quelltest 0.4.4

Your docstrings say what your code should do. Quell proves it.

quelltest

Your docstrings say what your code should do. Quell proves it.

Quell reads specifications that already exist in your codebase — docstrings, Pydantic models, bug reports — extracts every testable requirement, checks which ones have no test, generates a test for each gap, proves the test actually catches violations, then writes it to disk. Every test is verified before it touches your files.

Why Quell is different

Tool	Spec source	Verified?
Qodo Gen	reads implementation	❌
GitHub Copilot	chat prompt	❌
Hypothesis	you write it manually	❌
Quell	reads existing specs	✅

The critical insight: Qodo reads your implementation and generates tests for what your code does. If your code has a bug, Qodo generates tests that bless it. Quell reads your specification (docstring says "must raise ValueError") and generates a test that proves the requirement — catching the bug.

How it works

docstrings + Pydantic models + bug reports
         ↓
   list[Requirement]
         ↓
   Coverage checker (AST scan — no execution)
         ↓
   Rule engine → verified test
         ↓
   Verification: PASS on correct code + FAIL on violated code
         ↓
   libcst injection into test file

Installation

pip install quelltest

Requires Python 3.11+. The CLI command is quell.

Quick start

# Scan your specs, find gaps
quell check src/

# Generate + verify + write tests for all gaps
quell check src/ --fix

# Reproduce a bug from a description
quell reproduce "payment accepts zero amount"

# Show confidence score for a file
quell prove src/payments.py

# Project-wide Quell Score
quell score --badge

Example

Given this existing code:

def process_payment(request: PaymentRequest) -> dict:
    """
    Process a payment transaction.

    Args:
        request: Amount must be greater than 0. Currency must be one of: USD, EUR, GBP.

    Returns:
        dict with transaction_id, status, amount.

    Raises:
        ValueError: If amount is zero or negative.
    """

And this Pydantic model:

class PaymentRequest(BaseModel):
    amount: float = Field(gt=0)
    currency: Literal["USD", "EUR", "GBP"]

Running quell check src/payments.py finds 5 requirements with no tests and generates a verified test for each one — before touching your files.

Configuration

quell init   # adds [tool.quell] to pyproject.toml

[tool.quell]
llm_provider = "anthropic"          # "anthropic" | "openai" | "ollama"
llm_model    = "claude-sonnet-4-5"
enable_docstring = true
enable_types     = true
enable_mutations = false            # mutmut optional
auto_write       = false

Set your LLM API key (only needed for complex/unstructured specs):

export ANTHROPIC_API_KEY=sk-ant-...
# or
export OPENAI_API_KEY=sk-...

For fully local/offline setup, use Ollama:

# In pyproject.toml: llm_provider = "ollama"
# ollama pull codellama

Python SDK

from quell import Quell

q = Quell()

# Find requirement gaps
result = q.check("src/")
print(f"Score: {result.score:.0%} | Gaps: {len(result.uncovered)}")

# Reproduce a bug
q.reproduce("payment accepts zero amount silently")

# Project score
score = q.score()
print(f"Project: {score.percentage}%")

Project structure

quell/
├── cli.py              # Typer CLI: check, reproduce, prove, score, ci, init
├── sdk.py              # Python API: Quell class
├── spec/               # Spec readers (docstring, type, bug, mutation)
├── core/
│   ├── models.py       # Requirement, ConstraintKind, VerificationResult
│   ├── verifier.py     # THE MOAT — proves every test catches violations
│   └── writer.py       # libcst injection, backup/restore
├── coverage/           # AST-based coverage checker
├── synthesis/          # rule_engine.py + llm_engine.py
├── score/              # Quell Score calculator + SVG badge
└── llm/                # Anthropic / OpenAI / Ollama providers

Development

git clone https://github.com/shashank7109/quelltest_lib.git
cd quelltest_lib
uv sync --dev

uv run pytest tests/ -v
uv run ruff check . --fix
uv run mypy quell/

Related

• Docs
• quell_frontend — Next.js website

License

MIT — see LICENSE