The gateway SSE handler (`_handle_gateway_sse_stream`) and session-events
SSE handler (`_handle_session_events_stream`) both set
`Connection: close` on the response. On long-lived SSE streams, browsers
(Chrome, Firefox, Safari) interpret that header as 'this is a one-shot
response, the EventSource lifecycle is over the moment the body ends'
and trigger an immediate reconnect when the server-side worker rotates.
The result, with EventSource's auto-reconnect on top, is a tight loop of
connect -> sessions_changed snapshot -> reconnect every ~1s that pegs
the worker, thrashes the session list, and prevents the in-app session
list from ever settling.
Removing the header lets the python BaseHTTPServer close the socket
naturally after the stream ends, which is what EventSource expects.
Regression introduced in 598fd4ff.
Per Opus advisor on stage-batch36: skip role='user' messages in
_session_media_token_allows_image_path so a user-injected MEDIA: token
cannot mint an allow-list entry for the user's own request. Preserves
the original use case (assistant/tool emitted artifacts outside the
active workspace) while making the implicit threat model explicit.
Defense-in-depth — the single-user WebUI scope means same-origin user
input already had the same effective access, but multi-user / shared
WebUI deployments would benefit from the restriction.
PR #3023 only updated Session.load() and Session.load_metadata_only(), leaving
three sibling validators (Session-internal _repair_stale_pending and the
/api/session/worktree/remove + /api/session/delete route handlers) still
gated on the old lowercase-only character set. That would have shipped a
confusing UX where api-* and reachy-voice-* sessions could be loaded into
the sidebar but rejected with HTTP 400 on delete or worktree removal.
This commit factors the validation into a single is_safe_session_id helper
in api.models and updates all five call sites to use it. Adds regression
coverage in tests/test_issue3023_safe_session_id_validators.py for both
the helper itself and a repo-wide guarantee that no narrow lowercase-only
magic string survives.
Closes the follow-up flagged by the parallel reviewer agent on #3023.
The .usage.json file is owned by hermes-agent (tools/skill_usage.py).
This change removes the webui-side increment logic to avoid:
1. File ownership conflict - both writing to same file
2. Schema mismatch - agent uses ISO strings, webui used floats
3. Concurrency issues - agent uses fcntl locks, webui had no locking
4. Double-counting - agent already increments counters server-side
Changes:
- api/skill_usage.py: keep only read_skill_usage(), remove increment functions
- api/streaming.py: remove skill usage counter hook
- api/routes.py: adapt response to pass through agent's format as-is,
with defensive coercion for None values and metadata preservation
- tests/test_skill_usage.py: remove increment tests (17→7 cases)
Use _merged_session_messages_for_display for is_messaging_session even in the
metadata-only (messages=0) path. This ensures message_count and last_message_at
match the full load path for Telegram / external messaging sessions that have
stitched or duplicate rows in state.db + sidecar.
Prevents spurious refresh loops, scroll resets, and open panel closures when
resuming cross-surface sessions in the WebUI.
No impact on CLI, non-messaging, or full-message paths. All 580 session tests pass.
Fixes the root cause identified in the SessionDB / render interaction changes.
Replace the hardcoded 4-option deliver dropdown (local/discord/telegram/slack)
with a dynamic select populated from a new GET /api/crons/delivery-options
endpoint that reads _KNOWN_DELIVERY_PLATFORMS from hermes-agent.
Key changes:
- Add GET /api/crons/delivery-options endpoint returning all known platforms
- Frontend loads options asynchronously on first cron form open, with caching
- Enable deliver editing for existing jobs (was previously disabled)
- Include deliver in update payload when editing cron jobs
- Fallback to local-only if API unavailable
- Custom deliver values (e.g. feishu:oc_xxx) shown with * suffix
- Add cron_deliver_custom i18n key to all 12 locales
- Add 5 integration tests for the new endpoint
Resolve supported reasoning efforts per active model/provider and pass
that context through /api/reasoning so Composer and other non-configurable
models no longer show a misleading effort picker.
Co-authored-by: Cursor <cursoragent@cursor.com>
Ensure cursor/composer IDs always resolve via @cursor-acp:, carry the
visible picker selection into POST /api/session/new, persist model
changes before a session exists, and evict cached agents on model switch.
