metadata
license: apache-2.0
AssertAI
A fine-tuned SLM to generate deterministic Python unit test plans in strict JSON. It's designed to act as a test-case planner, rather than a full code generator.
Base model: Llama3.2.
What it does
Given a function signature + docstring/spec, Assert-AI outputs:
- a compact list of 2–5 high-signal unit tests
- each test includes args, kwargs, and either an expected value (expect) or expected exception (error)
Output format
Assert-AI outputs only this JSON object (no extra keys, no markdown):
{
"fn": "safe_divide",
"tests": [
{ "name": "divides_when_nonzero", "args": [19, -3], "kwargs": {"default": 0.0}, "expect": -6.333333333333333 },
{ "name": "returns_default_on_zero", "args": [19, 0.0], "kwargs": {"default": 1.5}, "expect": 1.5 }
]
}
Example User Input
Function spec:
def clamp(n: int, lo: int, hi: int) -> int:
\"\"\"Return n bounded between lo and hi inclusive. Precondition: lo <= hi.\"\"\"
Author
Author: Sai Teja Erukude
Role: Developer & Maintainer