42 Commits (9685cdaa4b87e2cd5be9584c46dca42d9703934d)

Author	SHA1	Message	Date
Rene Cannao	9685cdaa4b	mcp discovery: enforce target-scoped run model and add protocol-aware static harvesting (mysql+pgsql) ... Refactor the MCP query/discovery stack to remove prototype-era MySQL-only assumptions and make discovery/catalog semantics explicitly target-scoped. This commit is intentionally not backward compatible with legacy single-target catalog metadata. Key implementation details: - Discovery_Schema: - runs table now stores target_id, protocol, and server_version - resolve_run_id() now requires target_id and resolves schemas within target scope - create_run() now records target/protocol/server version - added legacy schema detection + destructive catalog table rebuild when old runs layout is found - fixed resultset lifetime/escaping issues in target-aware run resolution - New PostgreSQL static harvester: - added PgSQL_Static_Harvester class and build wiring - harvests schemas/objects/columns/indexes/fks/view definitions into Discovery_Schema - maps pg metadata to object_id-based insert APIs used by Discovery_Schema - Query_Tool_Handler: - constructor simplified to catalog_path only (no mcp-mysql_* runtime ctor deps) - discovery.run_static now requires target_id and dispatches protocol-aware harvester at runtime - catalog.*, agent.run_start, and llm.* tool contracts now require target_id when resolving run_id - all run_id resolution call sites switched to resolve_run_id(target_id, ...) - list_schemas catalog query now scoped by target_id via runs join - digest logging path now resolves run_id in target context - ProxySQL_MCP_Server: - updated Query_Tool_Handler construction to new signature - Docs updated: - added explicit target_id scoping guidance for discovery/catalog/agent/llm workflows - clarified protocol-aware routing and noted legacy examples still present in large script docs Build validation: - Recompiled changed objects with PROXYSQLGENAI=1: - Query_Tool_Handler.oo - Discovery_Schema.oo - PgSQL_Static_Harvester.oo - ProxySQL_MCP_Server.oo	6 days ago
Rene Cannao	7e54883ba5	mcp query debugging: surface rule-id context for blocks and add backend failure SQL details ... Observed confusion: - MCP failures like 'Error 100' / 'Error 101' were visible at endpoint level but gave no direct clue whether they came from query-rule blocks or backend execution failures. - Connection/firewall-style failures also lacked enough target/query context in lower-level logs. Changes in this commit: 1) Log MCP query-rule block/OK matches with rule id and context - In Discovery_Schema::evaluate_mcp_query_rules(): - when error_msg rule action is applied, log: rule_id, tool, target_id, schema, error_msg, query preview. - when OK_msg rule action is applied, log: rule_id, tool, target_id, schema, ok_msg. - This makes it explicit which runtime_mcp_query_rules row produced the returned error. 2) Add detailed backend execution error logs in Query_Tool_Handler - For MySQL and PostgreSQL execution paths (with and without schema switching), failures now log: - target_id - schema (when applicable) - backend error text - SQL query text - Added context for PG search_path failures too. - This disambiguates rule-block errors from actual backend connectivity/authorization/firewall failures. Behavioral notes: - API responses are unchanged (to avoid breaking tests expecting exact messages). - Logging is now sufficiently descriptive to trace each failure to either: - a specific MCP rule id/action, or - a concrete backend execution error. Validation: - Recompiled successfully: - lib/obj/Discovery_Schema.oo - lib/obj/Query_Tool_Handler.oo	7 days ago
