Spaces:
Runtime error
Runtime error
Commit History
-
fix: Playwright chromium deps for Debian trixie (ttf-unifont removed)
3ef4aa4 -
Create sentinel_audit.yml
Unverified signaturee2225ae -
Create ci.yml
Unverified signaturebc50d23
-
chore(cf): add KV_ROUTING namespace binding to wrangler.toml
8043a14 -
fix(search): accept TAVILY_KEY_0..N env var pattern (project convention)
61e0019 -
v26: Chat page (cross-project persistent), Memory live, brain/__init__.py, Sidebar updated
9a87c53 -
v26: add GET /infra/events endpoint (SpacePromoter Redis event log)
f6ec2d4 -
v26: complete website — Projects live, Sentinel live topology, 4 new components, CF Worker updated, api.js complete
37f231b
-
v26: live Projects + Sentinel pages, updated api.js
44737f9 -
v26: voice __init__.py + requirements-voice.txt
cde8149 -
v26: wire SpacePromoter + Discord bot into lifespan
3e62ed4
-
v25: patch main.py+discord_bot.py (lifecycle/gt/rdloop wired), add voice/server.py, add infrastructure/space_promoter.py
190b4da -
feat: memory upgrades (MAGMA/lifecycle/ground_truth), rd_loop, voice Dockerfile
Sources read: FredJiang0324/MAGMA (graph_db.py), BAI-LAB/MemoryOS (mid_term.py), microsoft/RD-Agent (core/) magma_graph.py: 4-graph MAGMA layer on NetworkX, Redis-persisted, async lifecycle.py: EverMemOS MemCell/MemScene/Foresight + heat-based promotion ground_truth.py: Raw episode store, drift-free, extraction on READ not write rd_loop.py: Autonomous R&D loop - Propose->Rank->Debate->Notify->Repeat docker/ultron-voice.Dockerfile: T1 fix (Kokoro weights pre-dl), VP2 fix (LiveKit WebRTC)1342767 -
feat: master website scaffold — Mission Control UI (Priority 2)
React + Vite + Tailwind. 6 pages: Dashboard, Credentials, Memory, Projects, Sentinel, Settings. CF Worker API proxy. Editorial aesthetic. Ghost's primary interface. Pages: - Dashboard: real /health data, provider pool bars, system health - Credentials: .env drag-drop import, key table per provider, show/mask, pool type - Memory: STM/MTM/LPM/MAGMA/Foresight tabs (MAGMA+Foresight stubbed, wire after magma_graph.py) - Projects: timeline stub (wire after Supabase schema) - Sentinel: trigger audit, incident log, weekly reports (wire after endpoints) - Settings: infra URLs, quick reference CF Worker (ultron-api): proxies /api/health, /api/infer, /api/sentinel/event Key mgmt + STM endpoints stubbed until Brain endpoints added. Deploy: CF Pages (website) + wrangler deploy (cloudflare/worker.js)1720172
-
feat: main.py v22 — wire Sentinel + Council into lifespan, /sentinel/event fully delegates to Sentinel.handle_event()
31d0de3 -
chore: requirements.txt — add PyNaCl (discord voice), kokoro, soundfile (Phase 5 voice)
08a1aab -
feat: Phase 8 brain/council.py — MOA Council Mode, parallel expert calls, synthesis, Redis state
28b5f98 -
feat: Phase 7 tools/code_exec.py — async sandboxed Python + shell execution
3fc262c -
feat: Phase 7 tools/browser_agent.py — Playwright LLM browser automation
5dbdc6c -
feat: Phase 5 voice/pipeline.py — VAD + STT + Brain + TTS pipeline
dfa7efd -
feat: Phase 5 voice/tts.py — Kokoro TTS WAV synthesis
9ddf241 -
feat: Phase 5 voice/stt.py — Groq Whisper STT
b3872bc -
feat: Phase 5 voice/__init__.py
4a31a0f -
feat: Phase 4 sentinel.py — V4 Sentinel God Layer with KV routing, incident reporting, weekly audit, Notion write
1415cd6 -
feat(brain/main): wire MemoryWorker + ZillizStore + RaptorTree into lifespan
b0e15a8 -
feat(deps): enable Phase 3 deps — pymilvus, umap-learn, scikit-learn
bb55cc9 -
feat(memory): Phase 3 — embedder, zilliz, raptor, worker, __init__
58ab44f
-
fix: wire real tavily_search into task_dispatcher, replace stub
47f2fa7 -
feat: search.py (Tavily+DDG fallback), requirements.txt, Dockerfile, README — deploy-ready
d59842f -
feat: discord_bot.py — per-channel context, slash+prefix cmds, chunked send, no internal leaks
a4a5261 -
feat: add config.py, config_loader.py, main.py — system boots end-to-end
47241e6 -
feat: add KeyPool dual sub-pool (general+sentinel) + shared exceptions
Task 2: packages/brain/key_rotation/pool.py — full V4 KeyPool - Dual sub-pool: self.general[] + self.sentinel[] - get_key(pool_type) → weighted RR with circuit breaker - get_sentinel_key() → dedicated Gemini 2.5 Pro key - report_success/report_failure with per-provider thresholds - asyncio.Lock thread-safety - AllKeysExhaustedError + SentinelKeyUnavailableError in shared/exceptions.py - _is_available() auto-resets expired cooldowns Weights: Groq=3, Cerebras=3, Together=2, OpenRouter=2, Gemini=2"7419178 -
fix: remove _make_groq_llm_fn, import make_provider_llm_fn from llm_router
Task 1 patch: pool now routes to all 5 providers (Groq/Cerebras/Together/OpenRouter/Gemini) not just Groq. Removes Groq-only LLM fn, wires make_provider_llm_fn(pool) from llm_router. Closes architecture gap identified in v16 session.63db9db
-
feat: add llm_router.py — multi-provider router (Groq/Cerebras/Together/OpenRouter/Gemini) session v16
65d8e8c -
feat: add packages/brain/task_dispatcher.py — orchestration layer v4 (session v15)
be4911f -
feat(brain): add react_loop.py — ReAct agentic loop engine (V4 Phase 2)
Informed by: - browser-use/browser-use: ActionLoopDetector, MessageCompaction, flash_mode schema - OpenHands/codeact_agent: pending_actions deque, condenser pattern, tool dispatch Features: - ActionLoopDetector port (rolling hash window, soft nudge) - MessageCompaction port (compact every 25 steps, keep_last 6) - flash_mode schema for Groq (memory+action only) - Tool registry hook (pluggable async tools) - Max 5 iterations default (hard ceiling 10) - Full AgentState tracking (n_steps, consecutive_failures, paused/stopped) - Groq function_calling compatible tool schemadac0057 -
Initial commit
Unverified signaturefb54f06