner(adapter)
gateway_run = importlib.import_module("gateway.run")
monkeypatch.setattr(gateway_run, "_hermes_home", tmp_path)
monkeypatch.setattr(gateway_run, "_resolve_runtime_agent_kwargs", lambda: {"api_key": "***"})
source = SessionSource(
platform=Platform.TELEGRAM,
chat_id="12345",
chat_type="dm",
thread_id=None,
)
result = await runner._run_agent(
message="hello",
context_prompt="",
history=[],
source=source,
session_id="sess-terminal-no-bash-block",
session_key="agent:main:telegram:dm:12345",
)
assert result["final_response"] == "done"
all_content = " ".join(call["content"] for call in adapter.sent)
all_content += " ".join(call["content"] for call in adapter.edits)
# Compact truncated preview, not a fenced bash block.
assert "```bash" not in all_content
assert 'terminal: "' in all_content
# The full multi-line command body must not reach the chat.
assert "npm install -g hyperframes@latest" not in all_content
@pytest.mark.asyncio
async def test_terminal_progress_no_bash_block_in_verbose_mode(monkeypatch, tmp_path):
"""#41215 also rendered the bash block in verbose mode. The revert removed it
from both branches, so verbose progress must not emit a fenced ```bash block
either (verbose still shows args by opt-in, just not as a code block)."""
monkeypatch.setenv("HERMES_TOOL_PROGRESS_MODE", "verbose")
fake_dotenv = types.ModuleType("dotenv")
fake_dotenv.load_dotenv = lambda *args, **kwargs: None
monkeypatch.setitem(sys.modules, "dotenv", fake_dotenv)
fake_run_agent = types.ModuleType("run_agent")
fake_run_agent.AIAgent = TerminalCommandAgent
monkeypatch.setitem(sys.modules, "run_agent", fake_run_agent)
import tools.terminal_tool # noqa: F401 - register terminal emoji
adapter = CodeBlockProgressAdapter(platform=Platform.TELEGRAM)
runner = _make_runner(adapter)
gateway_run = importlib.import_module("gateway.run")
monkeypatch.setattr(gateway_run, "_hermes_home", tmp_path)
monkeypatch.setattr(gateway_run, "_resolve_runtime_agent_kwargs", lambda: {"api_key": "***"})
source = SessionSource(
platform=Platform.TELEGRAM,
chat_id="12345",
chat_type="dm",
thread_id=None,
)
result = await runner._run_agent(
message="hello",
context_prompt="",
history=[],
source=source,
session_id="sess-terminal-verbose-no-bash",
session_key="agent:main:telegram:dm:12345",
)
assert result["final_response"] == "done"
all_content = " ".join(call["content"] for call in adapter.sent)
all_content += " ".join(call["content"] for call in adapter.edits)
assert "```bash" not in all_content
+48
View File
@@ -75,6 +75,54 @@ async def test_capabilities_advertises_session_control_surface(adapter):
}
@pytest.mark.asyncio
async def test_run_agent_binds_api_session_context_for_tool_env(adapter, monkeypatch):
"""API-server request sessions should reach tools and terminal subprocess env."""
monkeypatch.setenv("HERMES_SESSION_ID", "stale-session")
observed = {}
class FakeAgent:
session_prompt_tokens = 0
session_completion_tokens = 0
session_total_tokens = 0
def __init__(self, session_id: str):
self.session_id = session_id
def run_conversation(self, user_message, conversation_history, task_id):
from gateway.session_context import get_session_env
from tools.environments.local import _make_run_env
observed["task_id"] = task_id
observed["context_session_id"] = get_session_env("HERMES_SESSION_ID")
observed["context_platform"] = get_session_env("HERMES_SESSION_PLATFORM")
observed["context_session_key"] = get_session_env("HERMES_SESSION_KEY")
observed["child_session_id"] = _make_run_env({}).get("HERMES_SESSION_ID")
return {"final_response": "ok"}
def fake_create_agent(**kwargs):
return FakeAgent(kwargs["session_id"])
monkeypatch.setattr(adapter, "_create_agent", fake_create_agent)
result, usage = await adapter._run_agent(
user_message="hello",
conversation_history=[],
session_id="request-session",
gateway_session_key="request-key",
)
assert result["session_id"] == "request-session"
assert usage == {"input_tokens": 0, "output_tokens": 0, "total_tokens": 0}
assert observed == {
"task_id": "request-session",
"context_session_id": "request-session",
"context_platform": "api_server",
"context_session_key": "request-key",
"child_session_id": "request-session",
}
@pytest.mark.asyncio
async def test_session_crud_and_message_history(adapter, session_db):
app = _create_session_app(adapter)
+11
View File
@@ -190,6 +190,17 @@ def test_session_key_falls_back_to_os_environ(monkeypatch):
assert get_session_env("HERMES_SESSION_KEY") == ""
def test_session_id_set_via_contextvars(monkeypatch):
"""set_session_vars should set HERMES_SESSION_ID via contextvars."""
monkeypatch.setenv("HERMES_SESSION_ID", "stale-env-session")
tokens = set_session_vars(session_id="ctx-session-456")
assert get_session_env("HERMES_SESSION_ID") == "ctx-session-456"
clear_session_vars(tokens)
assert get_session_env("HERMES_SESSION_ID") == ""
def test_set_session_env_includes_session_key():
"""_set_session_env should propagate session_key from SessionContext."""
runner = object.__new__(GatewayRunner)
+20 -1
View File
@@ -205,6 +205,13 @@ def test_corr_id_pending_set_self_trims():
@pytest.mark.asyncio
async def test_send_dm():
"""DMs use the bare ``@<id> text`` chat-command form.
The bracketed form ``@[<id>] text`` is what the daemon's man page
documents, but in practice both addressing styles route through
the same chat-command parser; bare ``@<id>`` matches what every
Hermes deployment has been using in production for months.
"""
from gateway.config import PlatformConfig
cfg = PlatformConfig(enabled=True, extra={"ws_url": "ws://localhost:5225"})
adapter = SimplexAdapter(cfg)
@@ -222,6 +229,14 @@ async def test_send_dm():
@pytest.mark.asyncio
async def test_send_group():
"""Groups use the structured ``/_send #<id> json [...]`` form.
The bracket chat-command form ``#[<id>] text`` *looks* like an exact
ID match in the daemon docs but is parsed as a display-name lookup
so messages to groups whose display name isn't literally the ID
silently drop. The structured ``/_send`` form addresses by numeric
ID and survives newlines/quoting through ``json.dumps``.
"""
from gateway.config import PlatformConfig
cfg = PlatformConfig(enabled=True, extra={"ws_url": "ws://localhost:5225"})
adapter = SimplexAdapter(cfg)
@@ -231,7 +246,11 @@ async def test_send_group():
result = await adapter.send("group:grp-99", "Hello, group!")
payload = json.loads(mock_ws.send.call_args[0][0])
assert payload["cmd"] == "#[grp-99] Hello, group!"
assert payload["cmd"].startswith("/_send #grp-99 json ")
msg_content = json.loads(payload["cmd"].split(" json ", 1)[1])[0][
"msgContent"
]
assert msg_content == {"type": "text", "text": "Hello, group!"}
assert result.success is True
+1 -1
View File
@@ -835,7 +835,7 @@ class TestEditMessageStreamingSafety:
assert second_call == {
"chat_id": 123,
"message_id": 456,
"text": "final **bold**",
"text": "final bold",
}
@pytest.mark.asyncio
+400 -3
View File
@@ -27,7 +27,7 @@ import hermes_cli.dashboard_register as dr
def _ns(**kw):
defaults = dict(name=None, redirect_uri=None)
defaults = dict(name=None, redirect_uri=None, portal_url=None)
defaults.update(kw)
return argparse.Namespace(**defaults)
@@ -76,7 +76,7 @@ def _fake_http_ok(payload: dict):
class TestHappyPath:
def _run(self, *, args, account_token="tok_abc", portal="https://portal.nousresearch.com",
response=None, captured=None):
response=None, captured=None, existing_client_id=None):
response = response or {
"client_id": "agent:selfhost-1",
"id": "selfhost-1",
@@ -98,12 +98,21 @@ class TestHappyPath:
def fake_save(key, value):
saved[key] = value
# get_env_value is consulted twice: once for the stored client_id
# (idempotency key) and once for HERMES_DASHBOARD_PORTAL_URL. Route by
# key so a test can seed a prior client_id while keeping the portal
# unset (the default-portal-not-persisted path).
def fake_get_env(key):
if key == "HERMES_DASHBOARD_OAUTH_CLIENT_ID":
return existing_client_id
return None
with patch(
"hermes_cli.auth.resolve_nous_access_token", return_value=account_token
), patch("hermes_cli.config.is_managed", return_value=False), patch.object(
dr, "_resolve_portal_base_url", return_value=portal
), patch(
"hermes_cli.config.get_env_value", return_value=None
"hermes_cli.config.get_env_value", side_effect=fake_get_env
), patch(
"hermes_cli.config.save_env_value", side_effect=fake_save
), patch.object(
@@ -157,6 +166,394 @@ class TestHappyPath:
)
class TestIdempotentRerun(TestHappyPath):
"""Re-running with a stored client_id updates instead of creating.
Inherits ``_run`` from TestHappyPath; the only new lever is
``existing_client_id`` (the HERMES_DASHBOARD_OAUTH_CLIENT_ID a prior run
persisted), which the CLI re-sends so the portal updates that row.
"""
def test_stored_client_id_is_sent_as_idempotency_key(self, capsys):
captured: dict = {}
# Portal echoes back the SAME id -> it updated in place.
self._run(
args=_ns(),
existing_client_id="agent:selfhost-1",
response={
"client_id": "agent:selfhost-1",
"id": "selfhost-1",
"name": "dreamy_tesla",
"kind": "SELF_HOSTED",
"custom_redirect_uri": None,
"created_at": "2026-06-04T12:00:00.000Z",
},
captured=captured,
)
assert captured["body"]["client_id"] == "agent:selfhost-1"
def test_rerun_without_name_omits_name_to_preserve_stored(self, capsys):
# No --name on a re-run: don't churn the portal-stored name. The CLI
# leaves `name` out of the body so the portal keeps what it has.
captured: dict = {}
self._run(
args=_ns(),
existing_client_id="agent:selfhost-1",
captured=captured,
)
assert "name" not in captured["body"]
assert captured["body"]["client_id"] == "agent:selfhost-1"
def test_rerun_with_explicit_name_still_sends_name(self, capsys):
captured: dict = {}
self._run(
args=_ns(name="renamed_box"),
existing_client_id="agent:selfhost-1",
captured=captured,
)
assert captured["body"]["name"] == "renamed_box"
assert captured["body"]["client_id"] == "agent:selfhost-1"
def test_rerun_prints_updated_when_same_id_returned(self, capsys):
self._run(
args=_ns(),
existing_client_id="agent:selfhost-1",
response={
"client_id": "agent:selfhost-1",
"id": "selfhost-1",
"name": "dreamy_tesla",
"kind": "SELF_HOSTED",
"custom_redirect_uri": None,
"created_at": "2026-06-04T12:00:00.000Z",
},
)
out = capsys.readouterr().out
assert "Updated dashboard" in out
assert "Registered dashboard" not in out
def test_rerun_persists_returned_client_id(self, capsys):
saved = self._run(
args=_ns(),
existing_client_id="agent:selfhost-1",
)
# Same id round-trips into .env -> idempotent, one record.
assert saved["HERMES_DASHBOARD_OAUTH_CLIENT_ID"] == "agent:selfhost-1"
def test_stale_id_falls_through_to_create_prints_registered(self, capsys):
# Stored id no longer resolves server-side -> portal created a fresh
# row and returns a DIFFERENT id. The CLI treats that as a create and
# persists the new id (re-run stays safe, never worse than first run).
captured: dict = {}
saved = self._run(
args=_ns(name="seed_name"),
existing_client_id="agent:selfhost-stale",
response={
"client_id": "agent:selfhost-new",
"id": "selfhost-new",
"name": "seed_name",
"kind": "SELF_HOSTED",
"custom_redirect_uri": None,
"created_at": "2026-06-04T12:00:00.000Z",
},
captured=captured,
)
# The stale id is still SENT (portal decides create-vs-update).
assert captured["body"]["client_id"] == "agent:selfhost-stale"
# Returned id differs from what we sent -> message is "Registered".
out = capsys.readouterr().out
assert "Registered dashboard" in out
assert "Updated dashboard" not in out
assert saved["HERMES_DASHBOARD_OAUTH_CLIENT_ID"] == "agent:selfhost-new"
def test_blank_stored_client_id_treated_as_first_run(self, capsys):
# A blank/whitespace stored value is not a usable key: treat as a
# first registration (auto-generate a name, don't send client_id).
captured: dict = {}
self._run(
args=_ns(),
existing_client_id=" ",
captured=captured,
)
assert "client_id" not in captured["body"]
assert captured["body"].get("name") # auto-generated
class TestCustomPortalPersistence:
"""`--portal-url` / HERMES_DASHBOARD_PORTAL_URL is persisted to .env.
