eval-builder

Your agent turns your app's real logs into an eval suite, and tells you which of its judges you can actually trust.

uvx eval-builder ingest ./logs                 # OpenAI, Anthropic, Langfuse, OpenTelemetry or JSONL
uvx eval-builder select -n 40 && uvx eval-builder draft
uvx eval-builder judge-check                   # after your agent has run each judge a few times

Not on PyPI yet. Until the first release, run it straight from GitHub by replacing uvx eval-builder with uvx --from git+https://github.com/Abelo9996/eval-builder eval-builder. setup registers uvx eval-builder mcp, so it works once the package is on PyPI.

eval-builder is not another eval platform. It builds the suite and checks the judges, then exports to the tools you already run: promptfoo, DeepEval, Inspect AI, or plain JSONL. It never calls a model. Your coding agent (Claude Code, Codex, Cursor) does the thinking through the MCP server; eval-builder does the selection, the bookkeeping and the statistics, and writes down the evidence.

Example: real output

Everything below comes from runs on this machine (Apple M4, 16 GB) that are committed in examples/mt-bench/. The data is the LMSYS MT-Bench human judgments dataset (CC BY 4.0): 80 two-turn questions answered by 6 models, plus pairwise verdicts from expert human judges.

Logs to suite. ingest read 480 conversations (OpenAI chat format, redaction on, 0 matches). select -n 24 --stratify category removed 400 exact duplicates (each question was asked to six models), clustered the 80 unique conversations into 9 topics and picked 24 covering all 8 categories, 19 of them failures. From suite/report.md:

| trace           | cluster | represents | why it was picked |
| q81-alpaca-13b  | 0 | 6 | failure (negative user feedback); 62% of unique traces are failures and at least 30% of picks are reserved for them |
| q82-alpaca-13b  | 0 | 6 | adds variety within cluster 0 (11 traces, 14%; response, previous response, previous); least similar to cases already picked there |
| q156-alpaca-13b | 3 | 6 | covers category=humanities (10 unique traces, 12%) |

The agent wrote expected behavior for the 24 cases, and export produced files that Inspect AI 0.3.277, DeepEval 4.2.8 and promptfoo 0.124.0 all load (see verify_exports.sh).

Which judges can you trust? Five local LLM judge setups (four models through Ollama, one of them also at temperature 0) and two controls that are not language models judged 80 answer pairs where human experts picked a winner. Each judge ran 5 times per pair, 5 more with the answers swapped, and 5 more with an irrelevant paragraph appended to one answer: 8,400 calls, 0 errors. eval-builder judge-check, from judges/report.md:

n = 80 pairs per judge; brackets are 95% intervals. No judge passed. The closest, qwen2.5 7B, reaches kappa 0.40 to 0.45 with the human experts, but its verdict flips on a quarter of the pairs when the answer order is swapped, and setting the temperature to 0 removes the run-to-run flips without removing that order effect. llama3.2 3B picked whichever answer it saw second 80% of the time. The longer-answer rule never flips and agrees with humans 66% of the time, which is why stability and accuracy alone are not enough: the padding probe catches it.

The same check on the suite's own pass/fail judge (qwen2.5 7B, 24 cases, no human labels) gives unstable: its verdict changed across 5 identical calls on 7 of 24 cases (29%, interval 15% to 49%).

How it works

The verdict thresholds are explicit flags with defaults: flip rate at most 20% of cases, position consistency at least 80%, padding helps at most 10% of cases, kappa at least 0.4 on at least 20 human-labeled cases. A judge without human labels is never called trustworthy.

Setup for agents

uvx eval-builder setup          # shows what it would change
uvx eval-builder setup --yes    # applies it

setup registers the MCP server (uvx eval-builder mcp) with Claude Code (claude mcp add --scope user), Codex ([mcp_servers.eval-builder] in ~/.codex/config.toml) and Cursor (~/.cursor/mcp.json), and copies the agent instructions to ~/.claude/skills/eval-builder/ and ~/.codex/skills/eval-builder/. It backs up any file it edits and does nothing on a second run. The workflow the agent follows is in skills/eval-builder/SKILL.md.

MCP tools: ingest, select, draft, list_cases, update_case, set_rubric, validate, judge_plan, judge_check, export, report, status. The judge runner is CLI only, because it executes a command.

What it can't do

• It does not write expected behavior or human labels. The agent drafts expected behavior with you; labels must come from people. Without labels, judge-check can tell you a judge is unstable or biased, but not that it is right.
• Selection is lexical. Two requests that mean the same thing in different words can land in different clusters, and near-duplicate detection only catches close textual matches.
• Redaction is pattern-based. Names, street addresses and free-form secrets get through. Look at traces.jsonl before sharing a workspace.
• The bias probes cover answer order and irrelevant length only. Self-preference, style bias and rubric misreadings are not measured.
• Small samples give wide intervals. The report prints them; read them.
• It does not run your app or your eval. The exported files do that in promptfoo, DeepEval or Inspect AI.

Privacy and safety

Everything runs locally. eval-builder makes no network calls and sends nothing anywhere. The only process it starts is the judge command you name with judge-run --enable-judge-plugin. Redaction is on unless you pass --no-redact, and the report lists what was replaced (counts and kinds, never the values).

License

MIT. See LICENSE.