Rene Cannao	f15460348c	mcp query diagnostics: include runtime hostgroup status breakdown when target lacks ONLINE backend ... Adds targeted debug context for non-executable MCP targets. When Query_Tool_Handler reports that a target has no ONLINE backend, it now appends a status summary read from the corresponding runtime server table for that hostgroup: - MySQL: runtime_mysql_servers - PostgreSQL: runtime_pgsql_servers Example diagnostics now include status aggregates such as: - ONLINE=0, SHUNNED=1 - OFFLINE_SOFT=1 - no rows in runtime_*_servers for hostgroup X This closes the observability gap where users saw 'target non-executable' but could not immediately tell why the hostgroup failed ONLINE eligibility. Validation: - Recompiled lib/obj/Query_Tool_Handler.oo successfully.	7 days ago
Rene Cannao	f4bc1943fb	mcp query diagnostics: restore strict ONLINE requirement and explain target non-executable failures ... This change does two things: 1) Reverts executable-target policy to strict ONLINE backend status - Query target resolution in Query_Tool_Handler is restored to require UPPER(status)='ONLINE' in runtime_mysql_servers/runtime_pgsql_servers. - This matches expected semantics: non-ONLINE backends should not be considered executable MCP targets. 2) Replaces generic non-executable errors with actionable diagnostics - Added Query_Tool_Handler::format_target_unavailable_error(target_id). - All query-tool paths that previously returned the generic: 'Target is not executable by this handler' now return a reasoned error with context, including: - target_id - protocol - hostgroup_id - auth_profile_id - concrete reason (for example: empty db_username in auth profile, or no ONLINE backend in runtime_*_servers) - Internal logs in get_connection()/get_pgsql_connection() now emit the same detailed reason. 3) Improves MCP endpoint failure logging with SQL/target context - In MCP_JSONRPC_Resource::handle_tools_call(), failed tool calls now log: - endpoint/tool/error - full arguments payload (existing) - additional parsed details when present: target_id, schema, sql (trimmed for safety) - This makes it explicit what SQL was requested when a tool call fails before execution. 4) Protocol normalization retained for robustness - MCP_Threads_Handler::load_target_auth_map() lowercases protocol values when loading runtime profile joins. - This avoids protocol-casing drift causing inconsistent routing behavior. Build validation: - Successfully recompiled modified objects: - lib/obj/Query_Tool_Handler.oo - lib/obj/MCP_Endpoint.oo - lib/obj/MCP_Thread.oo Resulting behavior expected on next test run: - If a target is non-executable, the response and logs will state exactly why (ONLINE/backend/auth reason), instead of a generic error. - Failure logs will also include the attempted SQL text (for tool calls that carry sql arguments), clarifying whether a backend query was actually executed or blocked before execution.	7 days ago
Rene Cannao	49f811a638	mcp query: stop misclassifying reachable targets as non-executable ... Observed failure: - MCP tools received valid target_id values, but /mcp/query returned 'Target is not executable by this handler'. - This prevented rule evaluation, digest accounting, and hit counters from advancing, causing multiple TAP phases to fail. Root cause addressed in this commit: - Target endpoint resolution required runtime server status to be exactly ONLINE. - In practice, reachable backends can temporarily appear in other statuses (for example monitor-related transitions), while MCP direct connections with profile credentials are still valid. - Protocol casing inconsistencies could also lead to fragile routing behavior when profile data is consumed. Changes: 1) Relax backend eligibility during endpoint resolution - In Query_Tool_Handler::refresh_target_registry(), endpoint discovery now accepts any server status except OFFLINE_HARD. - ONLINE is still preferred deterministically via ORDER BY, but non-ONLINE candidates remain eligible for MCP execution. 2) Normalize protocol values to lowercase - In MCP_Threads_Handler::load_target_auth_map(), target protocol is normalized to lowercase when loaded from runtime profiles. - In Query_Tool_Handler::refresh_target_registry(), protocol is normalized again defensively before routing decisions. 