Co-authored-by: Cursor <cursoragent@cursor.com>
Inline fixes for 4 of 5 Opus SHOULD-FIX items before tag:
1. /api/auth/status now gates passkeys_enabled / passwordless_enabled on
_passkey_feature_flag_enabled() — when flag is off, status reports
no credentials even if passkeys.json has legacy entries. New
passkey_feature_flag field added to the response for the frontend.
2. Settings → System Passkeys block (passkeysSettingsBlock) now starts
display:none and loadPasskeys() reveals it only when the server
confirms passkey_feature_flag === true AND /api/auth/passkeys
doesn't return {disabled: true}. Stops the broken-affordance trap
where users would see Add passkey → click → 404.
3. /api/settings/save now refuses to set passwordless mode when the
passkey feature flag is off. Closes the auth-bypass path Opus flagged:
user goes passwordless while flag on → admin unsets flag → restart
serves the WebUI fully unauthenticated.
4. CHANGELOG entries added for PR #2685 (replayed-context dedup +
per-turn metering cap) and PR #2824 (Stop server affordance,
relocated to Settings) — both PRs had functional changes but no
release-notes entries. Also enriched the rate-limit detail on the
#2739 entry (30 events / 60s / 4KB body cap).
Deferred to follow-up issue (#5 in Opus review):
- Live tool metering cumulative cap across many tool calls — non-trivial
refactor of _bump_live_prompt_estimate, will be a separate PR
Per the stage-batch14 ship plan, passkey/WebAuthn support is shipped
opt-in default-off behind an explicit feature flag so deployments can
disable the entire surface (UI + endpoints + credential storage) without
needing to delete code.
Enable via either:
- HERMES_WEBUI_PASSKEY=1 environment variable, OR
- webui_passkey_enabled: true in config.yaml
With the flag off:
- are_passkeys_enabled() returns False even if credentials exist
- is_auth_enabled() falls back to password-only checking
- /login renders password-only (no passkey button)
- All 6 /api/auth/passkey/* endpoints return 404 with a clear message
- Settings → System → Passkeys section is hidden
Mirrors the #2527 notes-drawer flag shape (env-or-config, truthy parse).
Auth is high-stakes; opt-in lets us land the code while keeping default
deployments on the well-tested password-only path.
Touches: api/auth.py (new _passkey_feature_flag_enabled helper, gated
are_passkeys_enabled), api/routes.py (6 endpoint guards).
Move POST /api/shutdown routing after the CSRF check so drive-by
cross-origin requests cannot bring down a dev server with auth off.
Also replace os._exit(0) with os.kill(os.getpid(), signal.SIGINT)
so atexit handlers and pending session writes run during shutdown.
Add a power button (⏻) in the title bar that gracefully stops the
WebUI server process from the browser.
- api/routes.py: POST /api/shutdown endpoint with threaded os._exit(0)
- static/boot.js: shutdownServer() with confirm prompt, BroadcastChannel
cross-tab notification, and _showServerStopped() placeholder UI
- static/index.html: shutdown button HTML in title bar (after reload btn)
- static/style.css: .app-titlebar-shutdown styles, hover turns red
Opus pre-release advisor caught a #2762 parity gap. api/streaming.py:5078
(_run_agent_streaming worker, background thread) correctly passes
profile= to sync_session_usage post-#2827. But the SECOND production
call site at api/routes.py:9007 (_handle_chat_sync, HTTP thread) does
not. Safe TODAY (HTTP thread sets TLS correctly), but it's a
defense-in-depth gap: anyone wrapping that handler in a worker pool
later silently regresses the fix. Closes the parity gap so the
threat-model invariant holds regardless of future threading changes.
Opus pre-release advisor MUST-FIX patched inline:
- api/routes.py:7290-7308 _handle_folder_download: add Connection: close
header before end_headers() to satisfy HTTP/1.1 framing on the on-the-fly
ZIP stream. Without it, post-#2836 protocol_version bump leaves clients
hanging waiting for the next pipelined response after central-directory
bytes finish. Opus verified this is the ONLY streaming response #2836
missed — all other paths (j/t helpers, 12 hand-written responses, 8 SSE
endpoints, auth flow) are already correctly framed.
