testgap

AI-powered test gap finder. Scans your codebase with tree-sitter, identifies untested functions, and uses Claude to suggest what tests to write.

Install

From crates.io (once published):

cargo install testgap

From source:

git clone https://github.com/jadessoriano/testgap.git
cd testgap
cargo install --path crates/testgap-cli

Quick Demo

$ testgap analyze --no-ai

  testgap — Test Gap Analysis
  ────────────────────────────────────────
  Project:   /home/user/my-project
  Languages: rust, typescript
  Coverage:  42/68 functions (61.8%)
  AI:        disabled

  CRITICAL (3) ─────────────────────────
    process_payment src/billing.rs:47
      Public function with high complexity (12) and no test coverage
      Signature: pub fn process_payment(order: &Order, method: PaymentMethod) -> Result<Receipt>
      Complexity: 12

    validate_schema src/api/handlers.rs:112
      Public function with high complexity (8) and no test coverage
      Signature: pub fn validate_schema(input: &Value, schema: &Schema) -> Result<()>
      Complexity: 8

  WARNING (5) ──────────────────────────
    ...

  Summary: 3 critical, 5 warning, 18 info

Usage

# Analyze current directory
testgap analyze

# Static analysis only (no AI, no API key needed)
testgap analyze --no-ai

# Analyze a specific project
testgap analyze ./my-project

# JSON output for CI
testgap analyze --format json --fail-on-critical

# Markdown report
testgap analyze --format markdown

# Filter by language
testgap analyze --languages rust,typescript

# Only show critical gaps
testgap analyze --min-severity critical

# Reduce AI cost — only analyze critical gaps
testgap analyze --ai-severity critical

# Create a config file
testgap init

Best Practices

Best use cases:

CI gate for untested public APIs
Pre-release audit of test coverage
Onboarding into an unfamiliar codebase
Prioritizing which tests to write first

When NOT to use testgap:

Not a replacement for runtime coverage tools (lcov, tarpaulin, istanbul) — testgap uses static analysis
Skip generated or vendored code via exclude patterns in .testgap.toml
Don't chase 100% — Info-level gaps on private helpers are usually fine

Tips:

Start with --no-ai to get a fast baseline without API costs
Use --fail-on-critical in CI to catch regressions
Use --ai-severity critical to send only critical gaps to the AI, dramatically reducing API cost
Pipe JSON output to jq for custom filtering: testgap analyze --format json | jq '.gaps[] | select(.severity == "critical")'
Add a .testgap.toml config file early so the whole team shares the same settings

Supported Languages

Language	Extensions	Test Detection
Rust	`.rs`	`#[test]`, `#[cfg(test)]`, `test_` prefix
JavaScript	`.js`, `.jsx`, `.mjs`, `.cjs`	`test()`, `it()`, `describe()`, test dirs
TypeScript	`.ts`, `.tsx`, `.mts`, `.cts`	`test()`, `it()`, `describe()`, test dirs
Python	`.py`	`test_` prefix, test dirs
Go	`.go`	`Test` prefix, `*testing.T` parameter

How It Works

Scan — walks your project and classifies files as source or test
Extract — uses tree-sitter to parse functions from source files
Map — matches test functions to source functions by name, file convention, and body references
Detect — identifies untested functions and classifies severity:
- Critical: public + complex + untested
- Warning: public + untested
- Info: private + untested
Analyze (optional) — sends gaps to Claude API for risk assessment and test suggestions

Configuration

Create a .testgap.toml in your project root:

testgap init

See .testgap.toml.example for all options.

CI Integration

- name: Check test gaps
  run: testgap analyze --format json --fail-on-critical --no-ai

Exit codes:

0 — no critical gaps (or --fail-on-critical not set)
1 — critical gaps found (with --fail-on-critical)
2 — runtime error

Environment Variables

ANTHROPIC_API_KEY — required for AI analysis (use --no-ai to skip)

Architecture

See docs/ARCHITECTURE.md for the data flow diagram, module responsibilities, and key design decisions.

Contributing

See CONTRIBUTING.md for development setup, running tests, and how to add a new language.

License

MIT

testgap-core 0.2.0