bun install requires network access which Nix sandbox blocks.
CI now validates flake structure and evaluates expressions only.
Full builds work locally with: nix build --option sandbox false
- Use DeterminateSystems/nix-installer-action for Nix installation
- Use DeterminateSystems/magic-nix-cache-action for caching (free, no setup)
- Update documentation to remove Cachix references
- Add nix branch to CI triggers
- Add 1-second delays between release creations to ensure distinct timestamps
- Prepend GitHub original publication date to release notes
- Improve logging to show chronological processing order
- Addresses Gitea API limitation where created_unix is always set to current time
Fixes#129
Fixes#141
The repository metadata field was missing from the database schema, which
caused the metadata sync state (issues, PRs, releases, etc.) to not persist.
This resulted in duplicate issues being created every time a repository was
synced because the system couldn't track what had already been mirrored.
Changes:
- Added metadata text field to repositories table in schema
- Added metadata field to repositorySchema Zod validation
- Generated database migration 0008_serious_thena.sql
Root cause analysis:
1. Code tried to read/write repository.metadata to track mirrored components
2. The metadata field didn't exist in the database schema
3. On sync, metadataState.components.issues was always false
4. This triggered re-mirroring of all issues, creating duplicates
The fix ensures metadata state persists between mirrors and syncs, preventing
duplicate metadata (issues, PRs, releases) from being created in Gitea.
Updated development documentation to reflect current project structure
and simplified setup process.
Changes:
- DEVELOPMENT_WORKFLOW.md: Updated repository URL, simplified setup steps,
improved project structure documentation, and clarified command descriptions
- README.md: Reorganized as a concise index of available guides, removed
redundant content now covered in main README and in-app help
- SHUTDOWN_PROCESS.md: Removed (content consolidated into GRACEFUL_SHUTDOWN.md)
These updates make the documentation more accurate and easier to navigate
for new contributors.
Updated the vendor lock-in prevention article with more accurate UI navigation
paths and technical details to match current application features.
Changes:
- Added missing title heading
- Updated navigation instructions to reference actual UI paths (Configuration → Connections, Content & Data, Automation)
- Improved technical accuracy of sync interval recommendations
- Added FAQ section with practical questions about auto-discovery, sync intervals, and cutover scripting
- Clarified activity log and API endpoint usage for monitoring
These changes ensure the article reflects the current application UI and features.
Updated Helm chart configuration to use the correct application port (4321)
instead of the incorrect default (8080). This aligns the Helm deployment
with the actual application configuration.
Changes:
- helm/gitea-mirror/values.yaml: Updated service.port and deployment.port to 4321
- helm/gitea-mirror/README.md: Updated all port references in documentation
- www/src/pages/use-cases/deploy-with-helm-chart.mdx: Fixed article to reflect
correct port, added GitHub links to Helm chart, and improved installation instructions
The application runs on port 4321 as defined in:
- Dockerfile (ENV PORT=4321, EXPOSE 4321)
- docker-compose.yml (4321:4321 mapping)
- .env.example (PORT=4321)
Tested with local Kubernetes cluster and confirmed the application is accessible
on port 4321.
Since githubConfig is stored as JSON in the database (not individual columns),
no database migration is needed. However, we need to handle reading old configs
that still use the 'skipStarredIssues' field name.
Changes:
- Added skipStarredIssues as optional field in Zod schema (marked deprecated)
- Updated config mapper to check both starredCodeOnly and skipStarredIssues
- Old configs will continue to work seamlessly
- New configs will use starredCodeOnly field name
This ensures zero-downtime upgrades for existing installations.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The previous name 'skipStarredIssues' was misleading as it now skips ALL
metadata (not just issues) for starred repositories. The new name
'starredCodeOnly' better reflects the actual behavior - mirroring only
source code for starred repos.
Changes:
- Renamed skipStarredIssues → starredCodeOnly in all files
- Updated UI label from "Don't fetch issues" to "Code-only mode"
- Updated description to clarify it skips ALL metadata types:
issues, PRs, labels, milestones, wiki, and releases
- Updated database schema, types, config mapper, and all components
- Updated Helm charts, CI configs, and documentation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Extended the skipStarredIssues flag to also skip releases for starred repos
when "code-only" mode is enabled. Previously, releases were still being
mirrored even when lightweight mode was selected.
