Spaces:
Sleeping
Sleeping
Commit History
-
Add Helm charts, Dockerfiles, and K8s docs
Prepare project for local Kubernetes deployment: add multi-stage Dockerfiles for backend, mcp, and frontend (builder + runtime), include healthchecks and a /health endpoint for the MCP service, and enable Next.js standalone output. Introduce an umbrella Helm chart (helm/doit) with frontend/backend/mcp subcharts, templates for Deployments, Services, ConfigMaps, Secrets, Ingress, values.yaml and a values.secret.yaml.example, plus .helmignore. Add .dockerignore files for frontend/backend/mcp and update root .gitignore to ignore the secret values file. Update docs and metadata: add Phase 4 deployment guide, update README, AGENTS.md, CLAUDE.md, specs/tasks and history prompt. These changes stage the repo for Minikube/Helm deployment (image build & cluster validation still required).7c7b1d1
-
Improve Gemini agent resilience and chat UI routing
Backend (mcp_service): - Add logging and ModelSettings for Gemini compatibility, omitting unsupported fields. - Harden run_agent_stream: sanitize/truncate conversation history, handle Responses vs ChatCompletions event shapes, and retry once with only the last user message on API errors (e.g. invalid argument or 400) to recover from history-related failures. - Ensure success returns early and errors yield clear messages. - Normalize stored timestamps to naive UTC to match DB column type (updated_at and _utc_now). Frontend (chat UI): - Replace Next.js router navigation for conversation selection/creation/deletion with local state + window.history.replaceState to avoid full Next navigation and suspense, and sync active conversation id from search params. - Memoize handlers with useCallback and wire loadConversations refresh after responses complete. - ChatInput: auto-focus textarea when re-enabled. - ChatPanel: avoid reloading messages if already present, add onResponseDone callback and…8239783
-
Add media delete, audio UI, and MCP fixes
Backend: add DELETE /media/{asset_id} endpoint and NoteService.delete_media_asset with ownership check via associated note to allow users to remove uploaded assets. Frontend: add CustomAudioPlayer and Markdown Pre component (with copy button); integrate audio playback and pre component into NoteCard and NoteViewDialog. Enhance NoteForm with voice recording (MediaRecorder), audio device selection/settings, UI for deleting existing media (mark-for-delete + send DELETE on save), and various UI/UX updates. MCP service: improve streaming handling to support Responses API text deltas and fallback to chat-completions format, remove an unused variable, add UserIdMiddleware to inject X-User-ID into MCP context and register it with FastMCP, and make conversation.updated_at timezone-aware. These changes enable media lifecycle control, in-app voice notes/playback, and more robust/compatible agent streaming and context propagation.b5736da
-
Add docker services, UI polish, and MCP updates
Restructure docker-compose to add backend, mcp and frontend services with named volumes and env files (exposes backend:8000, mcp:8080, frontend:3000). Improve frontend: use next/navigation query param routing for chat, add Suspense wrapper, refine ChatInput/ChatMessageList/ChatPanel/ChatWidget/Sidebar UI and behavior (conversation selection, optimistic updates, delete actions), and make API proxy env var fallbacks more flexible. Update several API routes to read MCP_URL from multiple env vars. Update MCP service: switch agent model to gemini-3-pro-preview, wire MCP ASGI app lifespan into FastAPI and mount FastMCP at root (preserve API routes precedence). Update docs/tasks: quickstart.md MCP_PORT corrected to 8080 and tasks.md marks T046 complete; add quickstart validation history prompt file.9312beb -
Standardize mcp_service and add uv scripts
Rename and standardize the MCP package and runtime: update pyproject to use mcp_service and add project.scripts (prod/dev); add mcp/src/mcp_service/__init__.py with development/production uvicorn entrypoints. Revise Dockerfile to use the uv runtime image, copy sources, install deps via uv sync, set PYTHONUNBUFFERED, and run the app with `uv run prod` (expose port 8080). Update READMEs and quickstart to reflect new service name, port and startup commands.5250b68
-
Add MCP chat UI, API proxies, and tools
Implements Phase 3 chat MVP: adds a /chat page, floating ChatWidget, and chat components (ChatPanel, ChatInput, ChatMessageList) with SSE streaming, optimistic updates, and markdown rendering. Adds Next.js API proxy routes to the MCP service (chat SSE, conversations list/delete, messages) and new Conversation/ChatMessage/SSEEvent types. On the MCP side, implements and registers tools for tasks, notes, projects, categories, and a dashboard summary; includes categories, dashboard, and tasks tool implementations. Fixes async DB URL/SSL handling in config/database for asyncpg, adds mcp .gitattributes and README, and a GitHub Actions workflow to deploy the mcp/ subtree to a Hugging Face Space. Also updates history/prompts and task specs to reflect completed work.ada1dfd -
Add MCP service, conversation models & migration
Introduce a new mcp/ service (FastAPI + FastMCP) with agent integration and conversation support. Adds SQLModel conversation and message models (both backend and mcp), Alembic migration to create conversation/message tables, DB/auth/config/database helpers, and streaming chat endpoints (SSE) that persist messages and conversation metadata. Also adds project scaffolding (pyproject, Dockerfile, .env.example, .dockerignore, uv setup), agent/provider code for Gemini via OpenAI-compatible client, and supporting model files (project, note, category, session, user imports). Add GitHub Actions workflow to deploy backend/ to Hugging Face Spaces, update docs/specs and change local MCP port references to 8080, and add minor backend .gitignore update. History/prompts and spec artifacts for the conversational MCP chatbot feature are included.d17923b