codira

codira: your code ferret

codira is a repository-local indexing and context retrieval tool for agent-assisted development. It gives coding agents a compact, deterministic map of the repository so they can spend fewer tokens rediscovering files, symbols, call paths, and docstring issues.

It builds a SQLite index inside the target repository and currently provides:

• exact symbol lookup
• docstring auditing
• deterministic local semantic embeddings
• deterministic context generation for natural-language queries
• static call-edge and callable-reference inspection
• plugin discovery for official and third-party analyzers/backends

Used next to a coding agent, codira is most useful before broad exploration: index once, ask a focused question, then hand the agent a small context pack instead of asking it to scan the whole tree.

For normal use, install the published PyPI package instead of a development checkout:

pip install codira-bundle-official

Use the source checkout only when you are developing codira itself or need a specific development branch.

Documentation scope

The repository documentation covers:

• how to install and use the current tool locally
• how contributors should validate changes
• the small set of repository-owned helper scripts
• the branch and ADR workflow used for repository governance
• a baseline architecture description for the accepted ADR-004 migration
• the active ADR trail, listed dynamically as new ADR files are added

Current architecture status

The current implementation is SQLite-backed and registry-driven.

It now supports:

• one active backend per repository index
• multiple analyzers in one indexing run
• deterministic indexing for tracked *.py, *.c, and *.h files

The accepted future architectural direction is documented in ADR-004.

The current architecture and migration state are documented under Architecture.