- Replace 3 S/M/L height buttons with one cycling button (like font size)
- Fix closeAllPanels() to include ctx-panel so Context and Settings menus
cannot be open simultaneously
- Fix ctxOpenBtn handler to use the same toggle-via-closeAllPanels pattern
as the settings button
- Align .hdr-dropdown shadow to var(--shadow) instead of hardcoded rgba
- Align #ctx-panel z-index to 200 (match .hdr-dropdown)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Replace height <select> with S/M/L buttons (data-height); active class shows
current setting; clicking an empty textarea briefly expands it as a preview
so the effect is immediately visible, then auto-shrinks back
- Add --amber/--amber-border/--amber-glow CSS vars to all 4 theme blocks:
dark=#f59e0b (bright), light=#b45309 (deep, 4:1 contrast on light bg)
Fixes local-on/tools-toggle/backend-hint being nearly invisible in light mode
- Rename "Backend" ctx-section to "Role" (matches the role-cycle toggle)
- Update backend-toggle title from stale "primary backend" to "Active role"
- Capitalize distill buttons (Short/Mid/Long/All) to match Memory layer style
- Improve all ctx-panel button titles for clarity
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- #mode-select changed from flex column to flex row (desktop + mobile unified)
- Chat/⚡ buttons now sit side-by-side at the same height as the textarea
- Removed stale mode-agent CSS rules (mode removed in prior commit)
- Mobile: simplified override — flex:1 only, direction/align already desktop default
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
OFF: very dim (nearly invisible) — makes it clear tools are inactive
ON: amber with glow — matches local-on pattern, clearly active
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Remove 'agent' from mode dropdown; Chat/Note/OTR remain
- Add ⚡ tools toggle button in input bar (persisted in localStorage)
When on: routes to POST /orchestrate (Gemini tool loop); send btn → "Run"
When off: routes to POST /chat (direct to active role); no change
- Role selector and tools toggle are now fully independent:
active chat_role sent in orchestrate payload → used for final response
- orchestrator_engine.run() accepts response_role param; passes it to
complete(role=...) instead of hardcoded model="claude"
- OrchestrateRequest gains chat_role field (default "chat")
- Migrate stored 'agent' mode/MRU entries to 'chat' on load
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add shell_exec to orchestrator tool suite (system.py + __init__.py)
Runs arbitrary shell commands on the Cortex host with timeout (1–120s),
combined stdout/stderr output, optional working_dir, and exit code reporting.
Enables system diagnostics (df, ls, ps, journalctl, etc.) from Agent mode.
- Fix orchestrator_engine.run() to use model_name from resolved registry entry
Previously used settings.orchestrator_model (.env hardcode) regardless of
what model was assigned to the orchestrator role. Now accepts model_name param
and falls back to settings value only when registry has no model_name.
- Update ARCH__FUTURE.md: date, running host, local orchestrator status,
model registry V2 progress, added Cortex Mesh concept (section 9)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The backend toggle now cycles through configured roles (chat, coder,
research, distill, etc.) instead of backup model slots within the chat
role. Each role uses its own primary→backup chain from the registry.
- ChatRequest.slot replaced by chat_role (default "chat")
- GET /backend returns available_roles instead of chat_models
- _available_roles_for_toggle() builds list from defined_roles, excluding
orchestrator (which has its own Agent mode)
- Model label on responses now reflects the actual role's assigned model
- Toggle is inert when only one role is configured (avoids useless cycling)
- Add "Clear browser cache" button to Account Settings (Connected Accounts)
- Add _role_model_label() helper for cleaner response tag labeling
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Backend toggle now cycles through chat role models by label instead of
cycling service type strings (auto/claude/gemini/local).
- model_registry: get_model_for_slot() — resolves a specific priority
slot without walking the fallback chain
- llm_client: complete() gains slot param; explicit slot selection
dispatches directly to that model with no silent fallback
- routers/chat.py: ChatRequest.slot; GET /backend returns chat_models
[{slot, label, type}] for the UI; _stream_chat uses resolved model
label for the response tag when a slot is pinned
- app.js: toggle loads chat_models from /backend, cycles by label,
sends slot in chat payload; legacy model field removed from payload
- app.js: fix Gap B — agent mode placeholder no longer says "Gemini
tool loop"; now says "orchestrator"
- DESIGN doc: updated to reflect phases 1+2 complete, catalog-as-code
decision, Gap A/B documented, Phase 3 implementation details
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add Claude Opus 4.6 and Sonnet 4.5 (previous versions, still available)
- Fix context_k for Opus 4.7 and Sonnet 4.6: both have 1M context (was 200)
- Haiku 4.5 context_k 200 is correct
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Remove outdated gemini-2.0-flash and gemini-1.5-pro.
