| --- |
| title: Demo Agentic Service Data Eyond |
| emoji: 🏆 |
| colorFrom: yellow |
| colorTo: pink |
| sdk: docker |
| pinned: false |
| --- |
| |
| Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference |
|
|
|
|
| How to run: |
| `uv run --no-sync uvicorn main:app --host 0.0.0.0 --port 7860` |
|
|
|
|
| Agent |
| Orchestrator : intent recognition, orchestrate, and plannings |
| Chatbot : have tools (retriever, and search), called by orchestrator |
|
|
|
|
| APIs |
| /api/v1/login -> login by email and password |
|
|
| /api/v1/documents/{user_id} -> list all documents |
| /api/v1/document/upload -> upload document |
| /api/v1/document/delete -> delete document |
| /api/v1/document/process -> extract document and ingest to vector index |
| |
| /api/v1/chat/stream -> talk with agent chatbot, in streaming response |
| /api/v1/rooms/{user_id} -> list all room based on user id |
| /api/v1/room/{room_id} -> get room based on room id |
| |
| |
| Config |
| - Agent: system prompt, guardrails |
| - others config needed |
| |
| DB |
| - using postgres as db |
| - we can use pg vector from this db also |
| - use redis for caching response, same question will not re-processed for 24 hour |
| |
| Document |
| - service to manage document, upload, delete, log to db |
| |
| Knowledge |
| - service to process document into vector, until ingestion to pg vector |
| |
| Middleware |
| CORS: |
| - allow all |
| Rate limiting: |
| - upload document: 10 document per menit |
| Logging: |
| - create clear and strutured logging for better debuging |
| |
| Models |
| - Data models |
| |
| Observability |
| - Langfuse traceability |
| |
| RAG |
| - retriever service to get relevant context from pg vector |
| |
| storage |
| - storage functionality to communicate with storage provider |
| |
| tools |
| - tools that can be use by agent |
| |
| Users |
| - Users management, to get user indentity based on login information. |
| |
| Utils |
| - Other functionality |