An *explicitly supplied* custom portal URL is an intentional choice the
user wants to survive across sessions, so it's always written (updating an
existing entry in place rather than appending a duplicate). When no custom
URL is supplied, the older conservative behaviour is preserved: an inferred
portal is only written when absent and non-default, and an existing entry
is never altered unexpectedly.
"""
def _run(self, *, args, portal, existing_portal):
"""Drive cmd_dashboard_register, capturing save_env_value calls.
`existing_portal` is what get_env_value returns for
HERMES_DASHBOARD_PORTAL_URL (None = not present in .env).
"""
response = {
"client_id": "agent:selfhost-1",
"id": "selfhost-1",
"name": "dreamy_tesla",
"kind": "SELF_HOSTED",
"custom_redirect_uri": None,
"created_at": "2026-06-04T12:00:00.000Z",
}
saved: dict = {}
def fake_save(key, value):
saved[key] = value
def fake_get_env_value(key, *a, **kw):
if key == "HERMES_DASHBOARD_PORTAL_URL":
return existing_portal
return None
with patch(
"hermes_cli.auth.resolve_nous_access_token", return_value="tok"
), patch("hermes_cli.config.is_managed", return_value=False), patch.dict(
dr.os.environ, {}, clear=False
), patch.object(
dr, "_resolve_portal_base_url", return_value=portal
), patch(
"hermes_cli.config.get_env_value", side_effect=fake_get_env_value
), patch(
"hermes_cli.config.save_env_value", side_effect=fake_save
), patch.object(
dr.urllib.request, "urlopen", return_value=_fake_http_ok(response)
):
# The ambient process env may carry HERMES_DASHBOARD_PORTAL_URL
# (e.g. staging dev shells); drop it so `custom_portal_supplied`
# is driven solely by the args.portal_url under test.
dr.os.environ.pop("HERMES_DASHBOARD_PORTAL_URL", None)
dr.cmd_dashboard_register(args)
return saved
def test_explicit_custom_url_persisted_when_var_absent(self, capsys):
saved = self._run(
args=_ns(portal_url="https://preview.example.com"),
portal="https://preview.example.com",
existing_portal=None,
)
assert saved["HERMES_DASHBOARD_PORTAL_URL"] == "https://preview.example.com"
def test_explicit_custom_url_updates_existing_in_place(self, capsys):
# An entry already exists with a different value; the explicit custom
# URL overwrites it (save_env_value updates the matching key in place).
saved = self._run(
args=_ns(portal_url="https://new-preview.example.com"),
portal="https://new-preview.example.com",
existing_portal="https://old-preview.example.com",
)
assert (
saved["HERMES_DASHBOARD_PORTAL_URL"] == "https://new-preview.example.com"
)
def test_explicit_custom_url_persisted_even_when_equals_default(self, capsys):
# User explicitly asked for the production portal — honour the explicit
# request and persist it (the no-flag path would skip the default).
saved = self._run(
args=_ns(portal_url="https://portal.nousresearch.com"),
portal="https://portal.nousresearch.com",
existing_portal=None,
)
assert (
saved["HERMES_DASHBOARD_PORTAL_URL"] == "https://portal.nousresearch.com"
)
def test_explicit_custom_url_equal_to_existing_is_noop(self, capsys):
# Already persisted with the same value → no redundant write.
saved = self._run(
args=_ns(portal_url="https://preview.example.com"),
portal="https://preview.example.com",
existing_portal="https://preview.example.com",
)
assert "HERMES_DASHBOARD_PORTAL_URL" not in saved
def test_no_flag_default_portal_not_written(self, capsys):
# No custom URL supplied, resolves to default → not written.
saved = self._run(
args=_ns(),
portal="https://portal.nousresearch.com",
existing_portal=None,
)
assert "HERMES_DASHBOARD_PORTAL_URL" not in saved
def test_no_flag_does_not_overwrite_existing_entry(self, capsys):
# No custom URL supplied and the var already exists → left untouched,
# even if the inferred portal differs (acceptance criterion 4).
saved = self._run(
args=_ns(),
portal="https://inferred-from-login.example.com",
existing_portal="https://already-set.example.com",
)
assert "HERMES_DASHBOARD_PORTAL_URL" not in saved
class TestPublicUrlPersistence:
"""`--redirect-uri` derives & persists HERMES_DASHBOARD_PUBLIC_URL in .env.
--redirect-uri is the full public callback (e.g.
https://hermes.example.com/auth/callback). At serve time the dashboard auth
layer reconstructs that callback by appending "/auth/callback" to
HERMES_DASHBOARD_PUBLIC_URL, so the value that's actually consumed is the
ORIGIN (scheme://host). We derive the origin from the supplied redirect URI
and persist THAT as HERMES_DASHBOARD_PUBLIC_URL the var the runtime reads
so the public-URL override is genuinely wired, not just stored.
An explicitly supplied value is always written (updating an existing entry
in place rather than appending a duplicate); a no-op when it already
matches; and never written on a localhost-only install (no --redirect-uri).
"""
def _run(self, *, args, existing_public=None):
"""Drive cmd_dashboard_register, capturing save_env_value calls.
`existing_public` is what get_env_value returns for
HERMES_DASHBOARD_PUBLIC_URL (None = not present in .env).
"""
response = {
"client_id": "agent:selfhost-1",
"id": "selfhost-1",
"name": "dreamy_tesla",
"kind": "SELF_HOSTED",
"custom_redirect_uri": getattr(args, "redirect_uri", None),
"created_at": "2026-06-04T12:00:00.000Z",
}
saved: dict = {}
def fake_save(key, value):
saved[key] = value
def fake_get_env_value(key, *a, **kw):
if key == "HERMES_DASHBOARD_PUBLIC_URL":
return existing_public
return None
with patch(
"hermes_cli.auth.resolve_nous_access_token", return_value="tok"
), patch("hermes_cli.config.is_managed", return_value=False), patch.dict(
dr.os.environ, {}, clear=False
), patch.object(
dr, "_resolve_portal_base_url", return_value="https://portal.nousresearch.com"
), patch(
"hermes_cli.config.get_env_value", side_effect=fake_get_env_value
), patch(
"hermes_cli.config.save_env_value", side_effect=fake_save
), patch.object(
dr.urllib.request, "urlopen", return_value=_fake_http_ok(response)
):
dr.os.environ.pop("HERMES_DASHBOARD_PORTAL_URL", None)
dr.cmd_dashboard_register(args)
return saved
def test_origin_derived_from_full_callback_path(self, capsys):
# The key behaviour: a full callback URL is reduced to its ORIGIN so
# the runtime's "public_url + /auth/callback" reconstruction matches.
saved = self._run(
args=_ns(redirect_uri="https://hermes.example.com/auth/callback"),
existing_public=None,
)
assert saved["HERMES_DASHBOARD_PUBLIC_URL"] == "https://hermes.example.com"
# The full callback path must NOT be persisted verbatim (would double
# the path at serve time).
assert "/auth/callback" not in saved["HERMES_DASHBOARD_PUBLIC_URL"]
def test_origin_preserves_port(self, capsys):
saved = self._run(
args=_ns(redirect_uri="https://hermes.example.com:8443/auth/callback"),
existing_public=None,
)
assert saved["HERMES_DASHBOARD_PUBLIC_URL"] == "https://hermes.example.com:8443"
def test_public_url_updates_existing_in_place(self, capsys):
# A stale public-url entry exists; the new derived origin overwrites it.
saved = self._run(
args=_ns(redirect_uri="https://new.example.com/auth/callback"),
existing_public="https://old.example.com",
)
assert saved["HERMES_DASHBOARD_PUBLIC_URL"] == "https://new.example.com"
def test_public_url_equal_to_existing_is_noop(self, capsys):
# Derived origin already matches what's stored → no redundant write.
saved = self._run(
args=_ns(redirect_uri="https://hermes.example.com/auth/callback"),
existing_public="https://hermes.example.com",
)
assert "HERMES_DASHBOARD_PUBLIC_URL" not in saved
def test_no_redirect_flag_not_written(self, capsys):
# Localhost-only install (no --redirect-uri) → var left untouched.
saved = self._run(
args=_ns(),
existing_public=None,
)
assert "HERMES_DASHBOARD_PUBLIC_URL" not in saved
def test_no_redirect_flag_does_not_overwrite_existing(self, capsys):
# No --redirect-uri supplied but a value already exists → never touch
# it (an existing entry is only changed by an explicit new value).
saved = self._run(
args=_ns(),
existing_public="https://already-set.example.com",
)
assert "HERMES_DASHBOARD_PUBLIC_URL" not in saved
def test_non_http_redirect_not_persisted(self, capsys):
# A malformed / non-http(s) redirect yields no derivable origin → skip.
saved = self._run(
args=_ns(redirect_uri="not-a-url"),
existing_public=None,
)
assert "HERMES_DASHBOARD_PUBLIC_URL" not in saved
def test_public_url_persisted_alongside_portal_url(self, capsys):
# Both --portal-url and --redirect-uri supplied → portal_url AND the
# derived public_url are both persisted (ADD semantics: the public-url
# write does not displace portal-url persistence).
response = {
"client_id": "agent:selfhost-1",
"id": "selfhost-1",
"name": "dreamy_tesla",
"kind": "SELF_HOSTED",
"custom_redirect_uri": "https://hermes.example.com/auth/callback",
"created_at": "2026-06-04T12:00:00.000Z",
}
saved: dict = {}
def fake_save(key, value):
saved[key] = value
with patch(
"hermes_cli.auth.resolve_nous_access_token", return_value="tok"
), patch("hermes_cli.config.is_managed", return_value=False), patch.dict(
dr.os.environ, {}, clear=False
), patch.object(
dr, "_resolve_portal_base_url", return_value="https://preview.example.com"
), patch(
"hermes_cli.config.get_env_value", return_value=None
), patch(
"hermes_cli.config.save_env_value", side_effect=fake_save
), patch.object(
dr.urllib.request, "urlopen", return_value=_fake_http_ok(response)
):
dr.os.environ.pop("HERMES_DASHBOARD_PORTAL_URL", None)
dr.cmd_dashboard_register(
_ns(
portal_url="https://preview.example.com",
redirect_uri="https://hermes.example.com/auth/callback",
)
)
assert saved["HERMES_DASHBOARD_PORTAL_URL"] == "https://preview.example.com"
assert saved["HERMES_DASHBOARD_PUBLIC_URL"] == "https://hermes.example.com"
class TestPortalResolution:
def test_override_arg_wins(self):
assert (
+48
View File
@@ -540,6 +540,54 @@ def test_run_doctor_accepts_hermes_provider_ids_that_catalog_aliases(
)
def test_run_doctor_accepts_vendor_slugs_for_named_custom_provider(monkeypatch, tmp_path):
home = tmp_path / ".hermes"
home.mkdir(parents=True, exist_ok=True)
(home / "config.yaml").write_text(
"model:\n"
" provider: custom:hpc-ai\n"
" default: deepseek/deepseek-v4-flash\n"
"custom_providers:\n"
" - name: hpc-ai\n"
" base_url: https://hpc-ai.example/v1\n"
" api_key: test-key\n",
encoding="utf-8",
)
monkeypatch.setattr(doctor_mod, "HERMES_HOME", home)
monkeypatch.setattr(doctor_mod, "PROJECT_ROOT", tmp_path / "project")
monkeypatch.setattr(doctor_mod, "_DHH", str(home))
(tmp_path / "project").mkdir(exist_ok=True)
fake_model_tools = types.SimpleNamespace(
check_tool_availability=lambda *a, **kw: ([], []),
TOOLSET_REQUIREMENTS={},
)
monkeypatch.setitem(sys.modules, "model_tools", fake_model_tools)
try:
from hermes_cli import auth as _auth_mod
monkeypatch.setattr(_auth_mod, "get_nous_auth_status", lambda: {})
monkeypatch.setattr(_auth_mod, "get_codex_auth_status", lambda: {})
monkeypatch.setattr(_auth_mod, "get_xai_oauth_auth_status", lambda: {})
except Exception:
pass
buf = io.StringIO()
with contextlib.redirect_stdout(buf):
doctor_mod.run_doctor(Namespace(fix=False))
out = buf.getvalue()
assert "model.provider 'custom:hpc-ai' is not a recognised provider" not in out
assert "model.provider 'custom:hpc-ai' is unknown" not in out
assert (
"model.default 'deepseek/deepseek-v4-flash' uses a vendor/model slug but provider is "
"'custom:hpc-ai'"
not in out
)
assert "Either set model.provider to 'openrouter', or drop the vendor prefix." not in out
def test_run_doctor_accepts_kimi_coding_cn_provider(monkeypatch, tmp_path):
@@ -65,6 +65,15 @@ def test_resolve_provider_full_finds_named_custom_provider():
assert resolved.source == "user-config"
def test_is_aggregator_recognizes_named_custom_provider():
assert providers_mod.is_aggregator("custom:hpc-ai") is True
assert providers_mod.is_aggregator("custom:litellm") is True
def test_is_aggregator_leaves_unknown_provider_non_aggregator():
assert providers_mod.is_aggregator("not-a-provider") is False
def test_switch_model_accepts_explicit_named_custom_provider(monkeypatch):
"""Shared /model switch pipeline should accept --provider for custom_providers."""
