Automated README summary
Automatically summarized from the README on 1 Nov 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
evalr is a Python library for typed evaluation of agent systems. Evaluators judge an input such as a chat thread or workflow run and return a Pydantic-typed verdict, matching the feedback people give, so they can be trained and measured against it.
How it uses Jev
A DecisionEvaluator wraps TypeSafe's Jev through pydantic-ai, taking the same Pydantic verdict type and input type with a min_confidence threshold. Jev returns calibrated confidence; a Fallback composes it with a DSPy judge so the judge answers only where Jev is unsure.
Primitives:choicescorenoul
Technique worth stealing
Compose a calibrated decision model in front of a trained judge via Fallback, so the judge only handles cases the decision model is unsure about.
Try it
Install with uv add "evalr[all] @ git+https://github.com/alexnodeland/evalr"; needs OPENAI_API_KEY and TYPESAFE_API_KEY.
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.14.0- Project signals
jev-1.14.0- Category
jev-1.14.0- Pattern tags
jev-1.14.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectnot found in the excerpt0.19
- Shows a System One patternnot found in the excerpt0.15
- Describes handling uncertaintynot found in the excerpt0.03
- Reports measured numbersnot found in the excerpt0.02
- Is a runnable toolFound in material0.92
- Top-10 learning shortlist signalnot found in the excerpt0.12
- Is a replica of the modelnot found in the excerpt0.18
- Solves a common problem (0–2)score on a 0–2 scale1.58 / 2
- About JevFound in material0.94
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 1 Nov 2026.