3) Improve diagnostics for target executability - Added explicit warnings when: - a target resolves to a backend but has empty db_username, - a target has no eligible backend for its hostgroup/protocol. - These logs make it clear why a target is being marked non-executable. Expected behavioral impact: - MCP query tools should execute for valid target_id entries even when monitor status is not strictly ONLINE. - Query-rules blocking/rewriting/OK actions should be exercised again. - stats_mcp_query_rules hits and stats_mcp_query_digest counters should resume incrementing once queries run. Validation performed: - Recompiled modified objects successfully: - lib/obj/Query_Tool_Handler.oo - lib/obj/MCP_Thread.oo	7 days ago
Rene Cannao	6a788e48c4	mcp: make /mcp/query self-healing when targets/backends appear after startup ... Problem addressed: - MCP query endpoint could stay unusable with 'Tool Handler not initialized' after restart/reload flows. - This was triggered when Query_Tool_Handler could not build an executable pool at init time (for example profiles loaded before ONLINE servers), leaving runtime commands like 'LOAD MCP QUERY RULES FROM MEMORY' blocked behind a NULL query tool handler. - Users had to manually toggle MCP enablement to recover, which is the opposite of expected self-healing behavior. What this commit changes: 1) Query_Tool_Handler pool init is now resilient and idempotent - init_connection_pool() now starts with close() so reinitialization fully resets stale mysql/pgsql pool state before rebuilding from runtime profile + server tables. - If no executable targets are available, init_connection_pool() now returns success with a warning instead of hard failure. This allows the query tool handler (and /mcp/query endpoint) to stay initialized even before backends are ready. 2) Lazy auto-rebuild on first query usage - get_connection() and get_pgsql_connection() now: - refresh target registry before resolution, - attempt to use an existing pooled connection, - if unavailable, trigger a full pool rebuild (init_connection_pool()) and retry once. - This provides automatic recovery when hostgroups/servers/profiles are loaded or changed after MCP startup, without requiring manual MCP disable/enable. 3) Admin runtime rule load path attempts MCP recovery - load_mcp_query_rules_to_runtime() now detects NULL query tool handler and calls load_mcp_server() once before failing. - This turns a hard, immediate admin error into a self-recovery attempt consistent with MCP runtime semantics. Behavioral impact: - /mcp/query endpoint remains online even when there are temporarily zero executable targets. - As soon as compatible runtime targets/backends exist, run_sql_readonly/explain_sql can recover automatically on demand. - LOAD MCP QUERY RULES TO/FROM RUNTIME no longer fails immediately on first NULL handler condition; it retries after MCP server recovery. Validation performed: - Recompiled modified objects successfully: - lib/obj/Query_Tool_Handler.oo - lib/obj/ProxySQL_Admin.oo - Full TAP runtime test execution is not possible in this sandbox due blocked local TCP socket creation; validation should be run in the normal test environment where ProxySQL/MySQL/PGSQL are reachable.	7 days ago
Rene Cannao	af0411bd46	MCP: add target-aware rules/stats tests, explain_sql rule coverage, and AI local docker TAP infra ... This commit completes end-to-end MCP query-rules validation for multi-target routing and introduces a self-contained TAP infra for the `ai` group that can run both in Jenkins and manually. Main MCP/runtime changes: - Extended MCP query-rule matching context to include both `target_id` and resolved backend `username`. - Added `target_id` column to `mcp_query_rules` and `runtime_mcp_query_rules` table definitions. - Extended `stats_mcp_query_rules` to include `username` and `target_id` alongside `rule_id` and `hits`. - Updated load/save/runtime refresh paths to persist and rehydrate the expanded MCP rule schema. - Wired the rule engine into `explain_sql` so MCP rules apply consistently across `run_sql_readonly` and `explain_sql`. - Included startup-order