Configuration

testpilot-ai supports multiple configuration methods with a clear priority order.

Priority Order

CLI flags > Config file > package.json "autotest" field > Auto-detected framework > Defaults

Config File

Create autotest.config.json or .autotestrc in your project root:

{
  "provider": "anthropic",
  "model": "claude-sonnet-4-20250514",
  "framework": "vitest",
  "edgeCases": true,
  "errorHandling": true,
  "maxTokens": 4096,
  "temperature": 0.2
}

Supported file names

• autotest.config.json
• .autotestrc
• .autotestrc.json

package.json

Add an autotest field to your package.json:

{
  "name": "my-project",
  "autotest": {
    "provider": "openai",
    "model": "gpt-4o",
    "framework": "vitest"
  }
}

Framework Auto-Detection

testpilot-ai automatically detects your test framework from package.json:

• If vitest is in devDependencies → uses Vitest
• If jest or @jest/core is in devDependencies → uses Jest
• If mocha is in devDependencies → uses Mocha
• If the test script contains vitest, jest, mocha, or node --test → uses that framework
• Default → Vitest

When --verify is used, testpilot also preflights that the chosen framework is actually installed in your project (node_modules/<framework> exists). If not, it fails fast with a clear FrameworkNotInstalledError instead of hanging on an npx install prompt.

TypeScript Path Aliases

When the analyzer follows imports for context, it resolves both relative paths and tsconfig path aliases. Given a tsconfig.json like:

{
  "compilerOptions": {
    "baseUrl": ".",
    "paths": {
      "@/*": ["src/*"],
      "~lib/*": ["src/lib/*"]
    }
  }
}

…imports such as import { foo } from '@/lib/foo' are followed back into the project so the LLM has the relevant type definitions in context. External node_modules imports are still skipped.

Environment Variables

API keys are resolved from standard environment variables:

Provider	Environment Variable
OpenAI	OPENAI_API_KEY
Anthropic	ANTHROPIC_API_KEY
Google	GOOGLE_API_KEY
Ollama	None needed (local)

All Configuration Options

Option	Type	Default	Description
provider	LLMProvider	"openai"	LLM provider — openai, anthropic, google, or ollama
model	string	Provider default	Model name
apiKey	string	From env var	API key
framework	"vitest" | "jest" | "mocha" | "node"	Auto-detected	Test framework
outDir	string	Same as source	Output directory
overwrite	boolean	false	Overwrite existing tests
edgeCases	boolean	true	Include edge case tests
errorHandling	boolean	true	Include error handling tests
instructions	string	—	Custom LLM instructions
maxTokens	number	4096	Max LLM tokens
temperature	number	0.2	LLM temperature