How It Works
GitAuto uses 50+ techniques across 7 categories to generate high-quality unit tests. Each technique solves a specific failure mode we encountered in production - from hallucinated imports to broken diffs to runaway CI costs.
Context Enrichment
How GitAuto gathers the right context before generating tests - reading full files, detecting naming conventions, cleaning CI logs, and loading existing test examples.
Output Auto-Correction
Post-processing steps that fix common model output issues - repairing diff hunks, correcting tool names, sorting imports, and preserving line endings.
Quality Verification
Checks that run before committing generated code - formatting, linting, type checking, test execution, coverage enforcement, and dead code removal.
Safety Guardrails
Protections that prevent unintended side effects - file edit restrictions, race condition prevention, bot loop detection, and webhook deduplication.
Token/Cost Management
Optimizations that reduce token usage and CI costs - trimming context windows, deduplicating CI logs, removing stale file reads, and skipping intermediate CI runs.
Resilience & Recovery
How GitAuto handles failures gracefully - model fallback, overload retry with exponential backoff, forced verification, and dynamic edit permissions.
Hallucination Prevention
Techniques that keep generated code grounded in reality - web search for fact checking, URL fetching, anti-hallucination prompts, and review response guardrails.
Need Help?
Have questions or suggestions? We're here to help you get the most out of GitAuto.
Contact us with your questions or feedback!