rajdeepofgithub/sec-filings-rag

How rajdeepofgithub/sec-filings-rag works

SEC Filings RAG ingests filings and earnings-call transcripts, structures their content into section-aware chunks, indexes them for dense and sparse search, and answers questions with cited evidence. The principal query path combines retrieval, deduplication, rank fusion, reranking, grounded generation, citation checks, and a confidence signal. Separate evaluation and baseline artifacts measure behavior. The graph includes documented capabilities even where implementation excerpts are absent; those links cite the README, and unsampled implementation details are not inferred.

Main entry points: generator.py Grounded Answer Generator · loader.py Document Loader · pipeline.py Search Orchestrator · indexer.py Index Builder

Components 22

Document Ingestion

Document Loader src/loader.py
Document IR src/ir.py
Table Profiler src/table_profile.py
Table Converter src/table_convert.py
Section Detection src/sections.py
Structure-aware Chunker src/chunker.py

Corpus and Indexing

Corpus Builder src/corpus.py
Text Embedder src/embedder.py
Index Builder src/indexer.py

Hybrid Search

Search Orchestrator src/pipeline.py
Dense and Sparse Retrieval src/retriever.py
Dedup and Rank Fusion src/fusion.py
Cross-encoder Reranker src/reranker.py

Answer and Verification

Grounded Answer Generator src/generator.py
Citation Verifier src/verify.py

Evaluation

Answer Grader src/grading.py
Baseline Runs baseline

Other components

Questioner
Filings and Transcripts
ChromaDB
BM25 Index
OpenAI API

Connections 25

Filings and Transcripts → Document Loader: loads src/loader.py
Document Loader → Document IR: builds blocks src/loader.py
Document Loader → Table Profiler: profiles tables src/loader.py
Document Loader → Table Converter: converts tables src/loader.py
Corpus Builder → Document Loader: loads documents src/corpus.py
Corpus Builder → Section Detection: attributes sections src/corpus.py
Corpus Builder → Structure-aware Chunker: chunks documents src/corpus.py
Structure-aware Chunker → Document IR: reads blocks src/chunker.py
Index Builder → Corpus Builder: builds from src/indexer.py
Index Builder → Text Embedder: embeds chunks src/indexer.py
Index Builder → ChromaDB: writes vectors src/indexer.py
Index Builder → BM25 Index: builds index src/indexer.py
Questioner → Search Orchestrator: submits query src/pipeline.py
Search Orchestrator → Dense and Sparse Retrieval: runs searches src/pipeline.py
Dense and Sparse Retrieval → Text Embedder: embeds query src/retriever.py
Dense and Sparse Retrieval → ChromaDB: queries vectors src/retriever.py
Dense and Sparse Retrieval → BM25 Index: queries terms src/retriever.py
Search Orchestrator → Dedup and Rank Fusion: deduplicates and fuses src/pipeline.py
Search Orchestrator → Cross-encoder Reranker: reranks candidates src/pipeline.py
Grounded Answer Generator → Search Orchestrator: retrieves context src/generator.py
Grounded Answer Generator → OpenAI API: generates answer README.md
Grounded Answer Generator → Citation Verifier: checks citations README.md
Grounded Answer Generator → Questioner: returns cited answer README.md
Answer Grader → Grounded Answer Generator: grades answers src/grading.py
Baseline Runs → Answer Grader: provides eval cases README.md

More Python diagrams: tenstorrent/tt-xla · trademaster-ntu/finworld · vishakhasadhwani/llm-deployment-demo · vitusbenson/greenearthnet · wdbm/spin

Drawn by GitDiagram on September 29, 2026 from the repository's file tree, README and source files. Read as Markdown ·