# Test Generator > **Purpose:** Analyze source code and generate comprehensive unit tests following the project's testing conventions. --- ## Invocation ``` /test-gen [--framework vitest|jest|pytest|go] ``` --- ## Process ### Step 1: Analyze the Target Read the source file and identify: - All exported functions/classes/methods - Input types and return types - Edge cases from conditional branches - Dependencies that need mocking - Existing test file (if any) to avoid duplicates ### Step 2: Determine Testing Framework Auto-detect from project config: - `vitest.config.*` → Vitest - `jest.config.*` or `package.json[jest]` → Jest - `pytest.ini` or `pyproject.toml[tool.pytest]` → pytest - `*_test.go` files → Go testing ### Step 3: Generate Test Cases For each function, generate: 1. **Happy path** — typical valid input → expected output 2. **Edge cases** — empty input, zero, null/undefined, boundary values 3. **Error paths** — invalid input → expected error/exception 4. **Type variations** — different valid types if function accepts union types ### Step 4: Structure the Test File ```typescript import { describe, it, expect, vi } from 'vitest' import { targetFunction } from '../path/to/source' describe('targetFunction', () => { it('returns expected result for valid input', () => { expect(targetFunction('valid')).toBe('expected') }) it('throws on null input', () => { expect(() => targetFunction(null)).toThrow() }) it('handles empty string', () => { expect(targetFunction('')).toBe('') }) }) ``` --- ## Mocking Strategy - **External APIs**: Always mock with `vi.mock()` or `jest.mock()` - **Database calls**: Mock the client/ORM, not the database - **File system**: Mock `fs` operations, never touch real files - **Time**: Use `vi.useFakeTimers()` for time-dependent logic - **Environment variables**: Set via `vi.stubEnv()` or test setup --- ## Rules - Match the project's existing test style (describe/it vs test, assertion style) - Place test files adjacent to source (`*.test.ts`) or in `tests/` mirror - Never generate snapshot tests unless explicitly asked - Aim for branch coverage, not just line coverage - Each test should be independent — no shared mutable state - Test names should describe behavior, not implementation ## Playground

Input: Function spec

Function: parseAmount(str)
Language: TypeScript / Vitest

Parses a currency string like
"$1,234.56" → 1234.56 (number).
Throws if input is not a valid
currency string.

Output: Test suite

describe('parseAmount', () => {
  it('parses USD with commas', () =>
    expect(parseAmount('$1,234.56'))
      .toBe(1234.56))

  it('parses zero', () =>
    expect(parseAmount('$0.00')).toBe(0))

  it('throws on empty string', () =>
    expect(() => parseAmount(''))
      .toThrow())

  it('throws on plain text', () =>
    expect(() => parseAmount('hello'))
      .toThrow())
})