How saikrishna2004205/rfp-extracter works
The product is a bid-package RFP extractor. Its workflow is to discover isolated bid folders, parse PDF/HTML documents, classify them, preserve their content as evidence, retrieve evidence for 20 fields, and ask Gemini for a schema-validated JSON result written to the output directory. The graph follows calls visible in the sampled entry point and module excerpts; Gemini use and the documented output contract are also represented. Some module internals were not sampled, so uncertain implementation links are omitted.
Main entry points: main.py Pipeline CLI · parser.py PDF/HTML parser · extractor.py Gemini extractor · classifier.py Document classifier
Components 13
Document ingestion
- Pipeline CLI
main.py - PDF/HTML parser
src/parser.py - Document classifier
src/classifier.py
Evidence pipeline
- Evidence generator
src/evidence.py - Evidence index
src/retriever.py
Extraction
- Gemini extractor
src/extractor.py - Output schema
src/schema.py
Shared support
- Runtime settings
src/config.py - Text and file utilities
src/utils.py
Other components
- Bid package provider
- Bid files
data/input - Google Gemini
- JSON results
data/output
Connections 20
- Bid package provider → Bid files: provides files
- Pipeline CLI → PDF/HTML parser: discovers and parses
- Bid files → PDF/HTML parser: reads documents
- PDF/HTML parser → Document classifier: parsed package
- Pipeline CLI → Document classifier: classifies package
- Pipeline CLI → Evidence generator: generates evidence
- PDF/HTML parser → Evidence generator: parsed content
- Document classifier → Evidence generator: classification metadata
- Pipeline CLI → Evidence index: builds index
- Evidence generator → Evidence index: evidence records
- Pipeline CLI → Gemini extractor: requests extraction
- Gemini extractor → Evidence index: retrieves field evidence
- Gemini extractor → Google Gemini: requests candidates
- Gemini extractor → Output schema: validates output
- Pipeline CLI → JSON results: writes JSON
- PDF/HTML parser → Text and file utilities: cleans text
- PDF/HTML parser → Runtime settings: reads input settings
- Document classifier → Runtime settings: checks reference path
- Pipeline CLI → Runtime settings: reads directories
- Gemini extractor → Runtime settings: reads model settings