Cherry-picked via 3-way apply (rebase had failed on static/index.html
conflict when applied via rebase commit chain; 3-way of the net delta
against stage HEAD applied cleanly).
Co-authored-by: mccxj <mccxj@github.users.noreply.github.com>
Agent reviewer 'LGTM. Ship it.'
- Bug A fix: _session_field helper handles dict-vs-object snapshot in pin-limit check
- Bug B fix: removed stale client-side pinLimitReached short-circuit
- Bug C recovery: renderSessionList() on pin/unpin failure refreshes from server
Co-authored-by: franksong2702 <146128127+franksong2702@users.noreply.github.com>
nesquena APPROVED 2026-05-22. Cherry-picked onto post-v0.51.127
master via 3-way apply. Resolved api/routes.py conflict: master had
the inline correctness fix from the deep-review iteration; PR
refactors it into _metadata_only_message_summary() helper. Took the
helper AND added profile= threading (post-#2827 master adds
profile-aware state.db reads). Kept master's pre-existing
test_api_session_reload_drops_stale_cached_user_tail_after_saved_assistant
alongside the PR's new test_metadata_fast_path_matches_reconciliation_for_restamped_replays.
Co-authored-by: dobby-d-elf <dobby.the.agent@gmail.com>
Add a complete Turkish locale to the WebUI and login page so users can
select Türkçe in Settings, with speech recognition via tr-TR.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add `PATCH /api/mcp/servers/{name}` endpoint that accepts `{"enabled": bool}`,
updates `mcp_servers.<name>.enabled` in config.yaml, and calls `reload_config()`.
Mirrors the existing DELETE pattern.
Also wire the previously-defined-but-unrouted `_handle_mcp_server_delete` into
`handle_delete`, and `_handle_mcp_server_update` into a new `handle_put` +
`do_PUT` in server.py — fixing a pre-existing bug where those handlers existed
but were never reachable over HTTP.
UI: add a toggle button in each MCP server row in the system settings panel
(panels.js). Clicking it calls PATCH and reloads the list. Toggle button is
styled with `.mcp-toggle-enabled` / `.mcp-toggle-disabled` CSS classes. The
`toggle_supported` flag in the list response is now `True`.
i18n: add 5 new keys (`mcp_enable_server`, `mcp_disable_server`,
`mcp_enabled_toast`, `mcp_disabled_toast`, `mcp_toggle_failed`) to all 9
non-English locales (English values as placeholder translations).
Tests: add `TestMcpToggle` class with 7 tests covering disable, enable,
404-not-found, empty name, missing field, response payload, and URL-encoded name.
Update `test_empty_config` and visibility panel assertions to reflect
`toggle_supported: True` and the new toggle button in panels.js.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Right-click any workspace file, folder, or root now shows
'Open in VS Code' alongside the existing Reveal in File Manager action.
- POST /api/file/open-vscode: resolves path via safe_resolve, finds VS
Code via shutil.which() with fallbacks for macOS (/usr/local/bin/code,
app bundle CLI), Linux (/usr/bin/code, /snap/bin/code), and Windows
(%LOCALAPPDATA% and %PROGRAMFILES% user/system installs). Returns a
descriptive error if not found rather than a bare OS error.
- Optional vscode block in config.yaml: command (default: code),
host_path_prefix + container_path_prefix for Docker path mapping.
- i18n: open_in_vscode and open_in_vscode_failed translated in all 10
locales (it, ja, ru, es, de, zh-CN, zh-TW, pt, ko).
- 26 tests in tests/test_2735_open_in_vscode.py covering source wiring,
command resolution, i18n completeness, and live endpoint error paths.
Two bugs combined to cause historical messages to vanish from the WebUI
after a session was continued in a later conversation.
**Bug 1 — missing `id` in state.db SELECT (models.py)**
`get_state_db_session_messages()` did not include the `id` column in its
SELECT, so every row got a `("legacy", ...)` merge key instead of
`("message_id", ...)`. The timestamp gate in
`merge_session_messages_append_only()` explicitly exempts `message_id`-keyed
rows from its "skip if older than newest sidecar message" rule, but
legacy-keyed rows are unconditionally dropped. With a session that has any
new sidecar messages (max_sidecar_timestamp == today), all older state.db
rows were silently discarded.