fix in `src/main.cpp` to initialize MCP/GenAI thread handlers early, preventing startup crashes in PROXYSQLGENAI builds. Test coverage changes: - Updated existing MCP TAP phases to assert target-aware and username-aware behavior: - `test_phase4_stats.sh` - `test_phase6_eval_block.sh` - Added new MCP TAP phases: - `test_phase10_eval_explain.sh` (rule engine coverage for `explain_sql`) - `test_phase11_pgsql_target.sh` (pgsql target routing/rule/stats coverage; graceful skip if no pgsql target is configured) - Updated `test_mcp_query_rules-t.sh` runner to execute new phases. AI group isolated infra (manual + CI compatible): - Added `test/tap/groups/ai/docker-compose.yml` with MySQL 9.0 and PostgreSQL 16 backends. - Added lifecycle scripts: - `docker-compose-init.bash` - `docker-compose-destroy.bash` - `pre-proxysql.bash` - `post-proxysql.bash` - Extended `test/tap/groups/ai/env.sh` with local default ports/credentials/target IDs used by MCP TAP tests. - Added `test/tap/groups/ai/README.md` documenting manual execution flow outside Jenkins. Outcome: - MCP tests now validate routing-aware rule enforcement and stats attribution for both MySQL and PostgreSQL targets. - The `ai` TAP group can be run with an isolated local backend stack without relying on external Jenkins infra repositories.	7 days ago
Rene Cannao	013864b36f	MCP: introduce profile-based target/auth routing and unified LOAD/SAVE MCP PROFILES commands ... This commit introduces the core MCP routing architecture that decouples client-facing target selection from backend connection details. Architecture introduced: - New MCP auth profile table to store backend credentials/policy (`mcp_auth_profiles`). - New MCP target profile table to map opaque `target_id` -> protocol/hostgroup/auth profile (`mcp_target_profiles`). - Runtime mirrors for both profile tables. Behavioral changes: - MCP query tools resolve execution context by `target_id` and route internally by protocol + hostgroup. - Client no longer needs backend-specific connection details; only logical target identifiers. - Unified admin commands for profile lifecycle management: - `LOAD MCP PROFILES TO/FROM MEMORY/RUNTIME/DISK` - `SAVE MCP PROFILES TO/FROM MEMORY/RUNTIME/DISK` Design goals: - Support multi-backend MCP execution (MySQL + PostgreSQL) behind one endpoint. - Provide server-side credential management and cleaner operational model. - Replace legacy one-backend POC configuration with scalable target/auth abstractions. Result: - MCP routing is now dynamic, protocol-agnostic at the client layer, and fully managed through profile tables and unified load/save commands.	7 days ago
Rene Cannao	1af768f932	Fix critical issues in GenAI code (PR #5339) ... This commit addresses critical issues identified in the PR review: 1. Build system redundancy (src/Makefile): - Remove redundant PROXYSQLGENAI conditionals in linking commands - The flag doesn't affect linking, so simplify to PROXYSQLCLICKHOUSE only 2. Missing PROXYSQLGENAI guard (lib/Admin_Handler.cpp): - Add #ifdef PROXYSQLGENAI around MCP VARIABLES DISK commands - Ensures MCP commands are only available when GenAI is enabled 3. Broken retry logic (lib/LLM_Clients.cpp): - Remove misleading is_retryable_error() checks that used stale values - last_curl_code and last_http_code were never updated - Simplify to retry on empty responses only (documented limitation) - Fix thread-safety: use thread_local std::mt19937 instead of rand() 4. Resource leak (lib/MySQL_Tool_Handler.cpp): - Clean up previously created connections on init failure - Both mysql_init and mysql_real_connect error paths now clean up 5. Race condition (lib/Query_Tool_Handler.cpp): - Add missing pool_lock in find_connection() - Prevents race condition when accessing connection_pool Fixes identified by automated PR review agents.	3 weeks ago
