2 Commits (f2db55f0c8aac088a1c40d51bc0630f4ef27eb86)

Author	SHA1	Message	Date
Rene Cannao	612ef326bc	Fix sqlite-rembed demonstration scripts and add environment variable support ... This commit addresses critical fixes to the sqlite-rembed demonstration scripts and adds environment variable support for API key configuration. Key Changes: 1. Fixed sqlite-rembed-demo.sh similarity search queries: - Changed FROM demo_embeddings e JOIN ... WHERE embedding MATCH pattern - To correct subquery pattern required by sqlite-vec: FROM (SELECT rowid, distance ... LIMIT) e JOIN ... - This resolves "A LIMIT or 'k = ?' constraint is required on vec0 knn queries" error - All three similarity search queries now use proper subquery structure 2. Added comprehensive cleanup at script start: - Added DROP TABLE IF EXISTS for all demo_embeddings related tables - Prevents "UNIQUE constraint failed on demo_embeddings primary key" errors - Uses INSERT OR REPLACE instead of INSERT for embedding storage 3. Added environment variable support for API_KEY: - Updated all demonstration scripts to use API_KEY="${API_KEY:-YOUR_API_KEY}" - Users can now set API_KEY environment variable: export API_KEY="actual_key" - Falls back to YOUR_API_KEY placeholder if environment variable not set - Improves security by avoiding hardcoded keys in scripts 4. Updated documentation: - Modified SQLITE-REMBED-TEST-README.md to document environment variable usage - Updated comments in all scripts to mention environment variable option Files Modified: - doc/sqlite-rembed-demo.sh: Fixed similarity search queries, added cleanup, added environment variable support - doc/sqlite-rembed-examples.sh: Added environment variable support - doc/sqlite-rembed-test.sh: Added environment variable support - doc/SQLITE-REMBED-TEST-README.md: Updated documentation for env var support Verification: - sqlite-rembed-demo.sh now runs successfully end-to-end - All similarity search queries execute without errors - Environment variable fallback works correctly - Scripts maintain backward compatibility with direct key replacement	4 months ago
Rene Cannao	e75bd7c84a	Add comprehensive sqlite-rembed examples and documentation ... This commit adds a complete set of well-documented examples and test scripts for the sqlite-rembed integration in ProxySQL. The sqlite-rembed extension enables text embedding generation via HTTP API calls directly within SQL queries, complementing the sqlite-vec extension for vector similarity search. Key additions: 1. sqlite-rembed-examples.sql - Primary SQL demonstration file - Standalone SQL file with 8 phases of examples - Demonstrates complete AI pipeline: client config → embedding gen → storage → search - Includes proper subquery pattern for similarity search JOINs (vec0 requirement) - Well-documented with clear explanations for each phase 2. sqlite-rembed-test.sh - Comprehensive test suite - 9-phase test covering all integration aspects - Color-coded output with test result tracking - Error handling and edge case testing 3. SQLITE-REMBED-TEST-README.md - Complete documentation - Detailed test suite documentation - Usage instructions, troubleshooting, CI/CD integration examples 4. Supporting bash scripts for different use cases - sqlite-rembed-examples.sh - Phase-by-phase interactive examples - sqlite-rembed-demo.sh - Single-session demonstration script Security considerations: - All hardcoded API keys replaced with YOUR_API_KEY placeholder - Clear comments instructing users to replace with actual credentials - Synthetic OpenAI endpoint used as example (public test endpoint) Technical details: - Embedding dimensions: 768 (nomic-embed-text-v1.5 model) - Embedding size: 3072 bytes (768 × 4 bytes per float) - Similarity search pattern: Uses subqueries for JOIN compatibility with vec0 - Client configuration: temp.rembed_clients virtual table (per-connection) The examples provide a solid baseline for building applications that leverage sqlite-rembed and sqlite-vec in ProxySQL for AI-powered applications.	4 months ago