Fix: include `id` when the column is present so rows get proper
`("message_id", ...)` keys and survive the timestamp filter.
**Bug 2 — always reads active profile's state.db, not the session's (models.py + routes.py)**
`get_state_db_session_messages()` always called `_active_state_db_path()`,
which returns the currently-active profile's database. Sessions belonging to
a different profile (e.g. `jump`) were read from the wrong state.db, returning
either no rows or unrelated ones.
Fix: add an optional `profile` parameter; when supplied, resolve the path via
`_get_profile_home(profile)` with a fallback to the active path if the
profile-specific db does not exist. The call-site in `routes.py` now reads
`session.profile` and passes it through.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Original PR: #2676 by @lucasrc
Adds POST /api/skills/toggle endpoint that flips skills.disabled in
config.yaml, and a UI toggle in the Skills panel that shows all skills
(including disabled ones) with a per-skill on/off control.
- Backend: new endpoint validates skill exists in filesystem before
toggling. Read-modify-write wrapped in _cfg_lock for thread safety.
Writes through to platform_disabled.webui when present.
- Frontend: each skill-item now has a toggle switch; disabled skills
appear muted but still listed (previously they were filtered out).
- i18n: new toggle keys translated across all 9 non-English locales.
- Tests: round-trip test for disabled list normalization + toggle
endpoint behavior.
Squash-merged from contributor's branch (19 commits + 1 merge commit)
onto current master via the cherry-pick-stale-contributor-prs procedure.
Two functions on the /api/session/handoff-summary hot path were opening
sqlite3.connect(...) inside a bare `with` statement, which commits the
transaction at scope exit but does NOT close the connection. Per-turn
invocations accumulated state.db / state.db-wal file descriptors and
CPython heap pages on long-lived worker threads, surfacing as the
multi-GB VmRSS and 6x duplicated state.db fds observed on the live
instance (D0 pre-restart baseline: VmRSS 1,334,248 kB, 55 fds; cold
baseline after restart: VmRSS 136,668 kB, 10 fds).
Wrap both call sites with contextlib.closing(...) (already imported and
used at seven other sites in the same files) so the connection is
closed deterministically:
- api/models.py :: count_conversation_rounds
- api/routes.py :: _persist_handoff_summary_to_state_db
Regression test:
tests/test_issue2233_sqlite_connection_leak.py loops both functions
20 times against a tmp state.db and asserts /proc/<pid>/fd count
does not grow more than 2. Linux-only via sys.platform skip.
D1 live soak against a freshly-built worktree server (port 8799,
isolated HERMES_HOME / HERMES_WEBUI_STATE_DIR) hitting
/api/session/handoff-summary 20 times:
fd_before = 5
fd_after = 5 (growth 0, threshold < 5)
vmrss_before = 52636 kB
vmrss_after = 52636 kB (growth 0 kB, threshold < 30 MB)
The patched fix curve trends below the leak curve.
Rollback: single git revert <this-sha> reverts both file edits.
Refs #2233.
The WebUI clarification popup had a response-delivery failure: users
submitted answers in the popup, but the agent still fell through to the
timeout fallback message. Three bugs conspired:
1. No stable clarify_id — _ClarifyEntry had no unique identifier, so
the frontend could not reference a specific pending prompt. The
backend used FIFO resolution which silently failed for stale/late
responses.
2. Frontend hid the card before confirmation — respondClarify() called
hideClarifyCard(true, 'sent') BEFORE the API call completed. If the
backend rejected the response, the card was already gone and the
user's draft was discarded.
3. Backend lied about success — _resolve_clarify_legacy() returned
bool(resolved) or not bool(clarify_id). Since the frontend never
sent clarify_id, the backend always reported ok:true even when
nothing was resolved.
