How tencent/weknora works
WeKnora is an enterprise knowledge framework for document ingestion, cited RAG answers, multi-step agent work, and knowledge wikis. The principal workflow brings documents into shared knowledge bases, parses and indexes them, then retrieves relevant content for user questions and returns answers with citations. The graph also preserves the README’s major agent, wiki, integration, storage, and operations capabilities. Relationships beyond the sampled source are based on README statements; the source excerpts do not expose the full request-to-retrieval implementation.
Main entry points: engine.go Agent execution · chat_pipeline.go RAG answer pipeline · src Web application · routes_knowledge.go Document ingestion
Components 25
User experience
- Web application
frontend/src - Chat experience
frontend/src/composables/useChatStreamHandler.ts - Knowledge management
frontend/src/api/knowledge-base/index.ts
Knowledge processing
- Document ingestion
internal/router/routes_knowledge.go - Document parsing
internal/infrastructure/docparser/engines.go - DocReader parsers
docreader/parser/registry.py - Data-source sync
internal/datasource/connector/feishu/core/engine.go - Semantic retrieval
internal/container/engine_factory.go - Knowledge wiki
frontend/src/api/wiki/index.ts
Answering and agents
- RAG answer pipeline
internal/application/service/chat_pipeline/chat_pipeline.go - Agent execution
internal/agent/engine.go - Agent skills
internal/agent/skills/manager.go - Session sandbox
internal/sandbox/session_manager.go - Browser skill
internal/browserskill/manager.go
Platform services
- Backend API
internal/router/routes_knowledge.go - Storage backends
internal/application/service/storagebackend.go - Model providers
internal/models/providers/volcengine.go - Session streaming
internal/stream/redis_manager.go - Tracing
internal/tracing/langfuse/manager.go
Channels and integrations
- MCP services
internal/mcp/manager.go - MCP server
internal/mcpserver/server.go - Messaging channels
internal/im/service.go - Embedded widget
frontend/src/embed-main.ts
Other components
- User
- External systems
Connections 24
- User → Web application: uses
README.md - Web application → Chat experience: asks questions
README.md - Web application → Knowledge management: manages knowledge
README.md - Knowledge management → Backend API: sends requests
- Backend API → Document ingestion: routes requests
internal/router/routes_knowledge.go - Document ingestion → Document parsing: parses documents
- Document parsing → DocReader parsers: delegates formats
internal/infrastructure/docparser/engines.go - Data-source sync → Document ingestion: syncs documents
- Chat experience → RAG answer pipeline: requests answers
README.md - RAG answer pipeline → Semantic retrieval: retrieves knowledge
- RAG answer pipeline → Model providers: generates answers
- RAG answer pipeline → Session streaming: streams responses
- RAG answer pipeline → Agent execution: dispatches tasks
README.md - Agent execution → Agent skills: uses skills
internal/agent/engine.go - Agent execution → Session sandbox: runs tools
internal/agent/engine.go - Agent execution → Browser skill: controls browser
README.md - Agent execution → MCP services: refreshes tools
internal/agent/engine.go - Agent execution → Tracing: records steps
internal/agent/engine.go - Knowledge wiki → Semantic retrieval: organizes knowledge
README.md - Storage backends → External systems: stores files
README.md - MCP server → External systems: serves AI tools
README.md - MCP services → External systems: connects services
README.md - Messaging channels → External systems: serves Q&A
README.md - Embedded widget → External systems: embeds chat
README.md