monkeypatch.setattr(
+410 -151
View File
@@ -1,8 +1,8 @@
"""Tests for the Photon auth module (device login + project + user creation)."""
"""Tests for the Photon auth module (device login + dashboard API)."""
from __future__ import annotations
import json
import time
import os
from pathlib import Path
from typing import Any, Dict
@@ -36,51 +36,91 @@ class _FakeResponse:
raise RuntimeError(f"HTTP {self.status_code}")
_PHOTON_ENV = (
"PHOTON_PROJECT_ID",
"PHOTON_PROJECT_SECRET",
"PHOTON_DASHBOARD_PROJECT_ID",
)
@pytest.fixture
def tmp_hermes_home(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> Path:
def tmp_hermes_home(tmp_path: Path, monkeypatch: pytest.MonkeyPatch):
home = tmp_path / "hermes"
home.mkdir()
monkeypatch.setenv("HERMES_HOME", str(home))
# The auth module memoises by reading get_hermes_home at call time
# so the env var is what matters.
return home
for key in _PHOTON_ENV:
monkeypatch.delenv(key, raising=False)
yield home
# save_env_value() mutates os.environ directly, so scrub any leakage.
for key in _PHOTON_ENV:
os.environ.pop(key, None)
# ---------------------------------------------------------------------------
# Credential storage
def test_store_and_load_photon_token(tmp_hermes_home: Path) -> None:
photon_auth.store_photon_token("abc123def456")
assert photon_auth.load_photon_token() == "abc123def456"
auth_json = json.loads((tmp_hermes_home / "auth.json").read_text())
assert "credential_pool" in auth_json
assert auth_json["credential_pool"]["photon"][0]["access_token"] == "abc123def456"
def test_store_and_load_project_credentials(tmp_hermes_home: Path) -> None:
def test_store_project_credentials_round_trip(
tmp_hermes_home: Path, monkeypatch: pytest.MonkeyPatch,
) -> None:
# Don't touch .env / os.environ here — exercise the auth.json path.
monkeypatch.setattr(photon_auth, "_persist_runtime_env", lambda *a, **k: None)
photon_auth.store_project_credentials(
"proj-uuid", "secret-key", name="Test Project",
spectrum_project_id="sp-123",
project_secret="secret-key",
dashboard_project_id="dash-456",
name="Hermes Agent",
)
pid, secret = photon_auth.load_project_credentials()
assert pid == "proj-uuid"
for key in _PHOTON_ENV:
monkeypatch.delenv(key, raising=False)
sid, secret = photon_auth.load_project_credentials()
assert sid == "sp-123"
assert secret == "secret-key"
assert photon_auth.load_dashboard_project_id() == "dash-456"
def test_store_project_credentials_writes_env(tmp_hermes_home: Path) -> None:
photon_auth.store_project_credentials(
spectrum_project_id="sp-789",
project_secret="sek-ret",
dashboard_project_id="dash-1",
)
env_text = (tmp_hermes_home / ".env").read_text()
assert "PHOTON_PROJECT_ID=sp-789" in env_text
assert "PHOTON_PROJECT_SECRET=sek-ret" in env_text
def test_load_project_credentials_env_override(
tmp_hermes_home: Path, monkeypatch: pytest.MonkeyPatch,
) -> None:
photon_auth.store_project_credentials("from-file", "secret-file")
monkeypatch.setattr(photon_auth, "_persist_runtime_env", lambda *a, **k: None)
photon_auth.store_project_credentials(
spectrum_project_id="from-file", project_secret="secret-file",
)
monkeypatch.setenv("PHOTON_PROJECT_ID", "from-env")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "secret-env")
pid, secret = photon_auth.load_project_credentials()
assert pid == "from-env"
sid, secret = photon_auth.load_project_credentials()
assert sid == "from-env"
assert secret == "secret-env"
def test_request_device_code(monkeypatch: pytest.MonkeyPatch) -> None:
# ---------------------------------------------------------------------------
# Device login flow
def test_request_device_code_uses_photon_cli(monkeypatch: pytest.MonkeyPatch) -> None:
captured: Dict[str, Any] = {}
def fake_post(url: str, *, json: Dict[str, Any], timeout: float) -> _FakeResponse:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
captured["url"] = url
captured["body"] = json
captured["body"] = kwargs.get("json")
return _FakeResponse(json_body={
"device_code": "dev-code-xyz",
"user_code": "ABCD-1234",
@@ -95,189 +135,408 @@ def test_request_device_code(monkeypatch: pytest.MonkeyPatch) -> None:
code = photon_auth.request_device_code()
assert code.device_code == "dev-code-xyz"
assert code.user_code == "ABCD-1234"
assert code.expires_in == 600
assert "/api/auth/device/code" in captured["url"]
assert captured["body"]["client_id"] == "hermes-agent"
# Hosted Photon allowlists registered device clients — an unregistered
# client_id is rejected with 400 invalid_client. We use Photon's published
# CLI device client and send the standard scope.
assert captured["body"]["client_id"] == "photon-cli"
assert captured["body"]["scope"] == "openid profile email"
def test_poll_for_token_via_header(monkeypatch: pytest.MonkeyPatch) -> None:
"""Token from set-auth-token header is the documented mechanism."""
def fake_post(url: str, *, json: Dict[str, Any], timeout: float) -> _FakeResponse:
return _FakeResponse(
status=200,
json_body={"session": {}, "user": {}},
headers={"set-auth-token": "bearer-xyz"},
)
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
code = photon_auth.DeviceCode(
def _device_code() -> "photon_auth.DeviceCode":
return photon_auth.DeviceCode(
device_code="d", user_code="u",
verification_uri="https://x", verification_uri_complete=None,
expires_in=10, interval=0,
)
token = photon_auth.poll_for_token(code, interval=0, timeout=2)
assert token == "bearer-xyz"
def test_poll_for_token_via_body_fallback(monkeypatch: pytest.MonkeyPatch) -> None:
"""If the header is absent we fall back to session.access_token."""
def fake_post(url: str, *, json: Dict[str, Any], timeout: float) -> _FakeResponse:
return _FakeResponse(
status=200,
json_body={"session": {"access_token": "from-body"}, "user": {}},
)
def test_poll_for_token_body_access_token(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(status=200, json_body={"access_token": "tok-body"})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
code = photon_auth.DeviceCode(
device_code="d", user_code="u",
verification_uri="https://x", verification_uri_complete=None,
expires_in=10, interval=0,
)
assert photon_auth.poll_for_token(code, interval=0, timeout=2) == "from-body"
assert photon_auth.poll_for_token(_device_code(), interval=0, timeout=2) == "tok-body"
def test_poll_for_token_propagates_access_denied(
monkeypatch: pytest.MonkeyPatch,
) -> None:
def fake_post(url: str, *, json: Dict[str, Any], timeout: float) -> _FakeResponse:
return _FakeResponse(
status=400, json_body={"error": "access_denied"},
)
def test_poll_for_token_session_fallback(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(status=200, json_body={"session": {"access_token": "tok-sess"}})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
assert photon_auth.poll_for_token(_device_code(), interval=0, timeout=2) == "tok-sess"
def test_poll_for_token_header_fallback(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(status=200, json_body={}, headers={"set-auth-token": "tok-hdr"})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
assert photon_auth.poll_for_token(_device_code(), interval=0, timeout=2) == "tok-hdr"
def test_poll_for_token_pending_then_success(monkeypatch: pytest.MonkeyPatch) -> None:
calls = {"n": 0}
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
calls["n"] += 1
if calls["n"] == 1:
return _FakeResponse(status=400, json_body={"error": "authorization_pending"})
return _FakeResponse(status=200, json_body={"access_token": "tok-eventual"})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
assert photon_auth.poll_for_token(_device_code(), interval=0, timeout=5) == "tok-eventual"
assert calls["n"] == 2
def test_poll_for_token_access_denied(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(status=400, json_body={"error": "access_denied"})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
code = photon_auth.DeviceCode(
device_code="d", user_code="u",
verification_uri="https://x", verification_uri_complete=None,
expires_in=10, interval=0,
)
with pytest.raises(RuntimeError, match="access_denied"):
photon_auth.poll_for_token(code, interval=0, timeout=2)
photon_auth.poll_for_token(_device_code(), interval=0, timeout=2)
# ---------------------------------------------------------------------------
# Projects
def test_list_projects_unwraps_list(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body=[{"id": "p1", "name": "Hermes Agent"}])
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
projects = photon_auth.list_projects("tok")
assert projects[0]["id"] == "p1"
def test_find_project_by_name_case_insensitive(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body={"data": [
{"id": "p1", "name": "Other"},
{"id": "p2", "name": "hermes agent"},
]})
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
proj = photon_auth.find_project_by_name("tok", "Hermes Agent")
assert proj is not None and proj["id"] == "p2"
def test_create_project_sends_spectrum_true(monkeypatch: pytest.MonkeyPatch) -> None:
captured: Dict[str, Any] = {}
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
captured["url"] = url
captured["body"] = kwargs.get("json")
captured["headers"] = kwargs.get("headers")
return _FakeResponse(json_body={"success": True, "id": "new-proj"})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
data = photon_auth.create_project("tok", name="Hermes Agent")
assert data["id"] == "new-proj"
assert captured["body"]["spectrum"] is True
assert captured["body"]["name"] == "Hermes Agent"
assert captured["headers"]["Authorization"] == "Bearer tok"
assert captured["url"].endswith("/api/projects")
def test_create_project_raises_without_id(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body={"success": True})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
with pytest.raises(RuntimeError, match="project id"):
photon_auth.create_project("tok")
def test_ensure_spectrum_enabled_toggles_when_off(monkeypatch: pytest.MonkeyPatch) -> None:
get_calls = {"n": 0}
posted = {"toggle": False}
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
get_calls["n"] += 1
if get_calls["n"] == 1:
return _FakeResponse(json_body={"id": "p", "spectrum": False, "spectrumProjectId": None})
return _FakeResponse(json_body={"id": "p", "spectrum": True, "spectrumProjectId": "sp-1"})
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
if url.endswith("/spectrum/toggle"):
posted["toggle"] = True
return _FakeResponse(json_body={"success": True})
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
proj = photon_auth.ensure_spectrum_enabled("tok", "p")
assert posted["toggle"] is True
assert proj["spectrumProjectId"] == "sp-1"
def test_ensure_spectrum_enabled_skips_toggle_when_on(monkeypatch: pytest.MonkeyPatch) -> None:
posted = {"toggle": False}
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body={"id": "p", "spectrum": True, "spectrumProjectId": "sp-1"})
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
if url.endswith("/spectrum/toggle"):
posted["toggle"] = True
return _FakeResponse(json_body={"success": True})
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
proj = photon_auth.ensure_spectrum_enabled("tok", "p")
assert posted["toggle"] is False
assert proj["spectrumProjectId"] == "sp-1"
def test_regenerate_project_secret(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
assert url.endswith("/regenerate-secret")
return _FakeResponse(json_body={"success": True, "projectSecret": "rotated"})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
assert photon_auth.regenerate_project_secret("tok", "p") == "rotated"
# ---------------------------------------------------------------------------
# Users
def test_create_user_rejects_invalid_phone() -> None:
with pytest.raises(ValueError, match="E.164"):
photon_auth.create_user(
"proj", "secret", phone_number="not-a-number",
)
photon_auth.create_user("tok", "proj", phone_number="not-a-number")
def test_create_user_posts_shared_type(monkeypatch: pytest.MonkeyPatch) -> None:
def test_create_user_posts_dashboard_shape(monkeypatch: pytest.MonkeyPatch) -> None:
captured: Dict[str, Any] = {}
def fake_post(url: str, *, json: Dict[str, Any], auth: tuple, timeout: float) -> _FakeResponse:
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
captured["url"] = url
captured["body"] = json
captured["auth"] = auth
return _FakeResponse(json_body={
"succeed": True,
"data": {
"id": "user-uuid",
"phoneNumber": "+15551234567",
"assignedPhoneNumber": "+15559999999",
},
})
captured["body"] = kwargs.get("json")
captured["headers"] = kwargs.get("headers")
return _FakeResponse(json_body={"success": True, "user": {
"id": "user-uuid", "phoneNumber": "+15551234567",
}})
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
user = photon_auth.create_user(
"proj-id", "proj-secret",
phone_number="+15551234567",
)
assert user["assignedPhoneNumber"] == "+15559999999"
assert captured["auth"] == ("proj-id", "proj-secret")
assert captured["body"]["type"] == "shared"
user = photon_auth.create_user("tok", "proj-id", phone_number="+15551234567")
assert user["id"] == "user-uuid"
assert captured["body"]["phoneNumber"] == "+15551234567"
assert "/projects/proj-id/users/" in captured["url"]
assert captured["headers"]["Authorization"] == "Bearer tok"
assert "/projects/proj-id/spectrum/users" in captured["url"]
def test_register_webhook_surfaces_secret(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_post(url: str, *, json: Dict[str, Any], auth: tuple, timeout: float) -> _FakeResponse:
return _FakeResponse(json_body={
"succeed": True,
"data": {
"id": "wh-uuid",
"webhookUrl": json["webhookUrl"],
"signingSecret": "0" * 64,
},
})
def test_register_user_if_absent_dedup(monkeypatch: pytest.MonkeyPatch) -> None:
posted = {"n": 0}
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body=[{
"id": "u1",
"phoneNumber": "+1 (555) 123-4567",
"assignedPhoneNumber": "+16282679185",
}])
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
posted["n"] += 1
return _FakeResponse(json_body={"success": True, "user": {}})
