evidentloop 0.1.0a2

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, record a decision, and ask your coding agent to update the same report without re-reviewing or changing code.

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 machine-readable decision export; the report's primary action copies the same payload for a coding agent.

Schema 0.4 report overview	Human adjudication and copy-to-AI flow

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

After recording decisions, click 复制给 AI 更新报告 (“Copy for AI to update report”). In the workspace that contains the original audit.json, paste the block into an EvidentLoop-enabled coding agent. By default, revise updates only audit.json and audit.html in the same directory; other files stay untouched. Explicit --out creates a new copy. Stale, conflicting, or ambiguous feedback stops without guessing; report revision never modifies business code or invokes model review.

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 0.4, report revision, exact changed-line evidence	Automatic fixes or model re-review from feedback
In-place feedback revision with paired JSON/HTML rollback and explicit copy output	Automatic stale-feedback merging or cross-workspace report search
Complete, partial, failed, and inconclusive states	Review targets other than Git diffs

Integration and development

Use EvidentLoop through the Skill as a standalone product; no workflow integration is required.

EvidentLoop can also provide verifiable audit results to a workflow: diff_version identifies the reviewed Git diff, while report_version identifies the formal audit.json revision. Sopify is the first workflow to integrate this capability, and other workflows are welcome to use the same public integration contract.

Integrators can follow the prepare -> external review -> finalize path 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

Alpha feedback and bug reports are welcome in GitHub Issues.

License

Licensed under the MIT License.