firefrost-gaming/firefrost-operations-manual
3
0
Fork 0
Code Issues 146 Pull Requests Packages Projects Releases Wiki Activity

docs: elevate Dify RAG tuning and n8n rebuild to active priorities

Browse Source 
Decision made March 4, 2026 by Michael:
- Dify RAG tuning: actively queued (Top K 10-15, score threshold 0.3)
- n8n: rebuild workflows from scratch, not restore old ones

Task #34 updated:
- Renamed from 'Factory Reset' to 'Rebuild - Workflows From Scratch'
- Status changed from DEFERRED to ACTIVE
- Procedure updated to reflect scratch rebuild (no JSON re-import)
- Added minimum workflow list to rebuild
- Estimated time extended to 3-5 hours

SESSION-HANDOFF-PROTOCOL.md updated:
- Dify RAG marked ACTIVE PRIORITY (not medium)
- n8n section updated to reflect rebuild decision
- Future Chroniclers will see these as active work, not deferred

Signed-off-by: The Chronicler #28
This commit is contained in:
Claude
2026-03-04 22:39:34 +00:00
parent bb4a261d44
commit fcb84b72d8
2 changed files with 40 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
SESSION-HANDOFF-PROTOCOL.md
View File

@@ -68,15 +68,18 @@ Read these files IN ORDER:
- **Time:** 1-2 hours setup, 15 min/week maintenance
- **Value:** HIGH for accessibility + knowledge synthesis
**Dify Chat RAG (Medium Priority):**
**Dify Chat RAG (ACTIVE PRIORITY — March 4, 2026):**
- Chat interface configured but retrieval accuracy needs tuning
- **Next Steps:** Increase Top K to 10-15, lower score threshold to 0.3
- Test with specific queries about infrastructure, tasks
- **Status:** Actively queued, not deferred
**n8n Service (Deferred - Task #34):**
- Node registry corrupted (core nodes broken)
- PHP workaround (`sync_codex.php`) handling Git sync successfully
- Factory reset procedure documented but not urgent
**n8n Rebuild (ACTIVE — March 4, 2026):**
- Decision made to rebuild workflows from scratch (not restore old ones)
- Node registry corrupted, core nodes broken
- PHP workaround (`sync_codex.php`) functional as interim
- **Next Steps:** Wipe volume, fresh container, rebuild workflows from scratch
- See Task #34 for full rebuild procedure
**SSL Certificate (Pokerole Wiki):**
- Certbot has urllib3 connection hang bug

58
docs/core/tasks.md
View File

@@ -775,8 +775,9 @@ Add custom "Modpack Version" field to Pterodactyl panel for version tracking vis
**Estimated Total Time:** ~78-99 hours across all tasks
**Critical Path:** Frostwall Protocol → Mailcow → Everything else
**Big Wins Available:**
- Firefrost Codex (RAG tuning, 2 hours remaining)
- NotebookLM Integration (accessibility + knowledge synthesis)
- Dify RAG tuning (ACTIVE — Top K + score threshold adjustments, ~2 hours)
- n8n Rebuild from scratch (ACTIVE — workflows, ~3-5 hours)
- NotebookLM Integration (accessibility + knowledge synthesis, ready to deploy)
**Recent Completions:**
- Pokerole Wiki (Feb 27) - 107 Pokémon LIVE in 6 hours!
@@ -855,41 +856,46 @@ Small improvements to Whitelist Manager:
---
### 34. n8n Factory Reset - Node Registry Recovery
**Time:** 2-3 hours
**Status:** DEFERRED
### 34. n8n Rebuild - Workflows From Scratch
**Time:** 3-5 hours
**Status:** ACTIVE — REBUILD FROM SCRATCH (decision: March 4, 2026)
**Priority:** Tier 2 - Major Infrastructure
**Documentation:** `docs/troubleshooting/n8n-node-registry-corruption.md`
Reset n8n instance on TX1 to resolve corrupted node registry preventing workflow execution.
Rebuild n8n instance on TX1 with fresh workflows from scratch. Original workflows are considered abandoned — do not attempt to restore them.
**Problem:** Core nodes (HTTP Request, Execute Command) fail with "Node not found" errors.
**Problem:** Core nodes (HTTP Request, Execute Command) fail with "Node not found" errors due to corrupted node registry.
**Current Workaround:** PHP script (`sync_codex.php`) handling Codex Git sync directly.
**Decision (March 4, 2026):** Rather than restoring old workflows, rebuild from scratch. Cleaner, more maintainable, aligned with current infrastructure state.
**Reset Procedure:**
1. Export all workflows to JSON backup
2. Backup credentials and settings
3. Stop n8n container
4. Backup existing volume to `.backup` folder
5. Wipe `./volumes/n8n/*` directory
6. Recreate container (fresh initialization)
7. Re-import workflows and credentials
8. Test core nodes functionality
9. Restore scheduled executions
**Current Workaround:** PHP script (`sync_codex.php`) handling Codex Git sync directly. Functional but not the long-term solution.
**Rebuild Procedure:**
1. Stop n8n container
2. Wipe `./volumes/n8n/*` directory completely
3. Recreate container (fresh initialization)
4. Verify core nodes functional (HTTP Request, Execute Command)
5. Rebuild workflows from scratch (do NOT re-import old JSON)
6. Document each new workflow as built
7. Test all workflows end-to-end
8. Re-enable scheduled executions
9. Monitor for 24 hours
**Workflows to Rebuild (minimum):**
- Codex Git sync (currently handled by PHP workaround)
- Discord notifications
- Any other automations identified during rebuild
**Prerequisites:**
- Workflow JSON exports backed up
- Credentials documented
- Maintenance window scheduled (low-traffic time)
- Credentials documented before wipe
- PHP workaround remains as fallback during rebuild
**Post-Reset:**
- Verify Git sync workflow works
**Post-Rebuild:**
- Verify Git sync workflow replaces PHP workaround
- Test Discord notifications
- Re-enable hourly scheduling
- Monitor for 24 hours
**Alternative:** Keep PHP workaround permanently if simpler/more reliable.
- Document all new workflows in task directory
- Remove PHP workaround once n8n confirmed stable
---