How noho/dayu-agent works
Dayu Agent is a buy-side financial research assistant. Its main workflow is to download or upload filings, process them into a local research corpus, answer questions using that corpus, and optionally produce auditable chapter-based reports and rendered documents. Interfaces include CLI, early-stage Web UI, and a text-based WeChat integration. The graph reflects documented capabilities and relationships visible in the README or indexed source references; unsampled implementation details are not inferred.
Main entry points: commands CLI commands · chat_service.py Chat service · async_agent.py Agent engine · pipelines Market pipelines
Components 20
User interfaces
- CLI commands
dayu/cli/commands - Web UI
dayu/web/streamlit/pages - WeChat interface
dayu/wechat
Research runtime
- Chat service
dayu/services/chat_service.py - Agent engine
dayu/engine/async_agent.py - Web research tools
dayu/engine/tools/web_tools.py
Financial documents
- Filing ingestion
dayu/fins/ingestion/service.py - Market pipelines
dayu/fins/pipelines - Market downloaders
dayu/fins/downloaders - Filing repository
dayu/fins/storage - Document processors
dayu/fins/processors - Document models
dayu/fins/domain/document_models.py - Filing search
dayu/fins/tools/search_engine.py
Report writing
- Report pipeline
dayu/services/internal/write_pipeline/pipeline.py - Document rendering
dayu/render/render.py
Host and workspace
- Workspace setup
dayu/cli/commands/init.py - Run and session host
dayu/host - Conversation state
dayu/host/conversation_store.py
Other components
- Investor
- Model provider
Connections 22
- Investor → CLI commands: runs commands
- Investor → Web UI: uses analysis UI
- Investor → WeChat interface: asks questions
- CLI commands → Workspace setup: initializes workspace
- CLI commands → Filing ingestion: requests filing work
- Filing ingestion → Market pipelines: dispatches ingestion
dayu/fins/ingestion/service.py - Market pipelines → Market downloaders: downloads filings
dayu/fins/pipelines/cn_pipeline.py - Market pipelines → Filing repository: uses repository
dayu/fins/pipelines/cn_pipeline.py - Market pipelines → Document processors: uses processors
dayu/fins/pipelines/cn_pipeline.py - Filing repository → Document models: stores document records
dayu/fins/storage/_fs_storage_infra.py - CLI commands → Chat service: submits questions
- WeChat interface → Chat service: supports questions
- Web UI → Chat service: supports analysis
- Chat service → Agent engine: runs agent
- Agent engine → Model provider: requests completion
- Agent engine → Web research tools: uses research tools
- Chat service → Run and session host: coordinates execution
dayu/services/chat_service.py - Run and session host → Conversation state: persists conversations
dayu/host/conversation_session_archive.py - Agent engine → Filing search: queries filings
- CLI commands → Report pipeline: starts report writing
- Report pipeline → Run and session host: executes chapters
dayu/services/internal/write_pipeline/pipeline.py - CLI commands → Document rendering: renders reports