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
data = photon_auth.register_webhook(
"proj", "secret", webhook_url="https://x.example.com/hook",
# Same number, different formatting — should match and NOT create.
user, created = photon_auth.register_user_if_absent(
"tok", "proj", phone_number="+15551234567",
)
assert data["signingSecret"] == "0" * 64
assert data["webhookUrl"] == "https://x.example.com/hook"
assert created is False
assert user["id"] == "u1"
assert posted["n"] == 0
# The reused user carries the assigned iMessage line ("TEXTS ON").
assert photon_auth.user_assigned_line(user) == "+16282679185"
def test_persist_webhook_signing_secret_writes_env(
tmp_hermes_home: Path,
) -> None:
"""The helper hands the secret to save_env_value, never returns it."""
summary: list = []
response = {
"id": "wh-uuid",
"webhookUrl": "https://x.example.com/hook",
"signingSecret": "ABCDEF1234567890" * 4,
}
ok = photon_auth.persist_webhook_signing_secret(
response, on_summary=summary.append,
def test_user_assigned_line() -> None:
assert (
photon_auth.user_assigned_line({"assignedPhoneNumber": "+16282679185"})
== "+16282679185"
)
assert ok is True
env_path = tmp_hermes_home / ".env"
assert env_path.exists()
env_text = env_path.read_text()
assert "PHOTON_WEBHOOK_SECRET=ABCDEF1234567890" in env_text
# The on_summary callback gets the redacted response + a saved-to path;
# none of those strings should leak the raw secret.
joined = "\n".join(summary)
assert "<redacted>" in joined
assert "ABCDEF1234567890" not in joined
# Own number present but no assignment yet (e.g. freshly created user).
assert photon_auth.user_assigned_line({"phoneNumber": "+15551234567"}) is None
assert photon_auth.user_assigned_line({"assignedPhoneNumber": ""}) is None
assert photon_auth.user_assigned_line({}) is None
assert photon_auth.user_assigned_line(None) is None
def test_persist_webhook_signing_secret_no_secret_no_write(
tmp_hermes_home: Path,
) -> None:
summary: list = []
ok = photon_auth.persist_webhook_signing_secret(
{"id": "wh-uuid", "webhookUrl": "https://x"},
on_summary=summary.append,
def test_register_user_if_absent_creates(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body=[])
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body={"success": True, "user": {"id": "u-new"}})
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
user, created = photon_auth.register_user_if_absent(
"tok", "proj", phone_number="+15551234567",
)
assert ok is False
# No env file written; summary callback still received the redacted
# response (without a signingSecret key, nothing to redact).
assert not (tmp_hermes_home / ".env").exists()
assert created is True
assert user["id"] == "u-new"
def test_credential_summary_returns_only_display_strings(
tmp_hermes_home: Path,
# ---------------------------------------------------------------------------
# Lines (assigned number)
def test_get_imessage_line_returns_existing(monkeypatch: pytest.MonkeyPatch) -> None:
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body=[
{"id": "l1", "platform": "imessage", "phoneNumber": "+15559999999", "status": "active"},
])
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
line = photon_auth.get_imessage_line("tok", "proj")
assert line is not None and line["phoneNumber"] == "+15559999999"
def test_get_imessage_line_provisions_when_missing(monkeypatch: pytest.MonkeyPatch) -> None:
added = {"n": 0}
def fake_get(url: str, **kwargs: Any) -> _FakeResponse:
return _FakeResponse(json_body=[])
def fake_post(url: str, **kwargs: Any) -> _FakeResponse:
added["n"] += 1
assert kwargs.get("json", {}).get("platform") == "imessage"
return _FakeResponse(json_body={"success": True, "line": {
"id": "l-new", "platform": "imessage", "phoneNumber": "+15558888888",
}})
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
line = photon_auth.get_imessage_line("tok", "proj")
assert added["n"] == 1
assert line["phoneNumber"] == "+15558888888"
# ---------------------------------------------------------------------------
# Credential summary (no secret leakage)
def test_credential_summary_no_secret_leak(
tmp_hermes_home: Path, monkeypatch: pytest.MonkeyPatch,
) -> None:
"""credential_summary must not leak raw token/secret material."""
monkeypatch.setattr(photon_auth, "_persist_runtime_env", lambda *a, **k: None)
photon_auth.store_photon_token("token-aaaaaaaaaaaaaaaa")
photon_auth.store_project_credentials("proj-uuid", "secret-bbbbbbbbbbb")
photon_auth.store_project_credentials(
spectrum_project_id="sp-uuid",
project_secret="secret-bbbbbbbbbbb",
dashboard_project_id="dash-uuid",
)
summary = photon_auth.credential_summary()
blob = "\n".join(summary.values())
assert "token-aaaa" not in blob
assert "secret-bbbb" not in blob
assert summary["device_token"].startswith("")
assert summary["project_key"].startswith("")
assert summary["project_id"] == "proj-uuid"
assert summary["spectrum_project_id"] == "sp-uuid"
assert summary["dashboard_project_id"] == "dash-uuid"
def test_print_credential_summary_emits_only_display_strings(
tmp_hermes_home: Path,
# ---------------------------------------------------------------------------
# Device-token candidate extraction + dashboard validation.
def test_device_response_candidates_covers_known_shapes() -> None:
candidates = photon_auth._device_response_token_candidates(
{
"access_token": "tok-snake",
"accessToken": "tok-camel",
"data": {"access_token": "tok-data"},
},
headers={"set-auth-token": "Bearer tok-header"},
)
by_source = {c.source: c.token for c in candidates}
assert by_source["access_token"] == "tok-snake"
assert by_source["accessToken"] == "tok-camel"
assert by_source["data.access_token"] == "tok-data"
# "Bearer " prefix is stripped from the header value.
assert by_source["set-auth-token"] == "tok-header"
def test_device_response_candidates_dedupes() -> None:
candidates = photon_auth._device_response_token_candidates(
{"access_token": "same", "accessToken": "same"},
)
assert [c.token for c in candidates] == ["same"]
def test_validate_photon_token_rejects_unrecognized_session(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""The emit callback must never receive raw credential bytes."""
photon_auth.store_photon_token("token-aaaaaaaaaaaaaaaa")
photon_auth.store_project_credentials("proj-uuid", "secret-bbbbbbbbbbb")
lines: list = []
photon_auth.print_credential_summary(lines.append)
blob = "\n".join(lines)
assert "token-aaaa" not in blob
assert "secret-bbbb" not in blob
assert "✓ stored" in blob # device token line
assert "proj-uuid" in blob # project id is intentionally surfaced
# Header is always emitted
assert any("Photon iMessage status" in line for line in lines)
def fake_get(url: str, *, headers: Dict[str, str], timeout: float) -> _FakeResponse:
if url.endswith("/api/auth/get-session"):
return _FakeResponse(json_body={}) # no "user" key
return _FakeResponse(json_body=[])
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
with pytest.raises(photon_auth.PhotonDashboardAuthError):
photon_auth.validate_photon_token("some-token")
def test_validate_photon_token_rejects_project_api_denial(
monkeypatch: pytest.MonkeyPatch,
) -> None:
def fake_get(url: str, *, headers: Dict[str, str], timeout: float) -> _FakeResponse:
if url.endswith("/api/auth/get-session"):
return _FakeResponse(json_body={"user": {"id": "u1"}})
return _FakeResponse(status=403) # project API rejects
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
with pytest.raises(photon_auth.PhotonDashboardAuthError):
photon_auth.validate_photon_token("some-token")
def test_login_device_flow_validates_before_persisting(
tmp_hermes_home: Path, monkeypatch: pytest.MonkeyPatch,
) -> None:
def fake_post(url: str, *, json: Dict[str, Any], timeout: float) -> _FakeResponse:
if url.endswith("/api/auth/device/code"):
return _FakeResponse(json_body={
"device_code": "dev", "user_code": "AAAA",
"verification_uri": "https://app.photon.codes/device",
"verification_uri_complete": None,
"expires_in": 600, "interval": 0,
})
# device/token approval
return _FakeResponse(json_body={"access_token": "good-token"})
def fake_get(url: str, *, headers: Dict[str, str], timeout: float) -> _FakeResponse:
if url.endswith("/api/auth/get-session"):
return _FakeResponse(json_body={"user": {"id": "u1"}})
return _FakeResponse(json_body=[]) # projects OK
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
token = photon_auth.login_device_flow(open_browser=False)
assert token == "good-token"
assert photon_auth.load_photon_token() == "good-token"
def test_login_device_flow_raises_when_token_invalid(
tmp_hermes_home: Path, monkeypatch: pytest.MonkeyPatch,
) -> None:
def fake_post(url: str, *, json: Dict[str, Any], timeout: float) -> _FakeResponse:
if url.endswith("/api/auth/device/code"):
return _FakeResponse(json_body={
"device_code": "dev", "user_code": "AAAA",
"verification_uri": "https://app.photon.codes/device",
"verification_uri_complete": None,
"expires_in": 600, "interval": 0,
})
return _FakeResponse(json_body={"access_token": "bad-token"})
def fake_get(url: str, *, headers: Dict[str, str], timeout: float) -> _FakeResponse:
return _FakeResponse(status=401) # session lookup rejects
monkeypatch.setattr(photon_auth.httpx, "post", fake_post)
monkeypatch.setattr(photon_auth.httpx, "get", fake_get)
with pytest.raises(photon_auth.PhotonDashboardAuthError):
photon_auth.login_device_flow(open_browser=False)
# A token that failed validation must never be persisted.
assert photon_auth.load_photon_token() is None
+159 -68
View File
@@ -1,12 +1,15 @@
"""Inbound dispatch + dedup tests for PhotonAdapter.
