Metadata-Version: 2.5
Name: evalstand
Version: 0.0.0.dev0
Summary: A local-first LLM evaluation tool for Python: write evals, run them like tests, watch results stream in.
License: MIT License
        
        Copyright (c) 2026 evalstand contributors
        
        Permission is hereby granted, free of charge, to any person obtaining a copy
        of this software and associated documentation files (the "Software"), to deal
        in the Software without restriction, including without limitation the rights
        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
        copies of the Software, and to permit persons to whom the Software is
        furnished to do so, subject to the following conditions:
        
        The above copyright notice and this permission notice shall be included in all
        copies or substantial portions of the Software.
        
        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
        SOFTWARE.
License-File: LICENSE
Keywords: ai,eval,evaluation,llm,pytest,testing
Classifier: Development Status :: 2 - Pre-Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Software Development :: Testing
Requires-Python: >=3.11
Requires-Dist: litellm>=1.55
Requires-Dist: pydantic>=2.10
Requires-Dist: pytest>=8.3
Requires-Dist: rapidfuzz>=3.11
Requires-Dist: rich>=13.9
Requires-Dist: tenacity>=9.0
Requires-Dist: textual>=1.0
Requires-Dist: typer>=0.15
Requires-Dist: watchfiles>=1.0
Provides-Extra: examples
Requires-Dist: faker>=33.1; extra == 'examples'
Requires-Dist: pypdfium2>=4.30; extra == 'examples'
Requires-Dist: reportlab>=4.2; extra == 'examples'
Description-Content-Type: text/markdown

# evalstand

> **Status: in development.** Phase 0 of 7. This release is a **name
> reservation placeholder** and contains no working code. The example below
> shows the intended API, which is not implemented yet.

Evaluating an LLM application should feel like running a test suite.

`evalstand` is a local-first LLM evaluation tool for Python. You write an eval
file, run a watch command, and results stream into a live terminal UI — scores,
nested call traces, token counts, latency, and cost. Everything runs on your
machine and persists to a local SQLite database, so you can compare a run
against the one before it.

```python
from evalstand import Case, evaluate
from evalstand.scorers import exact, levenshtein


def load_cases() -> list[Case]:
    return [
        Case(id="q1", input="What is the capital of France?", expected="Paris"),
        Case(id="q2", input="What is 2 + 2?", expected="4"),
    ]


async def answer(question: str) -> str:
    resp = await llm.acall("gpt-4o-mini", [{"role": "user", "content": question}])
    return resp.text


evaluate(
    name="basic-qa",
    cases=load_cases,
    task=answer,
    scorers=[exact, levenshtein],
)
```

Save that as `qa_eval.py` and run it either way:

```bash
evalstand run qa_eval.py     # live TUI, watch mode, traces
pytest qa_eval.py            # plain test runner, CI-friendly
```

## Why

Existing Python options are either heavyweight platforms that push you toward a
hosted service, or bare metric libraries with no runner, no persistence, and no
live feedback loop. `evalstand` is the middle: a real runner with a real UI that
stays on your machine.

## Planned capabilities

See [PLAN.md](PLAN.md) for the full build plan and the capability checklist that
defines v1.

## Limitations

Stated up front, and kept accurate as the project grows:

- Score differences between runs are reported as plain deltas. There is **no
  statistical significance testing** in v1, so a delta is not evidence of a real
  regression or improvement.
- LLM-as-judge scorers are **unvalidated** — they have not been calibrated
  against human labels.

## Licence

MIT. See [LICENSE](LICENSE).

---

<sub>Inspired by [evalite](https://github.com/mattpocock/evalite) (MIT), which
showed that local LLM evals could feel like running tests. `evalstand` is an
independent Python implementation.</sub>
