- Fixed dark mode visibility by setting proper text and tag luminosities in docs/static/v2_elite.css.
- Eliminated triple 'Explore Related' blocks by moving the injection outside recursion in src/v2_optimizer.py.
- Fixed 'Technical Deep-Dive' rendering by correcting Markdown indentation for collapsible blocks.
- Explicitly set response_format='text' to prevent JSON parsing failure loops.
- Added safety delay (1s) to respect global RPM quota.
- Tagged with role='PR-Guardian' for improved telemetry.
Provides a safe, low-cost way to publish the current cached state of the V2 portal without running the full AI pipeline. Guaranteed read-only cache access.
- Removed conflictive 'workflows: write' permission.
- Added 'Safety Reset for Workflow Files' step to prevent GHA from trying to self-modify configuration files during the PR sync.
- This ensures the PR will only contain technical metadata and V2 documentation, avoiding permission rejections.
- Added PYTHONUNBUFFERED: '1' to Agentic Curation and Backup Curation workflows.
- This ensures consistent real-time logging across the entire ecosystem, complementing the previous fixes in V2 Builder and the Logger core.
- Implemented manual sys.stdout.flush() in log_event.
- Enabled flush=True for all print statements in the logger.
- Set PYTHONUNBUFFERED: '1' in the V2 Builder workflow environment.
- These changes ensure immediate log delivery and reduce GHA agent stuttering.
- Clarified that inventory data is saved without waiting for workflow completion.
- Specified save intervals for both Metadata (500 links) and AI (20 batches) phases.
- Added auto-save every 500 GitHub API fetches to prevent data loss on early cancellation.
- This ensures that subsequent runs skip already enriched repositories.
- Added detailed trigger strategies for Curation, Cleaner, and Backup workflows.
- Defined default values and 'Fast/Standard' configurations for all manual triggers.
- Synchronized technical variables with UI labels for maintainer clarity.
- Added detailed explanation of manual workflow flags in README.md.
- Defined 'Fast/Standard' configuration for optimal cache utilization.
- Updated Mandate 47 in GEMINI.md to ensure AI agents respect manual overrides.
- Synchronized Table of Contents.
- Set cancel-in-progress to false for V2 Builder.
- This ensures that if a workflow is triggered multiple times (e.g., via push and manual dispatch), subsequent runs will queue cleanly instead of attempting to cancel the running one, which was causing duplicate process ghosting and double API quota consumption.
- Reduced BATCH_SIZE_FAST to 50 to prevent AI request timeouts.
- Hardened call_gemini_with_retry to strictly use Flash/Lite models when prefer_flash=True.
- Ensured newest model versions are prioritized within the Flash tier.
- Added periodic save (every 20 batches) to prevent data loss in long runs.
- Verified Flash-First settings (Batch: 100, Delay: 2s, prefer_flash=True).
- Transitioned V2 Optimizer to high-density batching (100 items/batch).
- Reconfigured Health Cleaner to use Flash/Lite for high-volume URL rescue.
- Codified mandates 48-50 in GEMINI.md for AI governance.
- Added Agentic Model Selection Matrix to README.md.
- Improved error handling to ensure Rate-Limit events trigger Circuit Breakers.
