Add-on: PDF RAG Pipeline (Legacy Narrow Variant)

Use this skill for PDF-only RAG. For mixed formats (Markdown, text, HTML, CSV), prefer addon-rag-ingestion-pipeline.

Compatibility

• Best with architect-python-uv-batch or architect-python-uv-fastapi-sqlalchemy.
• Can also support Next.js by keeping ingestion/indexing in a Python worker service.

Inputs

Collect:

• DOC_SOURCE: local folder, object storage, or uploaded files.
• EMBED_PROVIDER: openai or sentence-transformers.
• VECTOR_STORE: pgvector, chroma, or existing store.
• CHUNK_SIZE: default 1000.
• CHUNK_OVERLAP: default 150.

Integration Workflow