Changes:
api/clarify.py:
- _ClarifyEntry now auto-generates a stable clarify_id (uuid4.hex[:12])
- submit_pending() injects clarify_id into the data dict visible to the
frontend via SSE and polling
- New resolve_clarify_by_id() for O(1) lookup by id instead of FIFO pop
api/routes.py:
- _resolve_clarify_legacy() uses resolve_clarify_by_id when clarify_id
is provided; returns actual bool result (no more unconditional True)
- _handle_clarify_respond() returns HTTP 409 + {ok:false, stale:true}
when resolution fails
static/messages.js:
- respondClarify() now sends clarify_id in the POST body
- Waits for a positive backend acknowledgement before hiding the card
- Saves a draft copy before POST and restores it on failure
- On 409/network error: re-enables controls, shows error toast
- Guards against parallel-SSE race where clearing the cache after a
successful response could erase a newly queued next prompt (codex P1)
tests:
- Updated test_sprint30.py for new ack-before-hide behaviour
- Updated test_clarify_unblock.py for 409 on stale responses
Closes#2639.
The Settings → Plugins panel keyed off `loaded.enabled` and the four
agentic visibility hooks, both of which are False/empty for exclusive
plugins (memory.provider, model-provider, etc.). Those plugins were
mislabeled as "Disabled" with "No registered lifecycle hooks" even when
fully functional as the active provider for their category.
Surface `manifest.kind` and a derived `activation` field
("enabled" | "disabled" | "exclusive" | "provider") in /api/plugins.
The card render picks a third badge state ("Active (provider)") and a
dedicated empty-hooks line for those rows. `enabled` is preserved in
the payload so older clients still work; new clients should prefer
`activation`.
Fixes#2659
Drop the redundant 'if gw_data else []' guard — gw_data is already
guaranteed to be a dict by the 'or {}' fallback above.
Add a one-line comment explaining the peek-without-pop race window:
a concurrent resolver may pop a different gateway entry, but
approve_session is idempotent over the session key set so the
outcome is the same regardless.
During active streaming, dangerous-command approvals go through the
gateway path and are stored in _gateway_queues as _ApprovalEntry
objects, not in _pending. The _resolve_approval_legacy helper only
looked at _pending, so 'Allow for this session' never called
approve_session() — the user clicked Allow, the card vanished, but
the next dangerous command asked again.
Now when _pending has no matching entry, the helper peeks into
_gateway_queues to extract pattern_keys, calls approve_session(),
and marks found_target=True so resolve_gateway_approval also fires.
This commit is re-scoped to peek-only (no agent_session_key round-trip,
no state_db metadata changes).
Includes:
- Import + fallback for _gateway_queues
- Null-safe key filtering in all_keys
- Source-contract test (static) + functional test with
@requires_agent_modules skip marker for CI
- All comments and docstrings in English
Per reviewer note: because the zip streams straight into handler.wfile
(no io.BytesIO buffering), peak memory is bounded by zipfile's per-file
read buffer, not the HERMES_WEBUI_FOLDER_ZIP_MAX_MB cap. Adds a comment
so the next reader doesn't have to trace it to learn the cap's actual
shape.
When API server runs append messages directly to state.db, reconcile WebUI sidecar sessions with those canonical rows across API responses, model-facing streaming context, and active browser refresh.
Add append-only state.db merge helpers, metadata-only counts for refresh polling, and regression coverage for API visibility, context incorporation, and frontend refresh behavior.
Adds a "Download Folder" item to the workspace file-tree right-click
menu and a GET /api/folder/download endpoint that streams the
directory as a zip with Content-Disposition: attachment.
Configurable caps:
HERMES_WEBUI_FOLDER_ZIP_MAX_MB (default 1024)
HERMES_WEBUI_FOLDER_ZIP_MAX_FILES (default 50000)
Pre-flights the walk so cap-exceeded returns 413 + JSON BEFORE any
zip bytes are sent. Symlinks resolving outside the workspace are
skipped. Mirrors the existing _handle_file_raw shape (session_id
resolution, safe_resolve, RFC 5987 filename via
_content_disposition_value). Stdlib zipfile only; no new dependencies.
Tests: 11 static-inspection tests matching the style of
tests/test_issue1867_upload_size_preflight.py. All passing on
Python 3.11/3.12/3.13.
