How andersonvc/alpha-pulse works
Project type and purpose
• This is a backend‐focused data‐processing pipeline (an “agent system”) that automates retrieval, parsing, and structuring of SEC 8‐K filings using LLMs. It is not a full‐stack web app but rather a CLI/orchestrated workflow tool that integrates with external APIs (SEC EDGAR, Polygon, OpenAI).
File structure
• Top‐level scripts/ directory: contains executable entry points (run_workflow, run_edgar_query, etc.).
• src/alpha_pulse/:
• agent_workflows/: high-level workflow definitions (e.g., doc_analyzer.py, parse_8k_502.py).
• agents/: LLM-driven “agent” implementations specialized for EDGAR filings and news.
• clients/: low-level HTTP clients for SEC EDGAR and Polygon, plus rate limiting.
• graphs/: orchestration layer that ties clients and agents into end‐to‐end pipelines.
• storage/: persistence layer (DB clients, schema, publisher modules).
• tools/: utility functions for news and polygon-specific logic.
• types/: Pydantic/TypedDict definitions modeling filings and DB tables.
• Configuration: requirements.txt, pyproject.toml, placeholder.env.
• Tests in tests/: unit tests for agents, clients, storage.
README for additional insights
• Confirms key workflows: fetch 8‐K, parse items with LLM, output structured JSON.
• Mentions “agent system,” “LLM‐Powered Understanding,” “Automatic EDGAR Integration.”
Elements
• Blue for external services
• Green for client modules
• Yellow for agent/workflow modules
• Orange for storage/persistence
• Rectangles for code modules
• Clouds for external APIs