Chunking a codebase by a fixed token count, the same way you'd chunk an article, routinely splits a function in half — and a half-function chunk is nearly useless for a 'how does X work' question.
Chunk by structure, not by token count
- Parse the code (using the language's AST or a simple structural parser) and chunk by function or class boundary, not by arbitrary token windows.
- Include a function's signature and docstring even when the body is chunked separately, so retrieval has something to match against beyond raw logic.
- Index file paths and import relationships as metadata — 'where is this function called from' is a common real question retrieval alone can't answer without that structure.
What this enables
A code-aware RAG assistant that retrieves whole, coherent functions answers 'how does X work' meaningfully better than one retrieving arbitrary fifty-line windows that cut across logic.
See RAG chunking strategies, a guide.
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU Kanpur) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. Email pranjulrathour41@gmail.com.
