10 Commits (ade0130e67e88b9052d0dc4e76d48d98d3328485)

Author	SHA1	Message	Date
René Cannaò	313f637cf0	Merge branch 'v3.1-vec' into v3.1-MCP1 ... Signed-off-by: René Cannaò <rene.cannao@gmail.com>	1 month ago
Rene Cannao	4eab519848	Implement MySQL connection pool for MySQL_Tool_Handler ... Added built-in connection pool to MySQL_Tool_Handler for direct MySQL connections to backend servers. Changes: - Added MySQLConnection struct with MYSQL* pointer, host, port, in_use flag - Added connection_pool vector, pool_lock mutex, pool_size counter - Implemented init_connection_pool() to create MYSQL connections using mysql_init/mysql_real_connect - Implemented get_connection() and return_connection() with thread-safe locking - Implemented execute_query() helper method for executing SQL and returning JSON results - Updated tool methods to use actual MySQL connections: - list_schemas: Query information_schema.schemata - list_tables: Query information_schema.tables with metadata - describe_table: Query columns, primary keys, indexes - sample_rows: Execute SELECT with LIMIT - sample_distinct: Execute SELECT DISTINCT with GROUP BY - run_sql_readonly: Execute validated SELECT queries - explain_sql: Execute EXPLAIN queries - Fixed MYSQL forward declaration (use typedef struct st_mysql MYSQL) The connection pool creates one connection per configured host:port pair with 5-second timeouts for connect/read/write operations.	1 month ago
Rene Cannao	fa301948b2	Remove genai_demo_event binary from tracking and update .gitignore	1 month ago
Rene Cannao	f0a32c00b0	Add rerank support to GenAI prototype via llama-server ... - Change OP_RAG to OP_RERANK operation type - Add RerankResult and RerankResultArray structs for index/score pairs - Add call_llama_rerank() function calling http://127.0.0.1:8012/rerank - Update listener_loop to read query string for rerank operations - Update worker_loop to handle OP_RERANK requests - Add send_rerank_request() method to Client class - Update has_response() to display rerank results (index, score) - Modify send_request() to randomly choose embedding (70%) or rerank (30%) - Add 7 sample queries for rerank testing - Update response header to use result_ptr (instead of embedding_ptr) - Pass query and documents via shared memory (minimal copying)	1 month ago
Rene Cannao	aa53610924	Add batch embedding support and scale up GenAI prototype ... - Add BatchEmbeddingResult struct for contiguous multiple embeddings - Add call_llama_batch_embedding() for batch API requests - Update worker loop to use batch processing for all documents - Scale config: 8 workers, 20 clients, 1-10 docs per request - Add 17 more sample documents (25 total) - Update client to display batch embedding sizes (N x 1024 floats)	1 month ago
Rene Cannao	2c0f3a2e64	Evolve genai_demo_event to working POC with real embeddings ... Transform genai_demo_event.cpp from skeleton to working POC that: - Integrates with real llama-server on port 8013 for embeddings - Uses shared memory (passing pointers, not copying data) - Supports single or multiple documents per request - Properly transfers memory ownership between GenAI and client Architecture changes: - Document struct: passed by pointer from client to GenAI - RequestHeader: includes document_count and operation type - ResponseHeader: includes embedding_size and embedding_ptr - EmbeddingResult: allocated by GenAI, owned by client after response libcurl integration: - HTTP POST to llama-server embedding API - JSON parsing of embedding responses - Error handling for network/API failures Key features: - Clients wait for response before sending next request - (ensures document pointers remain valid) - GenAI workers handle multiple concurrent requests - Embedding dimension: 1023 floats (from llama-server) - Processing time: 30-250ms (real API latency) Results: - 5 clients completed 9 embedding requests - All embeddings successfully retrieved - Zero-copy data transfer via shared memory pointers - Early termination when all work completed Future-ready: - Operation enum (OP_EMBEDDING, OP_COMPLETION, OP_RAG) - Extensible for other GenAI operations - Document count supports batch processing	1 month ago
Rene Cannao	012142eeed	Fix event-driven GenAI demo and add early termination ... Fixed a critical bug in socketpair usage where clients were reading back their own requests instead of receiving processed responses from GenAI. Bug fix: - Changed client to write requests to read_fd_ instead of genai_fd_ - GenAI now correctly receives requests from client's write fd - Responses flow properly: client -> GenAI -> client Enhancements: - Added early termination when all work is done (no more waiting) - Reduced request send probability to 8% (more realistic spacing) - Increased requests per client to 5-15 (longer demo) Results: - Processing times now show realistic 100-500ms values - Queue depth occasionally > 0 when workers are busy - Clients properly transition through WAITING state - Early termination triggers when all clients complete	1 month ago
Rene Cannao	11d183a340	Add event-driven GenAI demo ... Implemented a new event-driven demonstration of the GenAI module that provides better testing isolation and observability. Unlike the thread-based version, this uses a single main event loop with epoll to manage all clients. Key features: - Client class as state machine (NEW → CONNECTED → IDLE → WAITING → DONE) - Single main event loop managing all client I/O via epoll - Random client addition at configurable intervals (probability-based) - Random request sending from idle clients (probability-based) - Periodic statistics printing showing active clients and states - Config struct for tuning behavior (workers, max_clients, probabilities) The demo demonstrates: 1. Multiple concurrent clients connecting to GenAI module 2. Non-blocking async request/response communication 3. Client lifecycle management and cleanup 4. Realistic traffic patterns with randomized behavior Files: - genai_prototype/genai_demo_event.cpp: New event-driven demo - genai_prototype/Makefile: Updated to build both demos	1 month ago
Rene Cannao	89285aa436	Add comprehensive Doxygen documentation to genai_demo.cpp ... Add detailed Doxygen comments throughout the GenAI prototype: File-level documentation: - Architecture overview and design principles - Component descriptions - Communication flow diagrams (MSC) - Build and run instructions Struct/Enum documentation: - RequestHeader: Detailed field descriptions and usage examples - ResponseHeader: Field descriptions and example code - Operation enum: Operation types with descriptions Class documentation: - GenAIModule: Architecture, threading model, thread safety, usage examples - Client: Communication pattern, async operation, usage example Method documentation: - All public and private methods with @param, @return, @pre, @post - Detailed descriptions of algorithms and flow - Thread safety notes - Usage examples where appropriate Member variable documentation: - All class members with ///< descriptions - Clear explanations of purpose and usage Total documentation: ~900 lines of detailed Doxygen comments	1 month ago
Rene Cannao	5dad6255d2	Add GenAI module prototype ... Create standalone prototype demonstrating the GenAI module architecture with thread pool, epoll-based listener, and async socket communication. Components: - genai_demo.cpp: Full working prototype with: * Thread pool (configurable workers) * epoll-based listener thread * Thread-safe request queue * socketpair communication * Multiple concurrent clients * Simulated async processing - Makefile: Build system with clean, run, debug targets - README.md: Documentation of architecture and enhancement roadmap - .gitignore: Exclude build artifacts This prototype will be enhanced before integration into ProxySQL main codebase, testing real LLM API calls, batching, metrics, and error handling.	2 months ago