These tests bypass the aiohttp server they call ``_dispatch_inbound``
and ``_is_duplicate`` directly. That keeps them fast and means we can
exercise the message-shape parsing logic without binding ports.
These bypass the loopback HTTP stream they call ``_dispatch_inbound`` /
``_on_inbound_line`` / ``_is_duplicate`` directly, exercising the
sidecar-event parsing without spawning the Node sidecar or binding ports.
"""
from __future__ import annotations
from typing import List
import base64
import json
from pathlib import Path
from typing import Any, Dict, List
import pytest
@@ -16,38 +19,39 @@ from plugins.platforms.photon.adapter import PhotonAdapter
def _make_adapter(monkeypatch: pytest.MonkeyPatch) -> PhotonAdapter:
# Avoid touching real auth.json / env.
monkeypatch.setenv("PHOTON_PROJECT_ID", "test-project-id")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "test-project-secret")
monkeypatch.delenv("PHOTON_WEBHOOK_SECRET", raising=False)
cfg = PlatformConfig(enabled=True, token="", extra={})
return PhotonAdapter(cfg)
@pytest.mark.asyncio
async def test_dispatch_text_dm(monkeypatch: pytest.MonkeyPatch) -> None:
adapter = _make_adapter(monkeypatch)
def _capture(adapter: PhotonAdapter, monkeypatch: pytest.MonkeyPatch) -> List[MessageEvent]:
captured: List[MessageEvent] = []
async def fake_handle(event: MessageEvent) -> None:
captured.append(event)
monkeypatch.setattr(adapter, "handle_message", fake_handle)
return captured
payload = {
"event": "messages",
"space": {"id": "any;-;+15551234567", "platform": "iMessage"},
"message": {
"id": "spc-msg-abc",
"platform": "iMessage",
"direction": "inbound",
"timestamp": "2026-05-14T19:06:32.000Z",
"sender": {"id": "+15551234567", "platform": "iMessage"},
"space": {"id": "any;-;+15551234567", "platform": "iMessage"},
"content": {"type": "text", "text": "hello world"},
},
def _dm_event(text: str, msg_id: str = "spc-msg-abc") -> Dict[str, Any]:
return {
"messageId": msg_id,
"platform": "iMessage",
"space": {"id": "+15551234567", "type": "dm", "phone": "+15551234567"},
"sender": {"id": "+15551234567"},
"content": {"type": "text", "text": text},
"timestamp": "2026-05-14T19:06:32.000Z",
}
await adapter._dispatch_inbound(payload)
@pytest.mark.asyncio
async def test_dispatch_text_dm(monkeypatch: pytest.MonkeyPatch) -> None:
adapter = _make_adapter(monkeypatch)
captured = _capture(adapter, monkeypatch)
await adapter._dispatch_inbound(_dm_event("hello world"))
assert len(captured) == 1
event = captured[0]
@@ -57,70 +61,157 @@ async def test_dispatch_text_dm(monkeypatch: pytest.MonkeyPatch) -> None:
src = event.source
assert src is not None
assert src.platform == Platform("photon")
assert src.chat_id == "any;-;+15551234567"
assert src.chat_id == "+15551234567"
assert src.chat_type == "dm"
assert src.user_id == "+15551234567"
@pytest.mark.asyncio
async def test_dispatch_group_id_detected(monkeypatch: pytest.MonkeyPatch) -> None:
async def test_dispatch_group_type(monkeypatch: pytest.MonkeyPatch) -> None:
adapter = _make_adapter(monkeypatch)
captured: List[MessageEvent] = []
captured = _capture(adapter, monkeypatch)
async def fake_handle(event: MessageEvent) -> None:
captured.append(event)
monkeypatch.setattr(adapter, "handle_message", fake_handle)
payload = {
"event": "messages",
"space": {"id": "any;+;group-guid-xyz", "platform": "iMessage"},
"message": {
"id": "spc-msg-grp",
"timestamp": "2026-05-14T19:06:32.000Z",
"sender": {"id": "+15551234567"},
"space": {"id": "any;+;group-guid-xyz"},
"content": {"type": "text", "text": "hi group"},
},
event = {
"messageId": "spc-msg-grp",
"space": {"id": "group-guid-xyz", "type": "group", "phone": None},
"sender": {"id": "+15551234567"},
"content": {"type": "text", "text": "hi group"},
"timestamp": "2026-05-14T19:06:32.000Z",
}
await adapter._dispatch_inbound(payload)
await adapter._dispatch_inbound(event)
assert captured[0].source.chat_type == "group"
# A real 1x1 transparent PNG (passes base.py's _looks_like_image magic check).
_PNG_1X1_B64 = (
"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNkYPhf"
"DwAChwGA60e6kgAAAABJRU5ErkJggg=="
)
def _attachment_event(
content: Dict[str, Any], msg_id: str = "spc-msg-att"
) -> Dict[str, Any]:
return {
"messageId": msg_id,
"space": {"id": "+15551234567", "type": "dm", "phone": "+15551234567"},
"sender": {"id": "+15551234567"},
"content": {"type": "attachment", **content},
"timestamp": "2026-05-14T19:06:32.000Z",
}
@pytest.mark.asyncio
async def test_dispatch_attachment_surfaces_marker(
async def test_dispatch_attachment_without_bytes_surfaces_marker(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""No inline ``data`` (over cap / failed sidecar read) -> text marker, no media."""
adapter = _make_adapter(monkeypatch)
captured = _capture(adapter, monkeypatch)
event = _attachment_event(
{"name": "IMG_4127.HEIC", "mimeType": "image/heic", "size": 12345}
)
await adapter._dispatch_inbound(event)
assert len(captured) == 1
ev = captured[0]
assert "Photon attachment received" in ev.text
assert "IMG_4127.HEIC" in ev.text
assert ev.message_type == MessageType.PHOTO
assert ev.media_urls == []
assert ev.media_types == []
@pytest.mark.asyncio
async def test_dispatch_attachment_downloads_image(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Inline base64 image bytes are decoded, cached, and exposed as media."""
adapter = _make_adapter(monkeypatch)
captured = _capture(adapter, monkeypatch)
raw = base64.b64decode(_PNG_1X1_B64)
event = _attachment_event(
{
"name": "photo.png",
"mimeType": "image/png",
"size": len(raw),
"data": _PNG_1X1_B64,
"encoding": "base64",
}
)
await adapter._dispatch_inbound(event)
assert len(captured) == 1
ev = captured[0]
assert ev.message_type == MessageType.PHOTO
assert ev.media_types == ["image/png"]
assert len(ev.media_urls) == 1
cached = Path(ev.media_urls[0])
try:
assert cached.is_file()
assert cached.read_bytes() == raw
assert ev.text == "(attachment)"
finally:
cached.unlink(missing_ok=True)
@pytest.mark.asyncio
async def test_dispatch_attachment_downloads_document(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Non-image attachments route through the document cache as DOCUMENT."""
adapter = _make_adapter(monkeypatch)
captured = _capture(adapter, monkeypatch)
raw = b"%PDF-1.4 hermes test document"
event = _attachment_event(
{
"name": "report.pdf",
"mimeType": "application/pdf",
"size": len(raw),
"data": base64.b64encode(raw).decode("ascii"),
"encoding": "base64",
}
)
await adapter._dispatch_inbound(event)
assert len(captured) == 1
ev = captured[0]
assert ev.message_type == MessageType.DOCUMENT
assert ev.media_types == ["application/pdf"]
assert len(ev.media_urls) == 1
cached = Path(ev.media_urls[0])
try:
assert cached.is_file()
assert cached.read_bytes() == raw
assert ev.text == "(attachment)"
finally:
cached.unlink(missing_ok=True)
@pytest.mark.asyncio
async def test_on_inbound_line_dispatches_and_dedups(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
captured: List[MessageEvent] = []
captured = _capture(adapter, monkeypatch)
async def fake_handle(event: MessageEvent) -> None:
captured.append(event)
line = json.dumps(_dm_event("ping", msg_id="dup-1"))
await adapter._on_inbound_line(line)
await adapter._on_inbound_line(line) # same messageId -> deduped
monkeypatch.setattr(adapter, "handle_message", fake_handle)
payload = {
"event": "messages",
"message": {
"id": "spc-msg-att",
"timestamp": "2026-05-14T19:06:32.000Z",
"sender": {"id": "+15551234567"},
"space": {"id": "any;-;+15551234567"},
"content": {
"type": "attachment",
"name": "IMG_4127.HEIC",
"mimeType": "image/heic",
"size": 12345,
},
},
}
await adapter._dispatch_inbound(payload)
assert len(captured) == 1
event = captured[0]
# Attachment carries metadata marker; mime → MessageType.PHOTO.
assert "Photon attachment received" in event.text
assert "IMG_4127.HEIC" in event.text
assert event.message_type == MessageType.PHOTO
assert captured[0].text == "ping"
@pytest.mark.asyncio
async def test_on_inbound_line_ignores_bad_json(monkeypatch: pytest.MonkeyPatch) -> None:
adapter = _make_adapter(monkeypatch)
captured = _capture(adapter, monkeypatch)
await adapter._on_inbound_line("{not json")
assert captured == []
def test_is_duplicate_window(monkeypatch: pytest.MonkeyPatch) -> None:
@@ -22,7 +22,6 @@ from plugins.platforms.photon.adapter import PhotonAdapter
def _make_adapter(monkeypatch: pytest.MonkeyPatch, extra: dict | None = None) -> PhotonAdapter:
monkeypatch.setenv("PHOTON_PROJECT_ID", "test-project-id")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "test-project-secret")
monkeypatch.delenv("PHOTON_WEBHOOK_SECRET", raising=False)
monkeypatch.delenv("PHOTON_REQUIRE_MENTION", raising=False)
monkeypatch.delenv("PHOTON_MENTION_PATTERNS", raising=False)
cfg = PlatformConfig(enabled=True, token="", extra=extra or {})
@@ -31,27 +30,21 @@ def _make_adapter(monkeypatch: pytest.MonkeyPatch, extra: dict | None = None) ->
def _group_payload(text: str) -> dict:
return {
"event": "messages",
"message": {
"id": f"grp-{abs(hash(text))}",
"timestamp": "2026-05-14T19:06:32.000Z",
"sender": {"id": "+15551234567"},
"space": {"id": "any;+;group-guid-xyz"},
"content": {"type": "text", "text": text},
},
"messageId": f"grp-{abs(hash(text))}",
"space": {"id": "group-guid-xyz", "type": "group", "phone": None},
"sender": {"id": "+15551234567"},
"content": {"type": "text", "text": text},
"timestamp": "2026-05-14T19:06:32.000Z",
}
def _dm_payload(text: str) -> dict:
return {
"event": "messages",
"message": {
"id": f"dm-{abs(hash(text))}",
"timestamp": "2026-05-14T19:06:32.000Z",
"sender": {"id": "+15551234567"},
"space": {"id": "any;-;+15551234567"},
"content": {"type": "text", "text": text},
},
"messageId": f"dm-{abs(hash(text))}",
"space": {"id": "+15551234567", "type": "dm", "phone": "+15551234567"},
"sender": {"id": "+15551234567"},
"content": {"type": "text", "text": text},
"timestamp": "2026-05-14T19:06:32.000Z",
}
@@ -126,7 +119,6 @@ def test_custom_mention_patterns_from_config(monkeypatch: pytest.MonkeyPatch) ->
def test_mention_patterns_env_comma_separated(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setenv("PHOTON_PROJECT_ID", "test-project-id")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "test-project-secret")
monkeypatch.delenv("PHOTON_WEBHOOK_SECRET", raising=False)
monkeypatch.setenv("PHOTON_REQUIRE_MENTION", "true")
monkeypatch.setenv("PHOTON_MENTION_PATTERNS", r"bot\b, assistant\b")
cfg = PlatformConfig(enabled=True, token="", extra={})
@@ -0,0 +1,69 @@
"""Tests for `hermes photon setup`'s access auto-configuration.
