* feat: Implement Claude orchestration with session management
- Add CLAUDE_WEBHOOK_SECRET for webhook authentication
- Fix Docker volume mounting for Claude credentials
- Capture Claude's internal session ID from stream-json output
- Update entrypoint script to support OUTPUT_FORMAT=stream-json
- Fix environment variable naming (REPOSITORY -> REPO_FULL_NAME)
- Enable parallel session execution with proper authentication
- Successfully tested creating PRs via orchestrated sessions
This enables the webhook to create and manage Claude Code sessions that can:
- Clone repositories
- Create feature branches
- Implement code changes
- Commit and push changes
- Create pull requests
All while capturing Claude's internal session ID for potential resumption.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Update SessionManager tests for new implementation
- Update test to expect docker volume create instead of docker create
- Add unref() method to mock process objects to fix test environment error
- Update spawn expectations to match new docker run implementation
- Fix tests for both startSession and queueSession methods
Tests now pass in CI environment.
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix: Standardize Docker image naming and improve environment variable handling
- Standardize on 'claudecode:latest' image name across the codebase
- Update build script to use claudecode:latest instead of claude-code-runner:latest
- Fix health check to use CLAUDE_CONTAINER_IMAGE env var dynamically
- Improve environment variable handling for git configuration
- Pass BOT_EMAIL and BOT_USERNAME to containers
- Entrypoint scripts already use these with appropriate defaults
- Add comprehensive environment variables documentation
- Document all 90+ environment variables used in the project
- Identify hard-coded values that could be made configurable
- Update .env.example with missing variables
This ensures consistency in Docker image naming and allows proper git
configuration in containers using the configured bot identity.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Add BOT_EMAIL to docker-compose.yml
- Add BOT_EMAIL environment variable to docker-compose.yml
- Ensures git configuration in containers uses proper email address
- Complements the previous changes for environment variable handling
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Use BOT_USERNAME environment variable in prompt creation
- Fix undefined BOT_USERNAME reference in createPrompt function
- Change prompt to use actual bot username instead of hardcoded "Claude"
- Makes the prompt more accurate: "You are @MCPClaude" instead of "You are Claude"
This fixes the PR review functionality that was broken due to the
undefined variable reference.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* feat: Add verbose and stream-json output to Claude CLI for better debugging
- Add --verbose flag to see detailed Claude processing
- Add --output-format stream-json for structured output
- Helps diagnose issues with PR review and other operations
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Use colon syntax for allowedTools in PR review to match auto-tagging
- Change from space syntax Bash(gh *) to colon syntax Bash(gh:*)
- This matches the working syntax used in auto-tagging
- Should fix the permission issues preventing PR reviews from posting
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* feat: Add Claude Code timeout configuration for unattended mode
- Add BASH_DEFAULT_TIMEOUT_MS (10 minutes) and BASH_MAX_TIMEOUT_MS (20 minutes)
- Pass timeout environment variables to Claude container
- Document new timeout settings in .env.example and environment-variables.md
- Better defaults for webhook mode where builds/tests may take longer
These timeouts are more suitable for unattended PR reviews and other
operations that might involve long-running commands like builds or tests.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* feat: Add @botaccount review command for manual PR reviews (#131)
- Add detection for 'review' command in PR and issue comments
- Implement handleManualPRReview function with authorization checks
- Reuse existing PR review logic with manual-pr-review operation type
- Configure PR review tools with broad research access and controlled write access
- Support manual triggering of comprehensive PR reviews on demand
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* style: Apply pre-commit formatting changes
* test: Update test expectation for new operationType parameter
* fix: Improve PR detection for manual review command
- Add pull_request property to GitHubIssue interface for PR comments
- Handle both direct PR objects and issues with pull_request metadata
- Fix TypeScript compilation errors and linting issues
* fix: Improve pre-commit hook to fail on issues instead of auto-fixing
- Use format:check instead of format to detect issues without auto-fixing
- Use proper error handling with clear error messages
