How qodo-ai/pr-agent works
PR-Agent is an AI-powered pull-request assistant. Users invoke commands locally or through platform integrations; the agent dispatches a command to a tool, which gathers PR context, uses an AI backend, and publishes a result through a Git provider. The graph includes documented tool capabilities and shared processing layers. Some subsystem paths are directories because their detailed implementations were not sampled; relationships are limited to README statements and SOURCE INDEX references.
Main entry points: pr_reviewer.py Review · pr_code_suggestions.py Code suggestions · pr_agent.py Command agent · pr_description.py PR description
Components 17
Entry points
- CLI
pr_agent/cli.py - Platform hooks
pr_agent/servers
Command workflow
- Command agent
pr_agent/agent/pr_agent.py - Review
pr_agent/tools/pr_reviewer.py - PR description
pr_agent/tools/pr_description.py - Code suggestions
pr_agent/tools/pr_code_suggestions.py - Other PR tools
pr_agent/tools
Review processing
- PR context
pr_agent/algo/pr_processing.py - Repository context
pr_agent/algo/repo_context.py - Diff processing
pr_agent/algo/git_patch_processing.py - Prompt budgeting
pr_agent/algo/token_budget.py
External integrations
- Provider interface
pr_agent/git_providers/git_provider.py - Platform providers
pr_agent/git_providers - AI handlers
pr_agent/algo/ai_handlers
Other components
- Developer
- LLM service
- Pull request platform
Connections 29
- Developer → CLI: runs command
README.md - Developer → Platform hooks: invokes commands
README.md - Platform hooks → Command agent: dispatches requests
pr_agent/servers/gitlab_webhook.py - CLI → Command agent: dispatches commands
pr_agent/cli.py - Command agent → Review: dispatches
pr_agent/agent/pr_agent.py - Command agent → PR description: dispatches
pr_agent/agent/pr_agent.py - Command agent → Code suggestions: dispatches
pr_agent/agent/pr_agent.py - Command agent → Other PR tools: dispatches
pr_agent/agent/pr_agent.py - Review → PR context: uses
pr_agent/tools/pr_reviewer.py - PR description → PR context: uses
pr_agent/tools/pr_description.py - Code suggestions → PR context: uses
pr_agent/tools/pr_code_suggestions.py - Review → Repository context: uses
pr_agent/tools/pr_reviewer.py - PR description → Repository context: uses
pr_agent/tools/pr_description.py - Code suggestions → Repository context: uses
pr_agent/tools/pr_code_suggestions.py - Review → Diff processing: uses
pr_agent/tools/pr_reviewer.py - Code suggestions → Diff processing: uses
pr_agent/tools/pr_code_suggestions.py - Review → Prompt budgeting: uses
pr_agent/tools/pr_reviewer.py - PR description → Prompt budgeting: uses
pr_agent/tools/pr_description.py - Code suggestions → Prompt budgeting: uses
pr_agent/tools/pr_code_suggestions.py - Review → AI handlers: requests completion
pr_agent/tools/pr_reviewer.py - PR description → AI handlers: requests completion
pr_agent/tools/pr_description.py - Code suggestions → AI handlers: requests completion
pr_agent/tools/pr_code_suggestions.py - AI handlers → LLM service: sends prompts
- Command agent → Provider interface: gets provider
pr_agent/agent/pr_agent.py - Provider interface → Platform providers: loads implementations
pr_agent/git_providers/__init__.py - Review → Provider interface: reads and publishes
pr_agent/tools/pr_reviewer.py - PR description → Provider interface: reads and publishes
pr_agent/tools/pr_description.py - Code suggestions → Provider interface: reads and publishes
pr_agent/tools/pr_code_suggestions.py - Platform providers → Pull request platform: accesses PR platform