DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to block an AI release when eval scores regress with Eval Forge

Run a baseline-versus-candidate gate that exits non-zero and writes CI-readable reports.

Who it is forA maintainer of a RAG app or AI assistant who needs a clear pass or fail check on each pull request before a model, prompt, or retrieval change ships.

The problemMetric numbers exist, but there is no shared answer to what changed, which limits were enforced, and where the evidence is.

The outcomeA gate result with JSON, JUnit, SARIF, and Markdown reports, and a non-zero exit code when a required check fails.

What you need first

  • Python 3 with pip and a virtual environment
  • A committed evaluation artifact, or results from promptfoo, Ragas, or DeepEval
  • Git, and a GitHub repository if you want the composite Action
  • Docker, only if you use the Compose path

Steps

  1. Install the gate with pip install evalforge-ci.
  2. Write a policy file with absolute thresholds and baseline deltas, or start from examples/quality_policy.json.
  3. Run evalforge gate examples/candidate_summary.json --policy examples/quality_policy.json --baseline examples/baseline_summary.json --json build/evalforge-report.json --junit build/evalforge-junit.xml --sarif build/evalforge.sarif --markdown build/evalforge-summary.md. (inferred, check the README)
  4. Check the exit code: 0 for pass, 1 for a failed or error check, 2 for invalid input or policy.
  5. Add the reusable Action with uses: jsdhwfmax/[email protected] and set candidate, baseline, and policy inputs.

Prompt for your agent

Using the EvalForge repo at https://github.com/jsdhwfmax/EvalForge, install evalforge-ci and run `evalforge gate <candidate-summary-path> --policy <policy-path> --baseline <baseline-path> --json build/evalforge-report.json`. Expected output is a pass or fail verdict and the JSON report; verify the exit code is 0 or 1 and not 2.

Caveat. Eval Forge is v0.6.0 alpha, and the gate only judges the numbers you feed it, so the quality of the check depends on the evaluator that produced them.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page