Rene Cannao	b965fc6df4	Fix PROXYSQLGENAI build - resolve circular includes and missing headers ... This commit fixes compilation errors when building with PROXYSQLGENAI=1: Header file fixes: - Add missing #endif /* PROXYSQLGENAI */ before header guards in multiple headers - Remove #include cpp.h from GenAI headers to avoid circular dependencies Source file fixes: - Add #include proxysql.h to GenAI .cpp files that were missing it - Add #include Static_Harvester.h to Query_Tool_Handler.cpp for forward decl Build system fixes: - Remove vec.o from libproxysql.a (it's linked separately in src/Makefile) - Prevents duplicate symbol errors during linking Test files: - Rename MCP test files with .sh suffix to prevent make clean deletion Both build modes now work: - make build_lib_debug (without GenAI) - PROXYSQLGENAI=1 make build_lib_debug (with GenAI)	3 weeks ago
Rene Cannao	48bc7dd7bf	Merge v4.0 GenAI features into v3.0 with conditional compilation ... This commit merges the experimental v4.0 GenAI/MCP features into the stable v3.0 branch using conditional compilation. All v4.0 features are disabled by default and only enabled when PROXYSQLGENAI=1 is set at compile time. Changes: Build System: - Modified main Makefile to pass PROXYSQLGENAI flag to sub-makefiles - Modified deps/Makefile to conditionally build sqlite-vec and sqlite-rembed - Modified lib/Makefile to add PSQLGA flag and include GenAI object files - Modified src/Makefile to add PSQLGA flag and conditional linking Headers (wrapped with #ifdef PROXYSQLGENAI): - All 20 new GenAI header files in include/ - Modified cpp.h, proxysql_glovars.hpp, proxysql_admin.h - Modified ProxySQL_Admin_Tables_Definitions.h for GenAI/MCP tables Source Files: - All 22 new GenAI source files in lib/ wrapped with #ifdef PROXYSQLGENAI - Modified src/main.cpp for conditional global variables and init/shutdown - Modified Admin_Handler.cpp for conditional command handlers - Modified Admin_Bootstrap.cpp for conditional table registration - Modified Admin_FlushVariables.cpp for conditional variable flushing - Modified ProxySQL_Admin.cpp for conditional admin methods - Modified ProxySQL_Admin_Stats.cpp for conditional MCP stats functions - Modified proxy_sqlite3_symbols.cpp to always compile (needed by core) - Modified MySQL_Session.cpp for conditional GenAI function calls Test Files: - Renamed test_mcp_query_rules-t to test_mcp_query_rules-t.sh - Renamed test_mcp_rag_metrics-t to test_mcp_rag_metrics-t.sh - Modified anomaly_detection-t.cpp for conditional test execution Usage: # Build without GenAI (v3.0 mode - default) make clean && make build_deps -j$(nproc) && make build_lib -j$(nproc) && make build_src -j$(nproc) # Build with GenAI (v4.0 mode) make clean && PROXYSQLGENAI=1 make build_deps -j$(nproc) && PROXYSQLGENAI=1 make build_lib -j$(nproc) && PROXYSQLGENAI=1 make build_src -j$(nproc)	3 weeks ago
Javier Jaramago Fernández	e351a0df74	feat: Add leading comments (--) detection for 'run_sql_readonly' ... Read only queries are no longer flagged as non-readonly when starting with double dash comments.	4 weeks ago
Javier Jaramago Fernández	a5ef787c7e	feat: Improve logging for MCP and RAG tools ... - Add new column 'endpoint' for 'stats_mcp_query_tools_counters'. - Add new table 'rag_search_log' to log 'log_rag_search_fts' operations.	4 weeks ago
Rene Cannao	3ccfa2bcc1	Address AI code review feedback for PR #5313 ... This commit addresses concerns raised by AI code reviewers (gemini-code-assist, Copilot, coderabbitai) on the initial security fixes. Critical fixes: - Fix lock.release() → lock.unlock() in GenAI_Thread.cpp worker_loop (lock.release() detaches without unlocking, causing deadlock) - Add missing early return after schema validation failure in Query_Tool_Handler.cpp Code quality improvements: - Improve escape_string() memory management in MySQL_Tool_Handler.cpp: - Use std::string instead of new[]/delete[] for buffer management - Check return value of mysql_real_escape_string() for errors - Remove redundant validation checks in validate_sql_identifier functions (character class loop already rejects unsafe characters) - Add backslash escaping to escape_string_literal() for defense-in-depth - Improve column list validation in MySQL_Tool_Handler sample_rows(): - Replace blacklist approach with proper column identifier parsing - Allow qualified identifiers (table.column) - Allow AS aliases (column AS alias) - No longer rejects legitimate column names containing "JOIN" These changes improve robustness while maintaining the security posture of the original SQL injection fixes.	1 month ago