Add gemini-2.5-flash-lite (GA) and the three Gemini 3.x preview
models (gemini-3.1-pro-preview, gemini-3-flash-preview,
gemini-3.1-flash-lite-preview). All have 1M token context windows.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Backends section: add local as third backend option, explain model
tag on responses, clarify auto vs explicit toggle behavior
- Agent Mode: remove hard-coded "Gemini" reference — orchestrator model
is now configurable via role assignments
- New Model Registry section: step-by-step for adding Google accounts,
local hosts, cloud/local model entries, and role assignments
- API reference: add local to model field, add /settings/models endpoint
- Remove outdated In Progress section (local backend + multi-user shipped)
- Header controls table: update Backend description
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds cloud provider management to /settings/models:
- Google Accounts section: add/remove Gemini API keys with labels
- Add Model form: provider tabs (Local / Google / Anthropic) with
catalog dropdowns that auto-fill label and context_k
- Provider badges on model rows (Anthropic / Google / Local)
- /settings/local now redirects to /settings/models (canonical URL)
- save_cloud_model() in model_registry for Anthropic/Google entries
- Distill role migration restored in _migrate_from_local_llm
- Test fixes: version assertions updated to V2
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a providers section to the per-user model registry for Anthropic and
Google as first-class providers alongside local hosts. Google accounts
(API keys) are now stored as a list so multiple Google accounts can coexist.
Changes:
- model_registry.py: V2 schema, auto migration V1→V2 (pulls gemini_api_key
from auth.json into providers.google.accounts), _resolve_model() merges
account API key for gemini_api type models
- routers/orchestrator.py: uses model-resolved api_key when orchestrator
role resolves to a gemini_api model with account_id
- ANTHROPIC_CATALOG and GOOGLE_CATALOG constants for model picker (Phase 2)
- New functions: get_google_api_key(), save/remove_google_account(), get_catalog()
- Documentation: ARCH__BACKENDS.md updated to V2 schema, DESIGN doc added
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The AsyncOpenAI client always appends /chat/completions to base_url.
Open WebUI's endpoint is at /api/chat/completions, so for openwebui
host_type the base_url must include the /api prefix — same logic as
_local() in llm_client.py.
Also strip non-standard metadata fields (backend, host, etc.) from
session_messages before passing them to the API.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previously any backend error would silently fall back to Claude.
Now if the user has a model configured via the model registry, errors
propagate to the UI so the actual problem is visible rather than hidden
behind a transparent backend switch.
Fallback still applies when using the default/auto backend with no
registry config.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Each assistant message in the session JSON now carries:
backend, backend_label, host (platform.node())
These fields are shown as model tags in the UI — on live responses and
when loading session history. Session log entries (sessions/YYYY-MM-DD.md)
include the backend label and host in the turn header.
The local (OpenAI-compat) backend strips non-standard fields before
sending messages to the API so extra fields don't leak upstream.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- openai_orchestrator.py — new ReAct tool loop engine for any
OpenAI-compatible endpoint (OpenRouter, Open WebUI, Ollama, LiteLLM);
model handles both tool loop and final response, no Claude handoff needed
- tools/__init__.py — auto-derive OpenAI JSON Schema from existing Gemini
FunctionDeclarations so tool definitions have a single source of truth
- routers/orchestrator.py — route to openai_orchestrator when model registry
"orchestrator" role resolves to a local_openai type host
- routers/chat.py — pass role to _backend_label(); fix fallback_used logic
(only meaningful for explicit backend overrides, not auto-routing)
- static/app.js — add null/"auto" to backend cycle; fetch local model hint
without overriding the auto default on page load
- model_registry.py — _normalize() back-fills host_type on old registry files
- requirements.txt — add openai>=1.0.0
- ARCH__BACKENDS.md — document OpenAI-compat backend and routing logic
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds host_type ("openwebui" | "openai") to the host schema so Cortex can
talk to both Open WebUI/Ollama and OpenRouter/standard-OpenAI endpoints.
