HiddenBag Pipeline Rework Prompt

Copy this into Claude Code from the hiddenbag-v2 directory:

I need to rework our entire betting picks pipeline. The goal: Discord server (1469086083701542956) becomes the SINGLE SOURCE OF TRUTH for all capper picks flowing into Supabase.

CURRENT STATE (research these files first):

• Schema: FULL_MIGRATION.sql (hb_cappers, hb_picks, hb_discord_messages, hb_game_odds, hb_consensus_snapshots)
• Consensus SQL: C:\Users\mpmmo\dailyai-picks-local\sql\supabase-consensus-v2.sql
• Consensus spec: CONSENSUS-TRACKER-SPEC.md
• Sheets sync: scripts/sheets-to-supabase.py
• Routing config: C:\Users\mpmmo\FULL-LOCAL-PRODUCTION\routing_config.json
• Capper list: cappersnames.txt (79 cappers)

DISCORD SERVER CHANNEL MAP (1469086083701542956 — "THB WORK SPLIT"):

VIP CAPPERS (36 channels):

free cappers (41 channels):

KNOWN MULTI-CAPPER CHANNELS (require author parsing):

• bonus-vip (1469088404104020079) — VIP bundle, multiple cappers post here
• bonusfree (1469438421058453659) — Free bundle, multiple cappers post here
• thecappercollective (1469086720900206817) — Collective channel

KNOWN CAPPER ALIASES (same person, different names):

NEW ARCHITECTURE:

Data Sources (Priority Order):

1. Discord Bot listening to server 1469086083701542956 — every channel, every message
2. Google Sheets (existing AllPicks/Picks tabs) for sources not in Discord
3. Both feed into the SAME Supabase tables

New/Modified Tables Needed:

hb_discord_channels (NEW — track every channel):

• channel_id, channel_name, server_id
• source_type: 'single_capper' | 'multi_capper' | 'algo' | 'aggregate'
• default_capper_id (for single-capper channels)
• is_active, created_at

hb_raw_messages (NEW — store EVERYTHING before parsing):

• message_id, channel_id, author_name, content, attachments_json
• source: 'discord' | 'sheets' | 'telegram'
• parsed_status: 'pending' | 'parsed' | 'failed' | 'skipped'
• parsed_at, error_message
• This is our audit trail — never delete raw data

hb_cappers (MODIFY):

• Add: discord_channel_id, aliases TEXT[] (array of known name variations)
• Add: source_priority ('discord' | 'sheets' | 'telegram')
• Add: auto_detected BOOLEAN (for new cappers found by parser)

hb_picks (MODIFY):

• Add: source_message_id (FK to hb_raw_messages)
• Add: source_type ('discord' | 'sheets' | 'telegram')
• Add: confidence_score (0-100, how sure parser is about the parse)
• Add: parse_method ('regex' | 'vision' | 'manual' | 'sheets_import')

hb_capper_aliases (NEW — solve the fragmentation problem):

• alias_text, canonical_capper_id
• This replaces the current broken duplicate-name system

Multi-Capper Channel Splitting:

Some Discord channels contain picks from MULTIPLE cappers. The bot needs to:

1. Check if channel is 'multi_capper' type
2. Parse author name / embedded capper name from message content
3. Match to hb_capper_aliases → resolve to canonical capper
4. If no match, flag as 'new_capper_detected' for review

Pipeline Flow:

Discord Message → hb_raw_messages (store raw) → Parser (regex + vision for images) → hb_picks (normalized, linked to capper) → Consensus SQL (find 2+ cappers on same side) → hb_consensus_snapshots → Auto-grading (scores from hb_game_odds)

CSV Export for Validation:

• Script to dump hb_picks to CSV with: date, capper, sport, team, line, odds, units, result, source, confidence
• Script to dump consensus plays to CSV
• Use these to verify data quality before trusting the pipeline

What I need you to build:

1. SQL migration for all new/modified tables above
2. Discord bot listener (Node.js, discord.js) that captures every message from the server
3. Parser module that extracts picks from messages (handle text + image attachments)
4. Alias resolution system for capper name matching
5. CSV export scripts for validation
6. Migration script to backfill existing hb_picks with source info
7. Dashboard page or API endpoint showing: messages collected, parse rate, unmatched cappers, confidence distribution

SEED DATA — Pre-populate hb_discord_channels with the channel map above on first run.

SEED DATA — Pre-populate hb_capper_aliases with the known aliases table above.