Rene Cannao	c914feb230	Fix security issues identified in PR #5312 code review ... This commit addresses critical and important security vulnerabilities found during comprehensive code review of the Gen AI features merge. Critical fixes: - SQL injection vulnerabilities in MySQL_Tool_Handler.cpp: - Added validate_sql_identifier() for schema/table validation - Added escape_string() for MySQL string escaping using mysql_real_escape_string - Fixed list_tables(), describe_table(), sample_rows(), sample_distinct() - SQL injection vulnerabilities in Query_Tool_Handler.cpp: - Added validate_sql_identifier_sqlite() for identifier validation - Added escape_string_literal() for SQLite string escaping - Fixed list_tables tool and catalog.get_relationships function - Use-after-free race condition in GenAI_Thread: - Changed shutdown_ from int to std::atomic<int> for proper memory ordering - Added additional shutdown check in worker_loop after popping request Important fixes: - Buffer overflow risks from sprintf usage: - Converted all sprintf() calls to snprintf() in GenAI_Thread.cpp - Converted sprintf() to snprintf() in MySQL_Session.cpp - Worker loop shutdown race condition: - Added shutdown check after popping request from queue - Properly clean up client_fd when shutdown is detected These fixes ensure: 1. All user input is properly validated before use in SQL queries 2. String values are properly escaped using database-specific escaping 3. Thread-safe shutdown with proper memory ordering guarantees 4. Bounds-safe string formatting to prevent buffer overflows	1 month ago
Rene Cannao	bbc04974f1	fix: Fix mysql_query failure path and affected_rows race condition ... Fix two issues in Query_Tool_Handler's execute_query functions: 1. mysql_query() failure path now returns immediately after return_connection() instead of continuing to process on bad state. 2. Capture affected_rows BEFORE return_connection() to avoid race condition. Previously, mysql_affected_rows() was called after return_connection(), potentially accessing a stale connection. Apply fixes to both execute_query() and execute_query_with_schema(). Addresses coderabbitai review comments.	1 month ago
Rene Cannao	6305537ba8	fix: Use delete instead of free for SQLite3_result deallocation ... Change free(resultset) to delete resultset in Query_Tool_Handler (extract_schema_name function). SQLite3_result is a C++ class allocated with new, so it must be deallocated with delete, not free(). Using free() causes mixed allocator UB. Addresses coderabbitai review comment.	1 month ago
Rene Cannao	bd6d34f52b	fix: Address SQL injection vulnerabilities from PR #26 review ... - lib/MySQL_Catalog.cpp: Convert search/list/remove to use SQLite prepared statements instead of string concatenation for user parameters - lib/RAG_Tool_Handler.cpp: Add escape_fts_query() function to properly escape single quotes in FTS5 MATCH clauses; update all FTS and vector MATCH queries to use escaped values - lib/Static_Harvester.cpp: Add is_valid_schema_name() validation function to ensure schema names only contain safe characters (alphanumeric, underscore, dollar sign) before using in INFORMATION_SCHEMA queries - lib/Query_Tool_Handler.cpp: Add clarifying comments to validate_readonly_query explaining the blacklist (quick exit) + whitelist (allowed query types) approach - Remove backup file lib/Anomaly_Detector.cpp.bak Addresses gemini-code-assist review comments from PR #26.	1 month ago
