Data-Driven AI Responses & Purpose

Release digest

Data-Driven AI Responses 🎯 - Chat now cites REAL numbers instead of hypotheticals
- AI tool get_crypto_trust_score now returns full project data: token_unlocks, investors, TVL, score_breakdown
- System prompt instructs: "NEVER say 'if' — state actual facts from the data"
- Example: "55.3% of tokens already unlocked" instead of "if there are large unlocks..."

Project Cleanup 🧹 - Organized root directory
- Moved debug files to .tmp_debug/
- Moved logs to .logs/
- Moved data exports to scripts/data/

Crypto Module API 🔧 - Fixed 404 errors on production
- Root cause: Fallback router had wrong prefix (/api/v2/crypto → /api/crypto)
- Restored hybrid search (DB + CoinGecko) for new projects
- New projects now show with "New" badge as expected
SQL Schema Alignment 🗄️ - Removed non-existent columns from queries
- Fixed: price_change_24h, volume_24h, github_repos, updated_at
- Added missing columns: project_stage, treasury_usd, certik_score, tweetscout_score

Purpose: AI assistant that cites REAL data and "sees" visual context.
Multimodal: Supports image (screenshots, charts) and file (PDF) analysis.
Implementation: Function Calling tools fetch data + Gemini 1.5 Pro Vision for visual tasks.
Key Rule: AI NEVER says "if" or "якщо" — it states actual facts from database.
Example: "55.3% of tokens already unlocked, next unlock: March 2026 (15M FXS)" instead of "if there are large unlocks..."