WeKnora
WeKnora is Tencent's open-source LLM knowledge framework for enterprise RAG Q&A, ReAct agents, Wiki mode, multi-source document ingestion (PDF/Word/Excel/etc.), and multi-workspace RBAC. This template deploys the core stack: Web UI + App API + Docreader + PostgreSQL (ParadeDB/pgvector) + Redis.
模板內的服務
WeKnora
Open-source knowledge framework by Tencent — RAG Q&A, ReAct Agent, Wiki mode, multi-format document parsing, and enterprise workspace RBAC.
Upstream: github.com/Tencent/WeKnora
Official site: weknora.weixin.qq.com
Services in this template
| Service | Image | Role |
|---|---|---|
| frontend | wechatopenai/weknora-ui | Web UI (public domain) |
| app | wechatopenai/weknora-app | Backend API (:8080) |
| docreader | wechatopenai/weknora-docreader | Document parse gRPC (:50051) |
| postgres | paradedb/paradedb:v0.22.2-pg17 | DB + vector (pgvector) |
| redis | redis:7.0-alpine | Cache / stream / MQ |
After deploy
- Open your PUBLIC_DOMAIN — complete first-time registration (unless disabled).
- In Settings, configure an LLM provider (OpenAI / DeepSeek / Qwen / Ollama / etc.).
- Create a knowledge base, upload documents, start Q&A or Agent mode.
- For production: set strong
JWT_SECRET/ DB / Redis passwords; prefer private network; considerDISABLE_REGISTRATION=trueafter bootstrap.
Resource tips
- Recommended ≥ 4 vCPU / 8 GB for comfortable document parsing + embeddings.
- Optional compose profiles (Neo4j / MinIO / Langfuse / Milvus …) are not included; add them later as extra services if needed.
- Configure external LLM/embedding API keys inside the product UI or via app env after deploy.
License
MIT — Tencent/WeKnora