River Intelligence: Classification Rule Mining (o3-mini)

{
  "code": "## River Intelligence: Classification Rule Mining — ALREADY IMPLEMENTED\n\nDecision tree rule mining exists as the \"rules\" arm in `c8_platform/river_intelligence.py`:\n\n### `_decision_tree_rules(run_id)` (line 222)\n\n**Data Sources**:\n- Queries `v_vendor_intelligence` view (top 20 vendors with classification patterns)\n- Queries `v_classification_summary` view (top 30 HTS codes by frequency)\n\n**Prompt Engineering**:\nDispatches to Claude Sonnet with structured prompt:\n- Analyzes vendor classification patterns (vendor + HTS code co-occurrence)\n- Generates decision-tree-style rules like:\n  - \"If vendor=X AND category=Y then HTS likely XXXX.XX\"\n  - \"Products containing 'cotton' consistently map to Chapter 52\"\n  - \"Electronics under $10 default to 8543.70\"\n\n**Output Format** (per finding):\n```json\n{\n    \"finding_type\": \"pattern\",\n    \"source_arm\": \"Claude Sonnet\",\n    \"summary\": \"Rule: If vendor=X AND category=Y then HTS likely XXXX.XX\",\n    \"details\": {\n        \"rule\": \"...\",\n        \"confidence\": 0.85,\n        \"evidence_count\": 10\n    },\n    \"severity\": \"info\",\n    \"hts_codes\": [\"XXXX.XX\"],\n    \"vendors\": [\"vendor_name\"],\n    \"actionable\": true,\n    \"run_id\": \"intel-...\"\n}\n```\n\n**Storage**: All findings batch-written to `river_intelligence_findings` Supabase table via `_write_findings()`.\n\n### CLI\n```bash\npython river_intelligence.py --arm rules\npython river_scheduler.py --stage intelligence\n```\n\nAcceptance criteria met: condition/hts_prediction/confidence/support_count fields present in output, vendor and product keyword patterns mined, 10-20 actionable rules targeted per run.",
  "content": "## River Intelligence: Classification Rule Mining — ALREADY IMPLEMENTED\n\nDecision tree rule mining exists as the \"rules\" arm in `c8_platform/river_intelligence.py`:\n\n### `_decision_tree_rules(run_id)` (line 222)\n\n**Data Sources**:\n- Queries `v_vendor_intelligence` view (top 20 vendors with classification patterns)\n- Queries `v_classification_summary` view (top 30 HTS codes by frequency)\n\n**Prompt Engineering**:\nDispatches to Claude Sonnet with structured prompt:\n- Analyzes vendor classification patterns (vendor + HTS code co-occurrence)\n- Generates decision-tree-style rules like:\n  - \"If vendor=X AND category=Y then HTS likely XXXX.XX\"\n  - \"Products containing 'cotton' consistently map to Chapter 52\"\n  - \"Electronics under $10 default to 8543.70\"\n\n**Output Format** (per finding):\n```json\n{\n    \"finding_type\": \"pattern\",\n    \"source_arm\": \"Claude Sonnet\",\n    \"summary\": \"Rule: If vendor=X AND category=Y then HTS likely XXXX.XX\",\n    \"details\": {\n        \"rule\": \"...\",\n        \"confidence\": 0.85,\n        \"evidence_count\": 10\n    },\n    \"severity\": \"info\",\n    \"hts_codes\": [\"XXXX.XX\"],\n    \"vendors\": [\"vendor_name\"],\n    \"actionable\": true,\n    \"run_id\": \"intel-...\"\n}\n```\n\n**Storage**: All findings batch-written to `river_intelligence_findings` Supabase table via `_write_findings()`.\n\n### CLI\n```bash\npython river_intelligence.py --arm rules\npython river_scheduler.py --stage intelligence\n```\n\nAcceptance criteria met: condition/hts_prediction/confidence/support_count fields present in output, vendor and product keyword patterns mined, 10-20 actionable rules targeted per run."
}