- Provide helpful instructions on how to fix issues
- Make commit behavior more predictable and transparent
* style: Fix whitespace formatting
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix: merge entrypoint scripts and fix auto-tagging tool permissions
- Merged duplicate claudecode-entrypoint.sh and claudecode-tagging-entrypoint.sh scripts
- Added dynamic tool selection based on OPERATION_TYPE environment variable
- Fixed auto-tagging permissions to include required Bash(gh:*) commands
- Removed 95% code duplication between entrypoint scripts
- Simplified claudeService.ts to use unified entrypoint
- Auto-tagging now uses: Read,GitHub,Bash(gh issue edit:*),Bash(gh issue view:*),Bash(gh label list:*)
- General operations continue to use full tool set
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: update Dockerfile to use unified entrypoint script
- Remove references to deleted claudecode-tagging-entrypoint.sh
- Update build process to use single unified entrypoint script
* fix: remove unnecessary async from promisify mock to fix lint error
* feat: add Husky pre-commit hooks with Prettier as primary formatter
- Added Husky for Git pre-commit hooks
- Configured eslint-config-prettier to avoid ESLint/Prettier conflicts
- Prettier handles all formatting, ESLint handles code quality only
- Pre-commit hooks: Prettier format, ESLint check, TypeScript check
- Updated documentation with pre-commit hook setup
- All code quality issues resolved
* feat: consolidate workflows and fix permission issues with clean Docker runners
- Replace 3 complex workflows with 2 lean ones (pull-request.yml, main.yml)
- Add Docker runner configuration for clean, isolated builds
- Remove file permission hacks - use ephemeral containers instead
- Split workload: GitHub-hosted for tests/security, self-hosted for Docker builds
- Add comprehensive pre-commit configuration for security
- Update documentation to be more pragmatic
- Fix credential file permissions and security audit
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: allow Husky prepare script to fail in production builds
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: update CI badge to reference new main.yml workflow
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Updates volume mounts and entrypoint scripts to properly mount Claude
authentication directory from ~/.claude-hub to /home/node/.claude in
containers, enabling proper credential access and token refresh capability.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
This comprehensive update adds support for Claude Max subscription authentication
and improves the overall authentication system with multiple methods:
🔐 Claude Authentication Enhancements:
- Add setup container method for Claude Max/20x subscription usage ($20-200/month)
- Create interactive authentication script (setup-claude-interactive.sh)
- Add authentication testing utility (test-claude-auth.sh)
- Support three authentication methods: Setup Container, API Key, AWS Bedrock
- Comprehensive authentication documentation
📁 Directory Configuration:
- Add CLAUDE_HUB_DIR environment variable (default: ~/.claude-hub)
- Update .gitignore to use .claude-hub/ instead of hardcoded paths
- Consistent environment variable usage across all scripts
🐙 GitHub Token Support:
- Add fine-grained GitHub token support (github_pat_) alongside classic tokens (ghp_)
- Update token validation in claudeService and githubService
- Enhanced token detection and authentication flow
📖 Documentation & Guides:
- Add complete Claude Authentication Guide with all three methods
- Add Setup Container Deep Dive documentation
- Update CLAUDE.md with quick start authentication section
- Comprehensive cost comparison and use case recommendations
🐳 Container & Docker Improvements:
- Update Dockerfile.claudecode with proper entrypoint script copying
- Add Dockerfile.claude-setup for interactive authentication
- Update docker-compose.yml with new port (3003) and environment variables
- Enhanced container volume mounting for authentication
🔧 Infrastructure Updates:
- Add TRUST_PROXY configuration for reverse proxy environments
- Update port configuration from 3002 to 3003
- Enhanced environment variable documentation in .env.example
- Debug utilities for troubleshooting authentication issues
This update enables Claude Max subscribers to use their existing subscriptions
for automation, potentially saving thousands in API costs while maintaining
full production capabilities.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit reorganizes all scripts in the repository into a more structured directory layout for better maintainability:
- Categorizes scripts by functionality (setup, build, aws, runtime, security, utils)
- Organizes test scripts into logical categories
- Consolidates redundant scripts with unified interfaces
- Adds backward compatibility wrappers
- Adds detailed SCRIPTS.md documentation
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
