PhoenixKit V34 Migration: AI Endpoints System

This migration replaces the Accounts + Slots architecture with a unified Endpoints system where each endpoint contains credentials, model selection, and generation parameters in a single entity.

Architecture Change

Before (V32)

• Accounts Table: Provider credentials only
• Slots: Stored in Settings as JSON (4 types × 3 slots each)
• Slots referenced account_id for credentials
• Fallback chain: tries slot 0 → 1 → 2

After (V34)

• Endpoints Table: Unified configuration
  • Provider credentials (name, api_key, base_url, provider_settings)
  • Model selection (single model per endpoint)
  • Generation parameters (temperature, max_tokens, etc.)
• No type categorization (user decides endpoint purpose)
• No fallback chain (explicit endpoint targeting)

Changes

New Table: phoenix_kit_ai_endpoints

• Unified configuration combining account + slot fields
• All generation parameters in one place
• Sort order for display

Modified Table: phoenix_kit_ai_requests

• Added endpoint_id for new requests
• Added endpoint_name for denormalized history
• Kept account_id/slot_index for backward compatibility

Cleanup

• Remove slot Settings entries (ai_text_processing_slots, etc.)
• Keep ai_enabled setting
• Keep accounts table for historical request references