Now starred repos with "code-only" mode will skip:
- Issues ✓
- Pull requests ✓
- Labels ✓
- Milestones ✓
- Wiki ✓
- Releases ✓ (this fix)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Previously, the skipStarredIssues flag (Lightweight mode for starred repos)
only applied to issues. This caused starred repos to mirror all metadata
(pull requests, labels, milestones, wiki) even when "code-only" was selected.
Fixed by applying the skipStarredIssues check to:
- Pull requests mirroring
- Labels mirroring
- Milestones mirroring
- Wiki mirroring (in migration payload)
Now starred repos with "code-only" mode truly mirror only source code,
skipping all metadata as intended.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added docker-entrypoint.sh to trigger paths
- Added drizzle/** for database migrations
- Added scripts/** for database management scripts
- Added src/** for source code changes
This ensures Docker images are rebuilt when critical runtime
files change, not just package dependencies.
## Problem
Forgejo 12.0+ rejects migration API calls with credentials embedded in URLs,
causing HTTP 422 errors when mirroring private GitHub repositories.
## Root Cause
Breaking security change in Forgejo 12.0 (July 2025) enforces credential
separation to prevent accidental exposure in logs/errors. Previous versions
(Forgejo 11.x, Gitea 1.x) accepted embedded credentials.
## Solution
- Use separate `auth_username` and `auth_token` fields instead of embedding
credentials in clone URLs
- Set `auth_username` to "oauth2" for GitHub token authentication
- Pass GitHub token via `auth_token` field
## Changes
- src/lib/gitea.ts:
- mirrorGithubRepoToGitea(): Use separate auth fields for private repos
- mirrorGitHubRepoToGiteaOrg(): Use separate auth fields for private repos
- .github/workflows/docker-build.yml:
- Enable PR image building and pushing to GHCR
- Tag PR images as pr-<number> for easy testing
- Add automated PR comment with image details and testing instructions
- Separate load step for security scanning
## Backward Compatibility
✅ Works with Forgejo 12.0+
✅ Works with Forgejo 11.x and earlier
✅ Works with Gitea 1.x
## Testing
Public repos: ✅ Working (no auth needed)
Private repos: ✅ Fixed (separate auth fields)
Fixes#102
- Prevent Automation UI from overriding schedule:
- mapDbScheduleToUi now parses intervals robustly (cron/duration/seconds) via parseInterval
- mapUiScheduleToDb merges with existing config and stores interval as seconds (no lossy cron conversion)
- /api/config passes existing scheduleConfig to preserve ENV-sourced values
- schedule-sync endpoint uses parseInterval for nextRun calculation
- Add AUTO_MIRROR_REPOS support and scheduled auto-mirror phase:
- scheduleConfig schema includes autoImport and autoMirror
- env-config-loader reads AUTO_MIRROR_REPOS and carries through to DB
- scheduler auto-mirrors imported/pending/failed repos when autoMirror is enabled before regular sync
- docker-compose and ENV docs updated with AUTO_MIRROR_REPOS
- Tests pass and build succeeds
- Implemented comprehensive GitHub API rate limit handling:
- Integrated @octokit/plugin-throttling for automatic retry with exponential backoff
- Added RateLimitManager service to track and enforce rate limits
- Store rate limit status in database for persistence across restarts
- Automatic pause and resume when limits are exceeded
- Proper user identification for 5000 req/hr authenticated limit (vs 60 unauthenticated)
- Improved rate limit UI/UX:
- Removed intrusive rate limit card from dashboard
- Toast notifications only at critical thresholds (80% and 100% usage)
- All rate limit events logged for debugging
- Optimized for GitHub's API constraints:
- Reduced default batch size from 10 to 5 repositories
- Added documentation about GitHub's 100 concurrent request limit
- Better handling of repositories with many issues/PRs
- Add missing database fields (language, description, mirroredLocation, destinationOrg) to repository operations
- Add missing organization fields (publicRepositoryCount, privateRepositoryCount, forkRepositoryCount) to schema
- Update GitRepo interface to include all required database fields
- Fix GitHub data fetching functions to map all fields correctly
- Update all sync endpoints (main, repository, organization, scheduler) to handle new fields
This fixes the "SQLite query expected X values, received Y" error when importing
large numbers (4.6k+) of starred repositories by ensuring all database fields
are properly mapped from GitHub API responses through to database insertion.
Arunavo Ray