Chat with your codebase with Rust framework Candle and Tree-sitter

In this 30-minute talk, we’ll take a step-by-step approach to building a tool that lets you ask questions about your codebase and get meaningful answers with references. Using Candle for embeddings and vector search, a Tree-sitter for syntax parsing, and an LLM to compose the responses, we’ll hack an intelligent code-querying system.

Questions we'll discuss and find answers to:

1) Why is querying large codebases difficult?

What does it involve?

- Introduction to Candle and Tree-sitter.

2) How to parse and extract structured data from your codebase (e.g., functions, classes, and comments).

- Choose the right model.

Generate embeddings for the extracted code and use vector search for context-aware queries.

- Connect parsed code with semantic embeddings to make it searchable.

3. How the query pipeline operates from input to output.  

This talk is tailored for Rust developers interested in combining Machine learning models and programming language parsing into their applications.

Starting from: $536

Renew Your Mind at LambdaConf 2025

Buy tickets