Skip to main content

The deterministic contract layer for AI-modified codebases — effects, purity, contracts, and blast radius, with zero LLM calls.

Project description

CodeGraph IR (CGIR)

The deterministic contract layer for AI-modified codebases. Agents write more of the code than you can review. CGIR reads a repo and — with zero LLM calls — tells you what each component is (effects, purity, contract, entrypoints, call surface) and whether a change altered it. Think ruff, but for architecture instead of style: fast, static, hallucination-free.

Distributed as codegraph-ir; the command and import package are both cgir.

Install

uv tool install codegraph-ir      # isolated CLI (recommended); or: pipx install codegraph-ir
cgir --version

For library/agent use in a project: uv pip install codegraph-ir (extras: [mcp] for the agent server, [api] for the HTTP surface, [llm] for regeneration).

The local loop

cgir scan .                          # build the .cgir index (Python, TypeScript, Go)
cgir watch .                         # keep it live: re-scan + show contract drift on save
cgir pack app.service.charge --repo .   # minimal context bundle for one component
cgir impact app.service.charge          # blast radius: affected callers, entrypoints, tests
cgir impact app.service.charge --candidate new.py --repo .   # radius narrowed by the real delta
cgir verify app.service.charge --candidate new.py --repo .   # contract-check an edit
cgir hook install                    # pre-commit seatbelt: block contract-breaking commits
cgir lsp                             # editor squiggles for pin violations + drift (cgir[lsp])

pack → edit → impactverifyhook, with watch keeping the index fresh underneath — an always-on membrane you and your agent both consult.

Gate CI on contract drift

The GitHub Action scans a PR's base and head and fails the build on drift — a pure function that starts hitting the network, a service that stops persisting, a new POST /admin route — deterministically, with no per-seat LLM cost:

- uses: asonkiya/llm-semantic-compilers@v0
  with:
    fail-on: "effect-gain:net effect-gain:fs effect-gain:db effect-loss:net effect-loss:fs effect-loss:db"

The default rule set is evidence-based: replaying real commit history, the I/O effect rules fire on ~0–10% of commits, each a genuine change in a component's I/O surface.

Agents as first-class users

cgir mcp --index .cgir serves the index over MCP. Instead of grepping, an agent calls search / pack to load minimal context, impact to see what a change touches, and verify / impact_of_change to contract-check its own edit before proposing it. See examples/ for a worked agent-PR case study.

Setup guide for agents (MCP config + CLAUDE.md snippet): docs/agents.md. Or via the pre-commit framework: hook id cgir-contract-check.

Docs

License

MIT — see LICENSE.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

codegraph_ir-0.3.0.tar.gz (209.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

codegraph_ir-0.3.0-py3-none-any.whl (132.1 kB view details)

Uploaded Python 3

File details

Details for the file codegraph_ir-0.3.0.tar.gz.

File metadata

  • Download URL: codegraph_ir-0.3.0.tar.gz
  • Upload date:
  • Size: 209.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for codegraph_ir-0.3.0.tar.gz
Algorithm Hash digest
SHA256 c55ba1ddc2422cac4a19c80a9595239a52f8653bf87e5a5704475c2733556b57
MD5 800a1f848750000ecd6dc96811312a3d
BLAKE2b-256 2acfb09e93d1f45cdaaba2b249319bcf68804488d9d8515aa04cf0199fb7f9dc

See more details on using hashes here.

Provenance

The following attestation bundles were made for codegraph_ir-0.3.0.tar.gz:

Publisher: release.yml on asonkiya/llm-semantic-compilers

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file codegraph_ir-0.3.0-py3-none-any.whl.

File metadata

  • Download URL: codegraph_ir-0.3.0-py3-none-any.whl
  • Upload date:
  • Size: 132.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for codegraph_ir-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 b0655833d42ba5623bdf6db87472e84c79aaebd2bfaa377a15d3e6e8560ccd4e
MD5 f543a5ac832dd1ee154a5d24cd494551
BLAKE2b-256 233a43d37f32dcdf8e0aa94bf3c33d9b1ee6288519427884bc44049211481f72

See more details on using hashes here.

Provenance

The following attestation bundles were made for codegraph_ir-0.3.0-py3-none-any.whl:

Publisher: release.yml on asonkiya/llm-semantic-compilers

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page