# ctx-optimize

> A deterministic code knowledge graph for coding agents: one static binary that gathers a repo into a plain-ndjson graph and answers structural questions — query, card, change-plan, affected, path, boundaries — each with an exact file:line.

ctx-optimize parses code with tree-sitter compiled to WASM (hosted by wazero, 12 languages embedded, others as grammar packs), stores a sorted newline-delimited JSON graph with a content-hash manifest, and answers with resolved symbols rather than matching lines. Every edge carries a confidence tier: EXTRACTED is parsed fact, INFERRED is name-matched, AMBIGUOUS is surfaced as a shortlist rather than guessed into the graph. The boundaries lane models a repo external surface — env reads, spawned binaries, HTTP hosts, served routes — as port nodes with direction, transport and credential flagging by NAME. The binary makes no LLM calls, holds no credentials at rest, and reaches the network only when explicitly asked.

## Documentation Sets

- [Abridged documentation](https://muthuishere.github.io/ctx-optimize/llms-small.txt): a compact version of the documentation for ctx-optimize, with non-essential content removed
- [Complete documentation](https://muthuishere.github.io/ctx-optimize/llms-full.txt): the full documentation for ctx-optimize

## Notes

- The complete documentation includes all content from the official documentation
- The content is automatically generated from the same source as the official documentation