`_autoconfigure_access` allowlists the operator and points the cron home
channel at their DM, writing to the per-test ~/.hermes/.env (the hermetic
HERMES_HOME fixture isolates this). It must fill only unset keys so a re-run
never clobbers a hand-tuned allowlist.
"""
from __future__ import annotations
import pytest
from hermes_cli.config import get_env_value, save_env_value
from plugins.platforms.photon.adapter import _env_enablement
from plugins.platforms.photon import cli
def test_autoconfigure_access_fills_unset(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.delenv("PHOTON_ALLOWED_USERS", raising=False)
monkeypatch.delenv("PHOTON_HOME_CHANNEL", raising=False)
cli._autoconfigure_access("+15551234567")
assert get_env_value("PHOTON_ALLOWED_USERS") == "+15551234567"
assert get_env_value("PHOTON_HOME_CHANNEL") == "+15551234567"
def test_autoconfigure_access_preserves_existing_allowlist(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.delenv("PHOTON_ALLOWED_USERS", raising=False)
monkeypatch.delenv("PHOTON_HOME_CHANNEL", raising=False)
# A hand-tuned allowlist already in place must survive a setup re-run.
save_env_value("PHOTON_ALLOWED_USERS", "+19998887777,+15551112222")
cli._autoconfigure_access("+15551234567")
assert get_env_value("PHOTON_ALLOWED_USERS") == "+19998887777,+15551112222"
# The still-unset home channel is filled.
assert get_env_value("PHOTON_HOME_CHANNEL") == "+15551234567"
def test_env_enablement_seeds_home_channel(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setenv("PHOTON_PROJECT_ID", "project_123")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "secret_123")
monkeypatch.setenv("PHOTON_HOME_CHANNEL", "+15551234567")
monkeypatch.setenv("PHOTON_HOME_CHANNEL_NAME", "Primary DM")
seed = _env_enablement()
assert seed is not None
assert seed["home_channel"] == {
"chat_id": "+15551234567",
"name": "Primary DM",
}
def test_env_enablement_home_channel_defaults_name(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setenv("PHOTON_PROJECT_ID", "project_123")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "secret_123")
monkeypatch.setenv("PHOTON_HOME_CHANNEL", "+15551234567")
monkeypatch.delenv("PHOTON_HOME_CHANNEL_NAME", raising=False)
seed = _env_enablement()
assert seed is not None
assert seed["home_channel"] == {
"chat_id": "+15551234567",
"name": "Home",
}
@@ -1,95 +0,0 @@
"""Signature verification tests for the Photon webhook receiver."""
from __future__ import annotations
import hashlib
import hmac
import time
import pytest
from plugins.platforms.photon.adapter import verify_signature
def _sign(secret: str, body: bytes, ts: int) -> str:
return "v0=" + hmac.new(
secret.encode(), f"v0:{ts}:".encode() + body, hashlib.sha256,
).hexdigest()
def test_accepts_valid_signature() -> None:
secret = "topsecret-32chars-or-whatever"
body = b'{"event":"messages"}'
ts = int(time.time())
sig = _sign(secret, body, ts)
assert verify_signature(
body=body, timestamp_header=str(ts), signature_header=sig,
signing_secret=secret,
)
def test_rejects_tampered_body() -> None:
secret = "s"
body = b'{"event":"messages"}'
ts = int(time.time())
sig = _sign(secret, body, ts)
assert not verify_signature(
body=body + b" tamper", timestamp_header=str(ts),
signature_header=sig, signing_secret=secret,
)
def test_rejects_wrong_secret() -> None:
body = b"x"
ts = int(time.time())
sig = _sign("right", body, ts)
assert not verify_signature(
body=body, timestamp_header=str(ts), signature_header=sig,
signing_secret="wrong",
)
def test_rejects_drifted_timestamp() -> None:
secret = "s"
body = b"x"
ts = int(time.time()) - 3600 # 1h old; drift window is 5 min
sig = _sign(secret, body, ts)
assert not verify_signature(
body=body, timestamp_header=str(ts), signature_header=sig,
signing_secret=secret,
)
def test_rejects_missing_v0_prefix() -> None:
secret = "s"
body = b"x"
ts = int(time.time())
raw_hex = hmac.new(
secret.encode(), f"v0:{ts}:".encode() + body, hashlib.sha256,
).hexdigest()
# Strip the "v0=" prefix — verify_signature must reject.
assert not verify_signature(
body=body, timestamp_header=str(ts), signature_header=raw_hex,
signing_secret=secret,
)
def test_rejects_empty_inputs() -> None:
assert not verify_signature(
body=b"x", timestamp_header="", signature_header="v0=abc",
signing_secret="s",
)
assert not verify_signature(
body=b"x", timestamp_header="123", signature_header="",
signing_secret="s",
)
assert not verify_signature(
body=b"x", timestamp_header="123", signature_header="v0=abc",
signing_secret="",
)
def test_rejects_non_integer_timestamp() -> None:
assert not verify_signature(
body=b"x", timestamp_header="not-an-int",
signature_header="v0=abc", signing_secret="s",
)
@@ -0,0 +1,20 @@
"""Parser-only tests for send_message targets.
These stay separate from ``test_send_message_tool.py`` because that module
skips wholesale when optional Telegram dependencies are not installed.
"""
from tools.send_message_tool import _parse_target_ref
def test_photon_e164_target_is_explicit() -> None:
chat_id, thread_id, is_explicit = _parse_target_ref("photon", "+15551234567")
assert chat_id == "+15551234567"
assert thread_id is None
assert is_explicit is True
def test_e164_target_still_requires_phone_platform() -> None:
assert _parse_target_ref("matrix", "+15551234567")[2] is False
+5
View File
@@ -1199,6 +1199,11 @@ class TestParseTargetRefE164:
assert chat_id == "+15551234567"
assert is_explicit is True
def test_photon_e164_is_explicit(self):
chat_id, _, is_explicit = _parse_target_ref("photon", "+15551234567")
assert chat_id == "+15551234567"
assert is_explicit is True
def test_signal_bare_digits_still_work(self):
"""Bare digit strings continue to match the generic numeric branch."""
chat_id, _, is_explicit = _parse_target_ref("signal", "15551234567")
+1 -1
View File
@@ -38,7 +38,7 @@ _NUMERIC_TOPIC_RE = _TELEGRAM_TOPIC_TARGET_RE
# below and falls through to channel-name resolution, which has no way to
# resolve a raw phone number. Keeping the '+' preserves the E.164 form that
# downstream adapters (signal, etc.) expect.
_PHONE_PLATFORMS = frozenset({"signal", "sms", "whatsapp"})
_PHONE_PLATFORMS = frozenset({"photon", "signal", "sms", "whatsapp"})
_E164_TARGET_RE = re.compile(r"^\s*\+(\d{7,15})\s*$")
# Email addresses — a valid email like "user@domain.com" should be treated as
# an explicit target for the email platform, not fall through to channel-name
+78 -1
View File
@@ -175,6 +175,7 @@ _LONG_HANDLERS = frozenset(
{
"browser.manage",
"cli.exec",
"plugins.manage",
"session.branch",
"session.compress",
"session.resume",
@@ -9020,7 +9021,83 @@ def _(rid, params: dict) -> dict:
return _err(rid, 5025, str(e))
# ── Methods: shell ───────────────────────────────────────────────────
@method("plugins.manage")
def _(rid, params: dict) -> dict:
"""List installed plugins with activation state, or toggle one on/off.
Backs the TUI Plugins Hub. Uses the same disk-discovery + enable/disable
primitives as ``hermes plugins`` / the dashboard, so the three surfaces
agree on what's installed and what's enabled.
Actions:
- ``list`` {"plugins": [{name, version, description, source,
status}], "user_count": N, "bundled_count": M}
- ``toggle`` flip ``name`` based on ``enable`` (bool). Returns the
refreshed row plus {"ok", "unchanged"}.
"""
action = params.get("action", "list")
try:
from hermes_cli.plugins_cmd import (
_discover_all_plugins,
_get_disabled_set,
_get_enabled_set,
_plugin_status,
)
def _rows():
enabled = _get_enabled_set()
disabled = _get_disabled_set()
out = []
for name, version, desc, source, _dir, key in sorted(
_discover_all_plugins()
):
out.append(
{
"name": name,
"version": str(version or ""),
"description": desc or "",
"source": source,
"status": _plugin_status(name, enabled, disabled, key=key),
}
)
return out
if action == "list":
rows = _rows()
user_count = sum(1 for r in rows if r["source"] != "bundled")
return _ok(
rid,
{
"plugins": rows,
"user_count": user_count,
"bundled_count": len(rows) - user_count,
},
)
if action == "toggle":
from hermes_cli.plugins_cmd import dashboard_set_agent_plugin_enabled
name = (params.get("name") or "").strip()
if not name:
return _err(rid, 4019, "plugins.toggle requires a 'name'")
enable = bool(params.get("enable"))
result = dashboard_set_agent_plugin_enabled(name, enabled=enable)
if not result.get("ok"):
return _err(rid, 5026, result.get("error") or "toggle failed")
row = next((r for r in _rows() if r["name"] == name), None)
return _ok(
rid,
{
"ok": True,
"unchanged": bool(result.get("unchanged")),
"name": name,
"plugin": row,
},
)
return _err(rid, 4017, f"unknown plugins action: {action}")
except Exception as e:
return _err(rid, 5026, str(e))
@method("shell.exec")
@@ -0,0 +1,91 @@
import { EventEmitter } from 'events'
import React, { useContext, useEffect } from 'react'
import { describe, expect, it } from 'vitest'
import StdinContext from './components/StdinContext.js'
import Text from './components/Text.js'
import Ink from './ink.js'
import { DISABLE_MOUSE_TRACKING } from './termio/dec.js'
class FakeTty extends EventEmitter {
chunks: string[] = []
columns = 80
rows = 24
isTTY = true
isRaw = false
ref(): void {}
unref(): void {}
read(): null {
return null
}
setEncoding(): this {
return this
}
setRawMode(mode: boolean): this {
this.isRaw = mode
return this
}
write(chunk: string | Uint8Array, cb?: (err?: Error | null) => void): boolean {
this.chunks.push(typeof chunk === 'string' ? chunk : Buffer.from(chunk).toString('utf8'))
cb?.()
return true
}
}
const tick = () => new Promise<void>(resolve => setImmediate(resolve))
// A child that grabs the last useInput consumer's raw-mode toggle. Mounting
// enables raw mode (count 0→1); unmounting disables it (count 1→0), which is
// the teardown path that must DISABLE_MOUSE_TRACKING so DEC 1003 hover can't
// leak as cooked-mode `35;col;row M` text over the prompt.
function RawModeConsumer({ active }: { active: boolean }) {
const { setRawMode, isRawModeSupported } = useContext(StdinContext)
useEffect(() => {
if (!active || !isRawModeSupported) {
return
}
setRawMode(true)
return () => setRawMode(false)
}, [active, isRawModeSupported, setRawMode])
return React.createElement(Text, null, 'x')
}
describe('App raw-mode teardown', () => {
it('disables mouse tracking when the last raw-mode consumer detaches', async () => {
const stdout = new FakeTty()
const stdin = new FakeTty()
const stderr = new FakeTty()
const ink = new Ink({
exitOnCtrlC: false,
patchConsole: false,
stderr: stderr as unknown as NodeJS.WriteStream,
stdin: stdin as unknown as NodeJS.ReadStream,
stdout: stdout as unknown as NodeJS.WriteStream
})
// Mouse tracking is asserted on the alt screen; the teardown path lives in
// App, independent of who enabled tracking.
ink.setAltScreenActive(true, 'all')
ink.render(React.createElement(RawModeConsumer, { active: true }))
ink.onRender()
await tick()
expect(stdin.isRaw).toBe(true)
stdout.chunks = []
// Drop the consumer → raw-mode count hits 0 → teardown runs.
ink.render(React.createElement(RawModeConsumer, { active: false }))
ink.onRender()
await tick()
expect(stdin.isRaw).toBe(false)
expect(stdout.chunks.join('')).toContain(DISABLE_MOUSE_TRACKING)
ink.unmount()
})
})
@@ -9,6 +9,7 @@ import type { DOMElement } from '../dom.js'
import { EventEmitter } from '../events/emitter.js'
import { InputEvent } from '../events/input-event.js'
import { TerminalFocusEvent } from '../events/terminal-focus-event.js'
import instances from '../instances.js'
import {
INITIAL_STATE,
type ParsedInput,
@@ -309,6 +310,21 @@ export default class App extends PureComponent<Props, State> {
}
})
})
// Re-assert mouse tracking on raw-mode re-entry. <AlternateScreen>
// owns the initial enable, but its effect only re-runs on a
// mode/writeRaw change — NOT on a raw-mode bounce (count 1→0→1, e.g.
