How dataelement/bisheng works
Project Type and Purpose
• This is a full-stack, containerized enterprise LLM application platform (“BISHENG”) that bundles:
• A monolithic Python backend (FastAPI + Celery)
• Two React-based frontends (“client” for user chat & “platform” for workflow & admin)
• Self-hostable infrastructure via Docker Compose, including DB, cache, vector stores, message broker, search, office server, and reverse proxies
• Purpose: Provide an end-to-end devops platform to author, orchestrate and run complex LLM workflows (chat, document parsing, multi-agent flows), with enterprise features (RBAC, SSO, observability).
Key Directories & Architectural Signals
• docker/: Compose files & container configs for bisheng backend, worker, MySQL, Redis, Elasticsearch, Milvus, OnlyOffice, Nginx
• src/backend/: Python service using FastAPI (API router, services, cache, database models), Celery tasks & workers, LLM/chat adapters, workflow engine
• src/bisheng-langchain/: LangChain integration library, custom agents, tools, document loaders, RAG pipelines
• src/frontend/client/: React “chat client” served by Nginx, communicates via REST & SSE/WebSockets
• src/frontend/platform/: React “platform UI” for building workflows, flows and admin functions
• Configuration / CI: .drone.yml, GitHub workflows, pre-commit, Dockerfiles