Multi-Repository Setup
Scale Simili Bot across your entire organization with centralized configuration.Overview
Single Repository:Setup
1. Configure All Repositories
2. Same Secrets Across Repos
All repositories use same Qdrant and Gemini instances:3. Shared GitHub Action
Deploy same workflow to all repositories:Benefits
✅ Unified Knowledge Base- Semantic search across all repos
- Find related issues across projects
- See organization-wide patterns
- Same quality standards
- Unified labeling
- Consistent triage
- One configuration file
- Update rules once, applies everywhere
- Centralized policy
- Issues go to correct repos automatically
- Organization-wide organization
Configuration Inheritance
Share configuration across repositories: Organization Base:Cross-Repository Search
Find issues across all repositories:- Current repository
- All other repositories in config
- Entire organization
- Duplicate issues across repos
- Common problems
- Related discussions
Coordination Example
Scenario: User opens issue “Login broken” in frontend repo Simili Bot Actions:- Searches backend, frontend, mobile repos
- Finds similar issues:
- frontend #234: “Sign in button not working” (95%)
- backend #45: “Auth endpoint down” (88%)
- mobile #67: “App can’t authenticate” (85%)
- Posts comprehensive analysis
- Optionally routes to central issue tracker
Organization Structure
Team-Based
Domain-Based
Best Practices
1. Use Shared Organization Secrets
- Settings → Secrets and variables → Actions
- Grant access to all repositories
2. Version Control Configuration
3. Index All Issues
Bulk index all historical issues once:Troubleshooting
Cross-Repo Search Not Working
Check configuration:Too Many Suggestions
Reduce results:Configuration Not Found
Verifysimili.yaml exists in all repositories.
Monitoring
View issues across organization:- Go to organization
- Use GitHub Issues global search
- See all issues with Simili Bot comments
- Monitor routing decisions
Performance at Scale
Organization with 50+ repositories:- Use multi-repo configuration
- One Qdrant collection for all
- ~1MB per 10,000 issues in vector DB
- Cross-repo search: 0.5-2 seconds