LLM code context (structured call sites)

Full workflow

Complete guide (limitations, CI, published lens): llm-code-context

Why use CCE

LLM agents reviewing or refactoring code need ground truth about what APIs are invoked — not guesses from imports or comments. Unstructured repo dumps exceed context windows and hallucinate call paths.

CCE emits compact JSON with file, line, method, and signature for every mapped call site — ideal input for RAG, agent prompts, and automated remediation workflows.

Example command

cce -folder . -language AUTO -filter all -format json

Analysis recipe

This use case is registered as recipe llm-code-context — run with:

cce run -folder . -language AUTO -recipes llm-code-context -output llm-code-context.json

Published lens

https://releases.stackgen.com/cce/lenses/llm-code-context/latest/llm-code-context_lenses.yaml