> ## Documentation Index > Fetch the complete documentation index at: https://www.greptile.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Developer Quick Reference > Developer guide to Greptile: learn how to trigger reviews, ask follow-up questions, train the AI with feedback, and customize review behavior for your workflow. ## Overview **What does Greptile do?** * AI code review agent that analyzes your pull requests * Reviews code with full context of your entire codebase * Provides three types of feedback: summary, inline comments, and suggested fixes * Learns from your team's preferences and feedback over time **How is it different from other tools?** * Understands your specific codebase patterns and architecture * References related code across your entire repository when making suggestions * Learns what your team cares about through reactions and feedback * Provides conversational interaction - you can ask follow-up questions ## Capabilities **What types of issues does it catch?** * Logic errors and potential bugs in your code * Security vulnerabilities like unvalidated inputs or SQL injection risks * Performance issues and inefficient patterns * Code style and consistency violations * Missing error handling or edge cases **What feedback does it provide?** * Summary of changes in your PR * Inline comments on specific lines with issues * Suggested code fixes you can click-to-accept * References to similar code in your codebase * Explanations of why something might be problematic **Can I ask it questions?** * Reply to any comment with `@greptileai ` and Greptile will respond * Ask for clarification on why something is flagged * Request alternative approaches to solve problems * Get explanations of unfamiliar code patterns ## When does it run? **When does Greptile automatically review?** * Your team configures trigger rules (labels, authors, branches, keywords) * Usually runs when PRs are first opened * May run on every commit push depending on team settings * Skips PRs that match ignore patterns (certain files, branches, etc.) **How do I manually trigger a review?** * Comment `@greptileai` on any PR to force an immediate review * Works on old PRs, even from before Greptile was enabled * Can re-trigger after making changes to get updated feedback * No rate limits - use as often as needed **Why isn't it reviewing my PR?** * Check if your PR matches your team's trigger rules * Your changes might be in excluded files or directories * Try manual trigger with `@greptileai` comment * Ask your team admin to check dashboard settings * Ensure your PR's repository is enabled and indexed in Greptile settings ## How do I give it more context? **What is custom context?** * Files your team uploads to help Greptile understand your standards * Style guides, architecture docs, security requirements, API standards * Greptile references these when reviewing your code * Ask your team admin what context files have been uploaded **How does Greptile use uploaded files?** * Searches context files when making suggestions * Enforces standards documented in uploaded guides * References architecture patterns from your docs * Ensures consistency with your team's documented practices **What should my team upload?** * Coding style guides and conventions * Architecture decision records and system design docs * Security policies and compliance requirements * API design standards and patterns * Testing strategies and requirements ## About memory and learning **How does Greptile learn?** * From thumbs up/down reactions on its comments * By reading other developers' comments on PRs * By analyzing which suggestions get implemented vs ignored * From team discussions about code patterns and preferences **What does it remember about my team?** * Code patterns you prefer (async/await vs Promises) * Architecture preferences (thin controllers, service layers) * Security standards (input validation requirements) * Style preferences (what matters vs what doesn't) **How do I train it effectively?** * Use thumbs up on helpful comments * Use thumbs down on unhelpful or wrong suggestions * React consistently - regular feedback helps it learn faster * Explain in comments why you're dismissing suggestions * Tag the bot in a thread and ask it to create a rule **How long does learning take?** * 2-3 weeks of consistent feedback for noticeable improvement * Early weeks may have more noise as it learns preferences * Gets better at filtering suggestions to what your team values * Becomes more accurate at understanding your codebase context ## About codebase context **How does Greptile understand my codebase?** * Builds a detailed graph of functions, classes, files, and their relationships * Analyzes dependencies and how different parts of code connect * Understands data flow and usage patterns across your repository * Maps out your architecture and common patterns **What context does it use when reviewing?** * Code affected by your changes and its dependencies * Related code that follows similar patterns * Files that import or are imported by your changes * Similar functions or components elsewhere in the codebase **Does it see my entire repository?** * Yes, Greptile analyzes your full codebase to build context * Uses this understanding to make more accurate suggestions * Can reference patterns from anywhere in your repository * Understands how your changes fit into the bigger picture ## Chatting with the agent with follow-up questions **How do I ask follow-up questions?** * Reply to any Greptile comment with `@greptileai ` * Ask on any line of your PR, even without an existing comment * Questions can be about the specific code or general patterns **What kinds of questions can I ask?** * "Can you explain why this is a security issue?" * "What's a better way to handle this error case?" * "How should I test this function?" * "Are there similar patterns elsewhere in our codebase?" * "What are the performance implications of this approach?" **What responses can I expect?** * Explanations tailored to your specific code and codebase * Alternative approaches with code examples * References to similar patterns in your repository * Context about why certain practices are recommended * Suggestions for testing, error handling, or optimization **Tips for effective conversations:** * Be specific in your questions * Reference the exact code or pattern you're asking about * Ask about alternatives if you disagree with a suggestion * Use conversations to learn about best practices for your codebase **Greptile is making too many comments, what should I do?** * You can go to `Settings` in the dashboard and use the `Severity Threshold` slider * You can turn off `style` or `syntax` comments in the same settings window