// an overlay that briefly drops the last useInput consumer). The
// teardown above now DISABLE_MOUSE_TRACKING's to stop the cooked-echo
// leak, so without this the terminal would be left with tracking off
// and the mouse silently dead until the next stdin-gap/resize
// re-assert. reassertTerminalModes() is gated on altScreenActive and
// idempotent, so it's a no-op when there's nothing to restore.
// Deferred (same setImmediate discipline as the XTVERSION probe) so it
// lands after any alt-screen enable writes in this render cycle.
setImmediate(() => {
instances.get(this.props.stdout)?.reassertTerminalModes()
})
}
this.rawModeEnabledCount++
@@ -324,6 +340,15 @@ export default class App extends PureComponent<Props, State> {
this.props.stdout.write(DFE)
// Disable bracketed paste mode
this.props.stdout.write(DBP)
// Disable mouse tracking. Tracking is asserted by <AlternateScreen> /
// the Ink instance, NOT here — but dropping raw mode + detaching the
// readable listener while DEC 1003 hover stays on means the terminal
// falls back to cooked-mode echo and every mouse move leaks as text
// (`35;col;row M` shards over the prompt). Same hazard handleSuspend()
// already guards against; this teardown path missed it. Idempotent
// (no-op if tracking was never on), and re-enabling raw mode below
// re-asserts tracking so a transient drop→re-add round-trips cleanly.
this.props.stdout.write(DISABLE_MOUSE_TRACKING)
stdin.setRawMode(false)
stdin.removeListener('readable', this.handleReadable)
stdin.unref()
+1
View File
@@ -93,6 +93,7 @@ export interface OverlayState {
confirm: ConfirmReq | null
modelPicker: boolean
pager: null | PagerState
pluginsHub: boolean
secret: null | SecretReq
sessions: boolean
skillsHub: boolean
+6 -2
View File
@@ -10,6 +10,7 @@ const buildOverlayState = (): OverlayState => ({
confirm: null,
modelPicker: false,
pager: null,
pluginsHub: false,
secret: null,
sessions: false,
skillsHub: false,
@@ -20,8 +21,10 @@ export const $overlayState = atom<OverlayState>(buildOverlayState())
export const $isBlocked = computed(
$overlayState,
({ agents, approval, clarify, confirm, modelPicker, pager, secret, sessions, skillsHub, sudo }) =>
Boolean(agents || approval || clarify || confirm || modelPicker || pager || secret || sessions || skillsHub || sudo)
({ agents, approval, clarify, confirm, modelPicker, pager, pluginsHub, secret, sessions, skillsHub, sudo }) =>
Boolean(
agents || approval || clarify || confirm || modelPicker || pager || pluginsHub || secret || sessions || skillsHub || sudo
)
)
export const getOverlayState = () => $overlayState.get()
@@ -46,6 +49,7 @@ export const resetFlowOverlays = () =>
agents: $overlayState.get().agents,
agentsInitialHistoryIndex: $overlayState.get().agentsInitialHistoryIndex,
modelPicker: $overlayState.get().modelPicker,
pluginsHub: $overlayState.get().pluginsHub,
sessions: $overlayState.get().sessions,
skillsHub: $overlayState.get().skillsHub
})
+28
View File
@@ -652,6 +652,34 @@ export const opsCommands: SlashCommand[] = [
}
},
{
help: 'view & toggle plugins (no arg opens the hub; enable/disable <name> for direct toggle)',
name: 'plugins',
run: (arg, ctx, cmd) => {
// No argument → open the interactive Plugins Hub overlay. Any
// subcommand (enable/disable/list/install/…) falls through to the
// text slash worker so it stays at parity with `hermes plugins`.
if (!arg.trim()) {
return patchOverlayState({ pluginsHub: true })
}
ctx.gateway.gw
.request<SlashExecResponse>('slash.exec', { command: cmd.slice(1), session_id: ctx.sid })
.then(r => {
if (ctx.stale()) {
return
}
const body = r?.output || '/plugins: no output'
const text = r?.warning ? `warning: ${r.warning}\n${body}` : body
const long = text.length > 180 || text.split('\n').filter(Boolean).length > 2
long ? ctx.transcript.page(text, 'Plugins') : ctx.transcript.sys(text)
})
.catch(ctx.guardedErr)
}
},
{
help: 'enable or disable tools (client-side history reset on change)',
name: 'tools',
+4
View File
@@ -151,6 +151,10 @@ export function useInputHandlers(ctx: InputHandlerContext): InputHandlerResult {
return patchOverlayState({ skillsHub: false })
}
if (overlay.pluginsHub) {
return patchOverlayState({ pluginsHub: false })
}
if (overlay.sessions) {
return patchOverlayState({ sessions: false })
}
+8
View File
@@ -11,6 +11,7 @@ import { FloatBox } from './appChrome.js'
import { MaskedPrompt } from './maskedPrompt.js'
import { ModelPicker } from './modelPicker.js'
import { OverlayHint } from './overlayControls.js'
import { PluginsHub } from './pluginsHub.js'
import { ApprovalPrompt, ClarifyPrompt, ConfirmPrompt } from './prompts.js'
import { SkillsHub } from './skillsHub.js'
@@ -125,6 +126,7 @@ export function FloatingOverlays({
overlay.pager ||
overlay.sessions ||
overlay.skillsHub ||
overlay.pluginsHub ||
completions.length
if (!hasAny) {
@@ -174,6 +176,12 @@ export function FloatingOverlays({
</FloatBox>
)}
{overlay.pluginsHub && (
<FloatBox color={theme.color.border}>
<PluginsHub gw={gw} onClose={() => patchOverlayState({ pluginsHub: false })} t={theme} />
</FloatBox>
)}
{overlay.pager && (
<FloatBox color={theme.color.border}>
<Box flexDirection="column" paddingX={1} paddingY={1}>
+238
View File
@@ -0,0 +1,238 @@
import { Box, Text, useInput, useStdout } from '@hermes/ink'
import { useEffect, useState } from 'react'
import type { GatewayClient } from '../gatewayClient.js'
import { rpcErrorMessage } from '../lib/rpc.js'
import type { Theme } from '../theme.js'
import { OverlayHint, useOverlayKeys, windowItems, windowOffset } from './overlayControls.js'
const VISIBLE = 12
const MIN_WIDTH = 44
const MAX_WIDTH = 96
interface PluginRow {
description?: string
name: string
source?: string
status?: string
version?: string
}
interface PluginsListResponse {
bundled_count?: number
plugins?: PluginRow[]
user_count?: number
}
interface PluginsToggleResponse {
name?: string
ok?: boolean
plugin?: PluginRow
unchanged?: boolean
}
type Scope = 'all' | 'user'
const GLYPH: Record<string, string> = {
disabled: '✗',
enabled: '✓'
}
export function PluginsHub({ gw, onClose, t }: PluginsHubProps) {
const [rows, setRows] = useState<PluginRow[]>([])
const [bundledCount, setBundledCount] = useState(0)
const [userCount, setUserCount] = useState(0)
const [idx, setIdx] = useState(0)
const [scope, setScope] = useState<Scope>('user')
const [busy, setBusy] = useState(false)
const [err, setErr] = useState('')
const [loading, setLoading] = useState(true)
const { stdout } = useStdout()
const width = Math.max(MIN_WIDTH, Math.min(MAX_WIDTH, (stdout?.columns ?? 80) - 6))
const load = () => {
gw.request<PluginsListResponse>('plugins.manage', { action: 'list' })
.then(r => {
setRows(r?.plugins ?? [])
setUserCount(Number(r?.user_count ?? 0))
setBundledCount(Number(r?.bundled_count ?? 0))
setErr('')
setLoading(false)
})
.catch((e: unknown) => {
setErr(rpcErrorMessage(e))
setLoading(false)
})
}
useEffect(load, [gw])
// Default to user plugins; fall back to all when there are none so the
// overlay is never empty when bundled plugins exist.
const visibleRows = scope === 'user' ? rows.filter(r => r.source !== 'bundled') : rows
const effectiveRows = scope === 'user' && !visibleRows.length && rows.length ? rows : visibleRows
const effectiveScope: Scope = effectiveRows === visibleRows ? scope : 'all'
const clampedIdx = Math.min(idx, Math.max(0, effectiveRows.length - 1))
useOverlayKeys({ disabled: busy, onClose })
const toggle = (row: PluginRow) => {
if (busy || !row) {
return
}
const enable = row.status !== 'enabled'
setBusy(true)
setErr('')
gw.request<PluginsToggleResponse>('plugins.manage', { action: 'toggle', enable, name: row.name })
.then(r => {
if (r?.plugin) {
setRows(prev => prev.map(p => (p.name === r.plugin!.name ? r.plugin! : p)))
} else {
load()
}
})
.catch((e: unknown) => setErr(rpcErrorMessage(e)))
.finally(() => setBusy(false))
}
useInput((ch, key) => {
if (busy) {
return
}
const count = effectiveRows.length
if (key.upArrow && clampedIdx > 0) {
setIdx(clampedIdx - 1)
return
}
if (key.downArrow && clampedIdx < count - 1) {
setIdx(clampedIdx + 1)
return
}
// Tab toggles user-only vs all (bundled) scope.
if (key.tab) {
setScope(s => (s === 'user' ? 'all' : 'user'))
setIdx(0)
return
}
if (key.return || ch === ' ') {
const row = effectiveRows[clampedIdx]
if (row) {
toggle(row)
}
return
}
const n = ch === '0' ? 10 : parseInt(ch, 10)
if (!Number.isNaN(n) && n >= 1 && n <= Math.min(10, count)) {
const next = windowOffset(count, clampedIdx, VISIBLE) + n - 1
const row = effectiveRows[next]
if (row) {
setIdx(next)
toggle(row)
}
}
})
if (loading) {
return <Text color={t.color.muted}>loading plugins</Text>
}
if (err && !rows.length) {
return (
<Box flexDirection="column" width={width}>
<Text color={t.color.label}>error: {err}</Text>
<OverlayHint t={t}>Esc/q close</OverlayHint>
</Box>
)
}
if (!rows.length) {
return (
<Box flexDirection="column" width={width}>
<Text bold color={t.color.accent}>
Plugins Hub
</Text>
<Text color={t.color.muted}>no plugins installed</Text>
<Text color={t.color.muted}>install: hermes plugins install owner/repo</Text>
<OverlayHint t={t}>Esc/q close</OverlayHint>
</Box>
)
}
const labels = effectiveRows.map(r => {
const status = r.status ?? 'not enabled'
const glyph = GLYPH[status] ?? '○'
const ver = r.version ? ` v${r.version}` : ''
const src = effectiveScope === 'all' && r.source === 'bundled' ? ' [bundled]' : ''
const state = status === 'enabled' ? '' : ` (${status})`
return `${glyph} ${r.name}${ver}${src}${state}`
})
const { items, offset } = windowItems(labels, clampedIdx, VISIBLE)
const scopeLabel =
effectiveScope === 'user'
? `${userCount} user plugin(s)${bundledCount ? ` · +${bundledCount} bundled (Tab)` : ''}`
: `all ${rows.length} plugins`
return (
<Box flexDirection="column" width={width}>
<Text bold color={t.color.accent}>
Plugins Hub
</Text>
<Text color={t.color.muted}>{scopeLabel}</Text>
{offset > 0 && <Text color={t.color.muted}> {offset} more</Text>}
{items.map((row, i) => {
const lineIdx = offset + i
const active = clampedIdx === lineIdx
return (
<Text
bold={active}
color={active ? t.color.accent : t.color.muted}
inverse={active}
key={effectiveRows[lineIdx]?.name ?? row}
wrap="truncate-end"
>
{active ? '▸ ' : ' '}
{i + 1}. {row}
</Text>
)
})}
{offset + VISIBLE < labels.length && (
<Text color={t.color.muted}> {labels.length - offset - VISIBLE} more</Text>
)}
{err ? <Text color={t.color.label}>error: {err}</Text> : null}
{busy ? <Text color={t.color.accent}>updating</Text> : null}
<OverlayHint t={t}>/ select · Enter/Space toggle · Tab user/all · 1-9,0 quick · Esc/q close</OverlayHint>
</Box>
)
}
interface PluginsHubProps {
gw: GatewayClient
onClose: () => void
t: Theme
}
+51 -73
View File
@@ -22,26 +22,30 @@ your account.
