AI-native stock screening engine with full-market discovery, LLM ranking, risk-aware scoring, and auditable evaluation. AlphaSift addresses the need for sophisticated, transparent, and agent-friendly stock selection by providing a multi-layered analysis framework. It benefits technical users and AI agents by automating complex screening processes with auditable strategies and deterministic scoring.

How It Works

AlphaSift operates through a multi-stage process: L1 deterministic screening applies hard filters and factor scoring across the entire market universe. L2 introduces optional LLM ranking for nuanced reasoning, thesis generation, and risk assessment. L3 offers pluggable post-analysis, defaulting to a local scorecard but extensible to external analyzers like DSA. The engine also features hotspot discovery for identifying market trends and daily feature enrichment for technical indicators, all driven by auditable YAML strategies and saved for T+N evaluation.

Quick Start & Requirements

• Installation: Install in editable mode using pip install -e ..
• Configuration: Copy .env.example to .env and configure LLM API keys (GEMINI_API_KEY, OPENAI_API_KEY, DEEPSEEK_API_KEY) or LiteLLM settings for LLM ranking. A Tushare token is recommended for enhanced data sourcing.
• Prerequisites: Python environment. LLM API keys are optional but required for LLM-based ranking.
• Usage: Run commands like alphasift screen dual_low --no-llm for basic screening, alphasift quickstart for a demo, or alphasift serve for a local JSON API.
• Documentation: Key documentation includes docs/usage.md, docs/configuration.md, and docs/strategy-guide.md.

Highlighted Details

• Agent-Native Interface: Includes SKILL.md detailing callable interfaces for AI agents, facilitating integration into automated workflows.
• Hotspot Discovery: Capable of identifying market hotspots, resolving topics into detailed payloads with timeline evidence, leader stocks, and explicit source confidence, including fallback metadata.
• Pluggable Post-Analysis: Supports a default local scorecard for L3 analysis, with optional integration for DSA or other external HTTP analyzers.
• T+N Evaluation Loop: Enables saving screening runs for later evaluation against newer market snapshots, incorporating transaction costs and outcome tagging for performance review.
• Robust Data Sourcing: Integrates multiple A-share market snapshot and daily K-line data sources with automatic fallback mechanisms and explicit marking of stale or cached data.

Maintenance & Community

The project is maintained by ZhuLinsen. Specific community channels (e.g., Discord, Slack) or sponsorship details are not explicitly mentioned in the provided README.

Licensing & Compatibility

• License: Apache License 2.0.
• Compatibility: The permissive Apache 2.0 license allows for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

Strategies requiring daily K-line features currently enrich only the top L1 candidates, not the entire historical market. AlphaSift is not a comprehensive backtesting engine or portfolio execution system. T+N evaluation is not a rigorous event-study backtest and does not model dividends, suspensions, slippage, or rebalancing constraints. Usage of the Tushare data source is contingent on the user's own token, point balance, and permissions.