Docwise — Hybrid RAG Documentation Assistant
Docwise is a documentation assistant that combines keyword and semantic search to answer technical questions with traceable sources. The hosted Python application fuses BM25 and MiniLM ONNX vector results through reciprocal rank fusion before generating a cited answer.
The problem
Docwise is a documentation assistant that combines keyword and semantic search to answer technical questions with traceable sources. The hosted Python application fuses BM25 and MiniLM ONNX vector results through reciprocal rank fusion before generating a cited answer.
My contribution
I integrated structure-aware ingestion, hybrid retrieval, citation validation, the Docwise browser interface, and a lightweight ONNX deployment for Vercel.
Product features
- Two Markdown documents provide 39 heading-aware indexed passages.
- BM25 keyword search and MiniLM ONNX embeddings combine through reciprocal rank fusion.
- Answers include validated source labels and expandable retrieved passages.
- Unsupported questions return an explicit documentation boundary.
- FastAPI validates inputs and Upstash Redis enforces shared public-demo limits.
Engineering work
- Hybrid retrieval: Fuse BM25 and MiniLM ONNX vector rankings through reciprocal rank fusion before building a source-grounded prompt.
- Traceable answers: Validate citation labels against retrieved passages and expose those passages in the interface.
- Serverless delivery: Package a lightweight read-only index and enforce shared Redis admission before model calls.
Outcome & scope
Makes 39 passages across 2 documents searchable with inspectable citations. Live checks verified a FastMCP answer against its retrieved passage and a refusal for an unrelated question; retrieval accuracy and business impact have not been benchmarked.
Public documentation demo · 2 documents · Shared daily allowance