## Architecture
Inbound messages arrive as **signed webhooks**: Photon POSTs JSON with
an `X-Spectrum-Signature` header to a URL you register, and Hermes'
aiohttp listener verifies the HMAC-SHA256 signature before dispatching
the event into the agent.
Photon is a **persistent-connection** channel, like Discord or Slack —
**no webhook, no public URL, no signing secret to manage.**
Outbound replies go through a small supervised **Node sidecar** that
runs the `spectrum-ts` SDK on loopback. Photon does not currently
expose a public HTTP send-message endpoint — that's a roadmap item on
their side — so until then the sidecar is the only way to call
`Space.send(...)`. The Python plugin starts, supervises, and shuts
down the sidecar automatically. When Photon ships an HTTP send
endpoint we'll retire the sidecar in a follow-up release.
The `spectrum-ts` SDK holds a long-lived **gRPC stream** to Photon for
both directions. Because the SDK is TypeScript-only, Hermes runs it in a
small supervised **Node sidecar** and talks to it over loopback:
- **Inbound** — the sidecar consumes the SDK's `app.messages` gRPC
stream and forwards each message to the Python adapter over a loopback
`GET /inbound` (NDJSON). The adapter dedupes and dispatches it to the
agent, reconnecting automatically if the stream drops.
- **Outbound** — replies are loopback POSTs to the sidecar, which calls
`space.send(...)` on the SDK.
The Python plugin starts, supervises, and shuts down the sidecar
automatically.
## Prerequisites
- A Photon account — sign up at [app.photon.codes][app]
- **Node.js 18.17 or newer** on PATH (`node --version`)
- A phone number that can receive iMessage (used to bind your account)
- A publicly reachable URL for the webhook receiver — Cloudflare
Tunnel, ngrok, or your own gateway hostname all work
That's it — there is no public URL or tunnel to set up.
## First-time setup
@@ -58,17 +62,24 @@ hermes gateway setup
hermes photon setup --phone +15551234567
```
The setup:
The setup, in order:
1. Opens `https://app.photon.codes/` for device approval
2. Creates a Spectrum-enabled project under your account
3. Calls the Spectrum `create-user` endpoint with `type: shared` so
Photon allocates an iMessage line from the free pool
4. Runs `npm install` inside the plugin's sidecar directory
1. **Device login** (`client_id=photon-cli`) — opens
`https://app.photon.codes/` for approval and stores the bearer token.
2. **Finds or creates** the `Hermes Agent` project on your account.
3. **Enables Spectrum**, reads the project's Spectrum id, and rotates
the project secret.
4. **Registers your phone number** as a Spectrum user — skipped if a
user with that number already exists, so re-running is safe.
5. **Prints your assigned iMessage line** — the number you text to reach
your agent.
6. **Runs `npm install`** inside the plugin's sidecar directory.
Credentials are stored in `~/.hermes/auth.json` under
`credential_pool.photon` (bearer token) and
`credential_pool.photon_project` (project id + secret).
Runtime credentials are written to `~/.hermes/.env`
(`PHOTON_PROJECT_ID` = the Spectrum project id, `PHOTON_PROJECT_SECRET`),
the same place every other channel keeps its token. Management metadata
(device token, dashboard project id) lives in `~/.hermes/auth.json` under
`credential_pool.photon` / `credential_pool.photon_project`.
## Authorizing users
@@ -131,26 +142,6 @@ Both keys also accept env vars (`PHOTON_REQUIRE_MENTION`,
`PHOTON_MENTION_PATTERNS`). This is the same mention-gating model the
BlueBubbles iMessage channel uses.
## Registering the webhook
Photon needs a public URL it can POST to. Expose your local listener
(default port 8788, path `/photon/webhook`) via Cloudflare Tunnel or
ngrok, then:
```bash
hermes photon webhook register https://YOUR-PUBLIC-URL/photon/webhook
```
The response includes a `signingSecret` — **Photon only returns it
once.** Save it to `~/.hermes/.env`:
```bash
PHOTON_WEBHOOK_SECRET=v0_64-char-hex...
```
The plugin verifies every inbound `POST` against this secret and
rejects deliveries with a timestamp drift greater than 5 minutes.
## Start the gateway
```bash
@@ -160,7 +151,7 @@ hermes gateway start --platform photon
You'll see something like:
```
[photon] connected — webhook at 0.0.0.0:8788/photon/webhook, sidecar on 127.0.0.1:8789
[photon] connected — sidecar on 127.0.0.1:8789, streaming inbound over gRPC
```
Send an iMessage to your assigned number and Hermes will reply.
@@ -177,9 +168,9 @@ Prints:
Photon iMessage status
──────────────────────
device token : ✓ stored
project id : 3c90c3cc-0d44-4b50-...
project key : ✓ stored
webhook key : ✓ set
dashboard project : 3c90c3cc-0d44-4b50-...
spectrum project id : sp-...
project secret : ✓ stored
node binary : /usr/bin/node
sidecar deps : ✓ installed
```
@@ -188,27 +179,19 @@ Common issues:
- **`sidecar deps : ✗ run hermes photon install-sidecar`** — Node is
installed but `spectrum-ts` isn't. Run the suggested command.
- **`webhook key : ⚠ unset — verification disabled`** — the
plugin will accept ANY POST to the webhook URL, which is unsafe.
Re-run `hermes photon webhook register` and store the secret.
- **`PHOTON_WEBHOOK_PORT` already in use** — set a different port via
`~/.hermes/.env`.
- **Webhook reachable from localhost but Photon can't deliver**
Photon needs a public hostname. Cloudflare Tunnel is the easiest
free option.
## Webhook management
```bash
hermes photon webhook list # show registered hooks
hermes photon webhook delete <webhook-id> # remove one
```
- **`device token : ✗ missing`** — run `hermes photon setup` to log in.
- **`No iMessage line assigned yet`** — Spectrum is enabled but no line
has been provisioned; re-run `hermes photon setup` or check the
[dashboard][app].
- **Sidecar won't start** — confirm `node --version` is 18.17+ and that
`hermes photon install-sidecar` completed without errors.
## Limits today
- **Inbound attachments are metadata-only.** Inbound webhooks carry the
filename + MIME type but no download URL — Photon documents an
attachment retrieval endpoint as roadmap.
- **Inbound attachments are metadata-only.** Inbound events carry the
filename + MIME type; the agent sees a marker but can't yet read the
bytes. The SDK exposes attachment bytes via `content.read()`, so this
is a sidecar follow-up.
- **Outbound attachments are supported.** Hermes sends images, voice
notes, video, and documents through spectrum-ts' `attachment()` /
`voice()` content builders via the sidecar's `/send-attachment`
@@ -222,22 +205,17 @@ hermes photon webhook delete <webhook-id> # remove one
| Variable | Default | Notes |
|---------------------------|--------------------|--------------------------------------------|
| `PHOTON_PROJECT_ID` | from `auth.json` | Set by `hermes photon setup` |
| `PHOTON_PROJECT_SECRET` | from `auth.json` | Set by `hermes photon setup` |
| `PHOTON_WEBHOOK_SECRET` | (unset) | From `hermes photon webhook register` |
| `PHOTON_WEBHOOK_PORT` | `8788` | Local port for the aiohttp listener |
| `PHOTON_WEBHOOK_PATH` | `/photon/webhook` | Path under which the listener mounts |
| `PHOTON_WEBHOOK_BIND` | `0.0.0.0` | Bind address for the listener |
| `PHOTON_SIDECAR_PORT` | `8789` | Loopback port for sidecar control |
| `PHOTON_PROJECT_ID` | from `.env` | Spectrum project id (the SDK's `projectId`); set by setup |
| `PHOTON_PROJECT_SECRET` | from `.env` | Project secret; set by setup |
| `PHOTON_SIDECAR_PORT` | `8789` | Loopback port for the sidecar control + inbound channel |
| `PHOTON_SIDECAR_AUTOSTART`| `true` | Whether the adapter spawns the sidecar |
| `PHOTON_NODE_BIN` | `which node` | Override the Node binary path |
| `PHOTON_HOME_CHANNEL` | (unset) | Default space ID for cron / notifications |
| `PHOTON_HOME_CHANNEL` | (unset) | Default space id for cron / notifications |
| `PHOTON_HOME_CHANNEL_NAME`| (unset) | Human label for the home channel |
| `PHOTON_ALLOWED_USERS` | (unset) | Comma-separated E.164 allowlist |
| `PHOTON_ALLOW_ALL_USERS` | `false` | Dev only — accept any sender |
| `PHOTON_REQUIRE_MENTION` | `false` | Require a wake word before responding in groups |
| `PHOTON_MENTION_PATTERNS` | Hermes wake words | JSON list / comma / newline regex patterns for group mentions |
| `PHOTON_API_HOST` | `spectrum.photon.codes` | Override the Spectrum management API host |
| `PHOTON_DASHBOARD_HOST` | `app.photon.codes` | Override the dashboard / device-login host |
[photon]: https://photon.codes/
+35 -1
View File
@@ -54,8 +54,11 @@ SIMPLEX_HOME_CHANNEL=<contact-id>
| `SIMPLEX_WS_URL` | Yes | WebSocket URL of the simplex-chat daemon |
| `SIMPLEX_ALLOWED_USERS` | Recommended | Comma-separated allowlist. Each entry can be a numeric `contactId` **or** a display name — both forms work. |
| `SIMPLEX_ALLOW_ALL_USERS` | Optional | Set `true` to allow every contact (use carefully) |
| `SIMPLEX_HOME_CHANNEL` | Optional | Default contact ID for cron job delivery |
| `SIMPLEX_AUTO_ACCEPT` | Optional | Auto-accept incoming contact requests (default: `true`) |
| `SIMPLEX_GROUP_ALLOWED` | Optional | Comma-separated group IDs the bot participates in, or `*` for any group. Omit to ignore group messages entirely |
| `SIMPLEX_HOME_CHANNEL` | Optional | Default contact/group ID for cron job delivery |
| `SIMPLEX_HOME_CHANNEL_NAME` | Optional | Human label for the home channel |
| `HERMES_SIMPLEX_TEXT_BATCH_DELAY` | Optional | Quiet-period seconds (default: `0.8`) used to concatenate rapid-fire inbound text messages into one event |
## Find your contact ID or display name
@@ -68,6 +71,37 @@ By default **all contacts are denied**. You must either:
1. Set `SIMPLEX_ALLOWED_USERS` to a comma-separated list of `contactId`s and/or display names (e.g. `SIMPLEX_ALLOWED_USERS=4,alice` matches either contactId 4 or the contact whose display name is "alice"), or
2. Use **DM pairing** — send any message to the bot and it will reply with a pairing code. Enter that code via `hermes pairing approve simplex <CODE>`.
## Group chats
By default the adapter ignores group messages — a bot in a group otherwise
processes every member's traffic. Opt-in explicitly:
```
SIMPLEX_GROUP_ALLOWED=12,34 # specific group IDs
# or
SIMPLEX_GROUP_ALLOWED=* # any group the bot is in
```
Address groups by prefixing the chat ID with `group:`, e.g.
`simplex:group:12` in `send_message` or as a cron `deliver=` target.
## Attachments
The adapter supports native SimpleX attachments in both directions:
- **Inbound** — incoming images, voice notes, and files are accepted via
the daemon's XFTP flow (`rcvFileDescrReady``/freceive` → wait for
`rcvFileComplete`) and surfaced as `MessageEvent.media_urls` with the
appropriate `MessageType` (`PHOTO`, `VOICE`, `TEXT` + document).
- **Outbound**`send_image_file`, `send_voice`, `send_document`, and
`send_video` all use the structured `/_send` form with `filePath`, so
the receiving SimpleX client renders images inline and plays voice
notes inline rather than offering them as downloads.
Agent replies can also embed `MEDIA:/path/to/file` tags in plain text —
the adapter strips the tag from the body and sends the file as either a
voice note (audio extensions) or a document.
## Using SimpleX with cron jobs
```python