evidentloop 0.1.0a0

Traceable code review artifacts for local Git diffs

EvidentLoop

Turn local Git diffs into evidence-backed, traceable audit reports.

English · 简体中文

EvidentLoop turns a local Git diff into a single HTML report. Any finding included in the risk score is tied to a real changed line, so you can verify it, mark false positives, add context, and export your decisions.

The Alpha report UI and review text are Simplified Chinese.

Quick start

Requirements: Git, Python 3.10 or newer, uv, Node.js/npx for Skill installation, and a coding agent with Skill support.

# Try the offline demo
uvx evidentloop demo

# Install the CLI and Skill
uv tool install evidentloop
npx skills@latest add evidentloop/evidentloop --skill evidentloop -g
evidentloop doctor

pipx install evidentloop is the CLI fallback.

Inside the Git repository to inspect, ask your coding agent:

Use EvidentLoop to audit my staged changes and generate the HTML report.

Output

File	Purpose
audit.json	Validated audit record linking Git changes to findings.
audit.html	Single-file report showing the relevant diff and browser-local decisions.
audit-feedback.jsonl	Optional decision export. In this Alpha, exported feedback does not update the report or change code.

EvidentLoop self-audit overview	Offline demo: finding, changed line, and decision

Open the HTML locally, or publish a redacted copy to static hosting. Feedback stays in each viewer's browser until exported; the report is shareable, but it is not a multi-user review service.

EvidentLoop uses the model already available in your coding agent. It never executes commands found in the diff or model output.

How it works

complete means the host returned a valid, complete result. What it finds still depends on the host model and the context it received.

Alpha scope

Supported	Not supported
Local Git staged, unstaged, ref, and range diffs	Folder diff, file-only review, or remote PR URLs
Added, modified, deleted, renamed, and binary-file metadata	Automatic fixes or command execution
Schema 0.3, findings tied to exact changed lines, nearby Git context	Feedback ingestion or automatic report revision
Complete, partial, failed, and inconclusive states	Review targets other than Git diffs

Integration and development

Most users should use the Skill. Host integrators can use the prepare -> external review -> finalize path documented in AI host integration. The public Python API is available from evidentloop.api.

For local development:

python -m pip install -e '.[dev]'
python -m pytest -q
python -m ruff check .
python -m build

References: Pages and sample report · V0 scope · Data model · AI host integration

License

Licensed under the MIT License.