Path differences per type:
openwebui (default): /api/chat/completions, /api/models
openai: /chat/completions, /models
model_registry.py:
- host_type added to host schema (default "openwebui", backward compat)
- save_host() accepts host_type parameter
- _resolve_model() passes host_type through with the merged host fields
llm_client._local():
- Reads host_type from resolved model_cfg
- Selects correct chat completions path accordingly
routers/local_llm.py:
- save_host route accepts host_type form field
- fetch-models uses /models for openai type, /api/models for openwebui
- Existing host rows show type selector pre-filled from stored value
local_llm.html:
- "Add host" form includes type selector
To use OpenRouter:
- Add host: URL = https://openrouter.ai/api/v1, Type = OpenAI-compatible
- API key from openrouter.ai (store in .env or model_registry.json only)
- Fetch models or add manually (e.g. anthropic/claude-sonnet-4-5-20251022)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Fixes:
- app.js was tracking primaryBackend locally but never included
model: primaryBackend in the /chat POST body, so the server always
used settings.primary_backend regardless of what the user clicked.
Now model: primaryBackend is sent on every chat request.
- Responses were only annotated when fallback occurred. Now every
assistant message shows a small model tag at the bottom right.
chat.py:
- _backend_label() resolves human-readable name:
claude → "Claude", gemini → "Gemini",
local → registry label (e.g. "Gemma 4 E4B") or model_name
- SSE payload now includes backend_label field
app.js:
- model: primaryBackend added to /chat fetch body
- After every response, appends .model-tag div with backend_label
- Fallback shows "⚡ fallback → <label>" in amber; normal is muted
- Removed separate system message for fallback (tag covers it)
style.css:
- .model-tag: small muted text, right-aligned, separated by thin line
- .model-tag.fallback: amber (#f59e0b)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Covers model_registry.py without requiring a running service or LLM:
Empty/fresh state: no files, missing user dir
Save/load: round-trip, corrupt file fallback
Migration: v1 hosts/models, v1 no active, v0 flat, v0 empty url,
distill_backend_mid=local → distill role, saves file after migrate
Built-in resolution: claude_cli, gemini_api, gemini_cli, unknown → None
User model resolution: local_openai merges host, missing host → None
get_model_for_role: registry primary, built-in from registry, skips missing,
walks full backup chain, .env fallback, hardcoded fallback,
custom roles
get_best_local_model: prefers role chain, falls back to first local, None if no local
Host CRUD: create, update, unknown ID creates new, remove + cascades to models
Model CRUD: create, update, remove + clears role refs
set_role: assign model, assign built-in, clear with None, invalid slot,
unknown model ID, creates new role key
get_defined_roles: returns all settings roles, fills gaps with {}
Multi-user isolation: registries don't bleed across users
All tests use tmp_path + patch.object(config.settings, ...) — no real files touched.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replaces the single-host local model settings page with a full model
registry interface at /settings/local.
Hosts section:
- List existing hosts with inline edit + save + remove
- Collapsible "Add host" form
- Per-host "Fetch models" button
Models section:
- List all models with label, model name, host, context_k badge, tags
- Remove button
Add Model section:
- Host dropdown, label, model name, context_k, tags (comma-separated)
- "Fetch models from host" with auto-fill picker
Role Assignments section:
- One row per defined role (chat, orchestrator, distill, coder, research)
- Primary + backup_1 + backup_2 dropdowns per role
- Dropdowns pre-filled from registry on load
- AJAX save on change (POST /api/models/role) with toast confirmation
- Built-in models (claude_cli, gemini_cli, gemini_api) always available in dropdowns
Backend:
- All user_settings references replaced with model_registry
- host/{id}/remove route added
- fetch-models now accepts host_id query param
- POST /api/models/role for AJAX role assignment
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Introduces model_registry.py as the single source of truth for all LLM
backend configuration. Replaces scattered backend settings across user_settings,
config distill_backend_*, and the UI toggle.
