Provides an abstract base class (BaseFetcher) for data collection stages. Each fetcher implements a standardized interface with auto-normalize, validation, error handling, and return of DimResult objects. The pattern is reusable for any multi-source data pipeline where consistent error handling and quality tracking are needed.

Implements a provider framework with a try_chain function that attempts data fetching from multiple backends (akshare, efinance, tushare, baostock, direct HTTP) in order, with automatic failover and environment variable overrides. The pattern is reusable for any application that needs resilient data access from multiple sources.

A lightweight utility to detect and reject junk text from LLM-generated content (e.g., placeholder text, prompt residuals, repetitive patterns). Uses heuristic checks like length, repetition, and known garbage phrases. Reusable in any system that consumes LLM output and needs to filter out invalid responses.

Pure Python functions implementing core financial models: DCF (2-stage + Gordon Growth), comparable company analysis, leveraged buyout, merger accretion/dilution, and 3-statement projection. All functions return structured dicts with methodology logs for auditability. The module is self-contained and can be extracted as a standalone financial modeling library.

A collection of 19 pure SVG rendering functions for financial charts: sparklines, donut charts, gauges, radar charts, candlestick charts, bar charts, etc. Each function takes data and returns SVG markup as a string, with color constants and no external dependencies. Reusable for any project needing inline SVG chart generation without browser rendering.