🏗️ Architecture: How should the Intelligence Pyramid be unblocked and all sys...

Date: 2026-02-22
Time: 03:27:02 UTC
Session: gspace1-133.24
Agent: backend
Confidence: 92%
Status: ✅ Documented

📋 Decision Context

Question:

How should the Intelligence Pyramid be unblocked and all system gaps be closed?

Task: Not specified

🎯 Decision Made

Chosen:

Fix 5 cascading blockers: wire OutcomeDetector before PatternMining, relax outcome queries from strict true to $ne:false, lower promotion thresholds 10/0.9 to 3/0.8, add daily promotion cron, lower trigger sample sizes 10 to 5. Also fixed hardcoded CORS URLs, upgraded silent error handlers, and reduced outcome detection delay from 1h to 15min.

Options Considered:

• No alternatives documented

💭 Reasoning

The pyramid wasn't growing because PatternMiningJob queried outcome.wasSuccessful:true but only 441/11924 decisions had outcomes. The OutcomeDetector runs every 5min but AFTER pattern mining. Relaxing queries to $ne:false and calling OutcomeDetector before mining means patterns can now be extracted from the majority of decisions. Lowered promotion thresholds because sparse data makes 10 uses + 90% success unreachable. Added scheduled daily promotion checker. Fixed 10 hardcoded localhost CORS URLs that would break Railway production. Upgraded 3 silent error catches to warn-level logging in task-scheduler.

📁 Files Involved

• apis/backend-ai-api/src/app/learning/jobs/pattern-mining.job.ts
• apis/backend-ai-api/src/app/learning/learning.service.ts
• apis/backend-ai-api/src/app/learning/triggers/trigger-config.ts
• apis/backend-ai-api/src/app/learning/triggers/adaptive-trigger.service.ts
• apis/backend-ai-api/src/app/learning/services/outcome-detector.service.ts
• apis/backend-ai-api/src/main.ts
• apis/backend-vitto-api/src/main.ts
• apis/backend-vitto-api/src/app/engine/task-scheduler.service.ts

🔗 Related

• Decision ID: dev_1771730822700_5gduh4k
• Session: 133
• Workspace: gspace1

Auto-generated by MCP Smart Consolidation