model_registry.py:
- Per-user home/{user}/model_registry.json with version, hosts, models, roles
- Models have: type (local_openai|claude_cli|gemini_cli|gemini_api), label,
model_name, host_id, context_k (tokens), tags (capability labels)
- Roles map to priority chains: primary, backup_1..backup_4
- Built-in IDs (claude_cli, gemini_cli, gemini_api) always resolvable
- Auto-migrates existing local_llm.json on first access
- CRUD: save_host, remove_host, save_model, remove_model, set_role
- get_model_for_role(): registry → .env default → hardcoded fallback
config.py:
- role_chat/orchestrator/distill/coder/research .env defaults
- defined_roles: comma-separated standard role list (extensible)
- get_defined_roles() and get_role_default() helper methods
llm_client.complete():
- New role= parameter (default "chat") for registry-based routing
- model= still accepted for explicit UI toggle override
- _claude() and _local() accept model_cfg dict instead of raw string
- _local() uses pre-resolved config from registry
memory_distiller.py:
- distill_mid/long now use role="distill" (no more distill_backend_* .env vars needed)
cron_runner.py:
- brief jobs use role="chat"
routers/chat.py + auth.py:
- Use model_registry instead of user_settings for local model info
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replaces the lone "← Back to Cortex" link with a consistent page-nav
on both pages: ← Chat | Help | Settings | Sign out
Active page is highlighted purple; others are muted gray.
Settings page gets a {{ help_href }} template var from settings.py.
Help page builds nav links from the existing cfg JS object.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Every persona now knows: direct chat has no tools, Agent mode (⚡) has
the full tool suite. If asked to write a reminder/task/etc in chat mode,
tell the user to switch modes rather than silently failing.
Updated: inara, tina, donut, wintermute, developer, cleo PROTOCOLS.md
Updated: persona_template.py so all future personas get this by default
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
NameError: name 'user' is not defined in orchestrator._run_job —
user was resolved in the endpoint but not forwarded to the background task.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- New cortex/tools/reminders.py with reminders_add, reminders_list, reminders_clear
- reminders_clear moved here from cron.py (cron still imports from same file)
- __init__.py: wired up new callables and Gemini declarations
- Inara can now add/read reminders in Agent mode via the orchestrator
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- New endpoints: POST /channels/google-chat/{username} and /webhook/nextcloud/{username}
- Channel secrets/config live in home/{username}/channels.json (gitignored)
- auth_utils: get_user_channels() helper reads channels.json
- Both routers load persona, audience/secret, backend, timeout per user;
set_context() wires the correct persona before building the system prompt
- Removed server-level channel settings from config.py and .env —
no user gets a channel until they create their own channels.json
- .gitignore: home/**/channels.json added
To migrate: update Google Chat Add-on webhook URL to /channels/google-chat/{username}
and re-register NC Talk bot at /webhook/nextcloud/{username}
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Removed orphaned root .env and .env.default (values already in cortex/.env,
which is what the systemd service actually loads)
- Replaced outdated cortex/.env.example with the comprehensive .env.default content
- Also tracks: tested/persona/cleo/ (new test persona), Inara memory updates
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- cortex/static/HELP.md: shared Help & Reference base served to all users
- help.html: loads shared base + appends persona-specific HELP.md if present
- inara/HELP.md: cleared (content moved to shared base)
- Google OAuth: registered scott.idem@oneskyit.com; flow now working end-to-end
- .gitignore: exclude home/**/sessions/ (runtime logs)
- New personas tracked: home/holly/persona/donut/, home/scott/persona/developer/
- Removed orphans: holly/, personas/, cortex-holly.service
- CLAUDE.md: updated current state and recently completed list to 2026-03-27
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Change type="password" to type="text" — the main signal password
managers use. Also add autocomplete="off", data-lpignore, data-1p-ignore
for broader coverage across Bitwarden, 1Password, LastPass.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The invite command reads email from profile.json, not auth.json.
google-add was only writing to auth.json so invite had no address
to send to. Now calls set_email() as well.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Settings page gains two new sections:
- Connected Accounts: shows linked Google email (read-only)
- Gemini API Key: paste personal key from aistudio.google.com,
shows masked hint of saved key, remove link to revert to server key
POST /settings/gemini-key saves/clears gemini_api_key in auth.json.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Users with Google accounts can now sign in without a password.
