Skip to content

RAG Pipeline

Planned Architecture (Future Phases)

The RAG pipeline is implemented in Phase 3 (Weeks 8–11).


Overview


Indexing

Document Chunking

# Chunking configuration
CHUNK_SIZE = 512
CHUNK_OVERLAP = 64
METADATA_FIELDS = ["company_id", "filing_id", "period", "section"]

Embedding Model

Indexing Pipeline


Retrieval

Dense Retrieval (pgvector)

SELECT content, 1 - (embedding <=> $1::vector) AS score
FROM document_chunks
ORDER BY embedding <=> $1::vector
LIMIT 10;

Sparse Retrieval (Elasticsearch)

Planned Architecture (Future Phases)

Elasticsearch hybrid retrieval is implemented in Phase 3.

Hybrid Fusion


Generation

Context Injection

LLM Configuration


Evaluation