How pandaai-tech/panda_factor works
Project Type & Purpose
• This is a full-stack, microservices-style Python platform for quantitative finance (“PandaAI量化因子库”), providing:
• Data ingestion & cleaning (panda_data_hub)
• Core data readers & utilities (panda_data)
• Factor‐construction library & workflows (panda_factor)
• REST API server to serve factors (panda_factor_server + “server” folder)
• LLM-driven chat service (panda_llm)
• Web frontend (panda_web)
• Shared common modules (panda_common)
• Persistent storage (MongoDB)
• External data sources (Tushare, RiceQuant, XTquant, etc.)
File Structure
• Top-level packages correspond to separate Python services or libraries:
• panda_common: config, DB handler, logging, shared models/utilities
• panda_data: low-level readers and quick-run scripts
• panda_data_hub: HTTP routes + scheduled tasks for nightly data cleaning/ingestion
• panda_factor: factor generation workflows & CLI entrypoint
• panda_factor_server (& server/): REST API layer exposing factor services
• panda_llm: chat endpoints, LLM service adapters, MongoDB integration
• panda_web: static UI assets + web server entrypoint
• Deployment artifacts: Dockerfile(.server), setup.py in each module, requirements.txt
README Insights
• Describes:
• Quant factor library + visualization
• Two factor‐writing modes (Python & formula)
• Data pipelines update nightly via schedulers
• MongoDB back-end
• Frontend + API for users to query factors or chat with LLM