Auth flow:
- GET /auth/google → Google consent page (CSRF state cookie)
- GET /auth/google/callback → exchange code, lookup user, set JWT
- auth.json gains google_sub + google_email fields
- set_password() no longer overwrites unrelated auth.json fields
Admin setup:
python manage_passwords.py google-add <username> <email>
# add GOOGLE_CLIENT_ID + GOOGLE_CLIENT_SECRET to .env
Per-user Gemini key:
- get_user_gemini_key() reads gemini_api_key from auth.json
- orchestrator_engine.run() accepts gemini_api_key param
- orchestrator router passes user's key, falls back to server key
login.html: "Sign in with Google" button above the password form.
manage_passwords.py list: now shows auth method columns (pw / google).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
app.js updates the <link rel="icon"> to the active persona's emoji on
load (CORTEX_EMOJI is already injected server-side). /favicon.ico route
added as a fallback for login/settings/help pages that don't have
persona context.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
beforeunload closes the EventSource explicitly so the browser doesn't
log "connection interrupted while page was loading". onerror handler
suppresses auto-reconnect noise if the connection temporarily drops.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Session ID is stored in localStorage keyed to user+persona. On page load
it's silently restored if within 30 min of last activity. Timestamp
updates on every sent message. New session / delete session clears the
stored ID so the TTL logic stays consistent.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
pencil → edit, trash-2 → del, copy → copy, check → copied feedback,
check → Save, x → Cancel. All small action buttons get inline-flex
alignment for consistent icon+label layout.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Visiting /scott (or any user root) now shows a clean card page listing
all their personas with emoji + name, each linking to /{user}/{persona}.
Previously the route was unhandled (404 or wildcard match).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pass ?persona= query param on the help link so the server knows which
persona to return to. Previously always defaulted to personas[0], causing
navigation back to the wrong persona.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Sync preload script font sizes to match app.js (21/25/17px)
- Send button variants now show icons: ↑ Send, 📝 Note, ⚡ Run
- Remove fixed width on send-col; add white-space:nowrap + padding
so "📝 Note" never wraps regardless of font size
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Load Inter variable font from Google Fonts on all 5 HTML pages
- font-weight: 450 on body (between regular and medium — fixes thin feel)
- -webkit-font-smoothing: antialiased for cleaner screen rendering
- Base font size: normal 16→17px, large 18→19px, small 14→15px
- Applies consistently to main UI, login, setup, settings, and help pages
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Header:
- Sessions, ⚙ context panel, ≡ settings dropdown (Files, Account,
Sign Out), and ❓ help — down from 6+ individual buttons
- Responsive: flex-row on desktop, wraps on mobile with labels hidden
Footer (input area):
- 4-way mode select replaces the button row — shows only the active
mode as [icon] [label] ▲; click opens an upward dropdown
- Options sorted by MRU: most recently used floats to the bottom
(closest to the trigger button) for quick re-selection
- Current mode marked with ✓
- Note mode shows a small prv/pub sub-toggle below the select button
- Mobile: textarea on top (full width), mode select + send on one row
Mode state consolidated from 3 booleans into a single current_mode
variable with localStorage persistence and MRU tracking.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a third input mode toggle alongside Note and Agent. When active:
- Textarea gets a subtle purple tint with dashed border
- OTR button highlights purple
- Placeholder reads "Off the record — not logged or distilled…"
- off_record=True is sent to /chat; session_logger is skipped
- In-memory session context is preserved within the session
Switching to Note or Agent mode deactivates OTR, and vice versa.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Hardcoded 'Inara' replaced with CORTEX_PERSONA in all placeholder
strings (chat mode and agent task mode).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Fix 'undefined' in auth banner: read access_token_hours_remaining (not hours_remaining)
- Fix false-positive warning on fresh tokens: when refresh token present, only warn
within 1 hour of expiry (not 24h) since the CLI should auto-rotate but sometimes misses
- Emit claude_auth_expired SSE event on 401 so UI shows inline red banner immediately
- app.js: handle claude_auth_expired SSE event with persistent top banner + dismiss button
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