Rene Cannao	709649232b	fix: Address AI code review concerns from PR #19 ... This commit addresses valid concerns raised by coding agents (Gemini, Copilot, CoderabbitAI): 1. Fix stats_mcp_query_digest naming conflict (ProxySQL_Admin.cpp): - Made reset and non-reset paths mutually exclusive using else block - Prevents both flags from being true, matching MySQL pattern - Ensures reset takes precedence over non-reset 2. Fix INSERT OR REPLACE sync issue (Admin_Handler.cpp): - Added DELETE before INSERT OR REPLACE in LOAD/SAVE MCP QUERY RULES - Prevents stale rules from persisting when syncing disk <-> memory - Ensures deleted source rows are also removed from target 3. Fix integer division truncation for timeout (Query_Tool_Handler.cpp): - Changed timeout_ms/1000 to (timeout_ms+999)/1000 for ceiling division - Ensures sub-second timeouts (e.g., 500ms) become at least 1 second - Prevents zero-second timeouts from causing unexpected behavior 4. Remove confusing comment (Discovery_Schema.cpp): - Simplified column count comment to be clear and accurate Note: The re_modifiers parsing code already correctly handles VARCHAR "CASELESS" to int conversion (lines 2414-2425), so that review comment was already addressed.	1 month ago
Rene Cannao	a1d9d2f1ba	docs: Add comprehensive documentation to MCP features ... Add detailed function-level documentation to all MCP query rules, query digest, static harvester, and catalog components. Static_Harvester.cpp: - Document all 18+ harvest functions (schemas, objects, columns, indexes, FKs, views) - Document lifecycle methods (init, close, connect, disconnect) - Document helper methods (is_time_type, is_id_like_name) - Document run management (start_run, finish_run, run_full_harvest) - Document statistics methods (get_harvest_stats) Query_Tool_Handler.cpp: - Document JSON helper functions (json_string, json_int, json_double) - Document digest tracking section with flow explanation MySQL_Catalog.cpp: - Document schema isolation architecture - Document CRUD operations (upsert, get, search, list, remove, merge) Discovery_Schema.cpp: - Document MCP query rules evaluation (evaluate_mcp_query_rules) - Document digest functions (compute_mcp_digest, fingerprint_mcp_args) - Document update/get functions for rules and digests ProxySQL_Admin_Stats.cpp: - Document stats collection functions ProxySQL_Admin.cpp: - Document load/save functions for query rules Admin_Handler.cpp: - Document MCP query rules command handlers include/ProxySQL_Admin_Tables_Definitions.h: - Add comments explaining table purposes	1 month ago
Rene Cannao	f01fc79584	feat: Add runtime_mcp_query_rules table and fix stats_mcp_query_rules schema ... - Add ADMIN_SQLITE_TABLE_RUNTIME_MCP_QUERY_RULES schema (17 columns, same as mcp_query_rules) - Fix STATS_SQLITE_TABLE_MCP_QUERY_RULES to only have rule_id and hits columns - Add runtime_mcp_query_rules detection and refresh in ProxySQL_Admin - Implement save_mcp_query_rules_from_runtime(bool _runtime) for both config and runtime tables - Update get_mcp_query_rules() to return 17 columns (no hits) - get_stats_mcp_query_rules() returns 2 columns (rule_id, hits) Mirrors the MySQL query rules pattern: - mcp_query_rules: config table (17 cols) - runtime_mcp_query_rules: runtime state (17 cols) - stats_mcp_query_rules: hit counters (2 cols)	1 month ago
Rene Cannao	f449c4236f	fix: Improve question learning fallback and error logging ... Two bug fixes for the question learning feature: 1. **Fallback to most recent agent_run across all schemas** - get_last_agent_run_id() now falls back to the most recent agent_run_id across ALL runs if none exists for the specific run_id - This allows adding questions even when the current schema's discovery didn't include an agent run - Adds logging to show when fallback is used 2. **Fix error message extraction for query_tool_calls logging** - Fixed bug where error messages weren't being extracted correctly - The old code checked for result["error"]["message"] but create_error_response only has result["error"] (no nested "message" field) - Now correctly extracts result["error"] as a string when present - This ensures failed tool calls are properly logged with error messages This fixes the issue where llm.question_template_add would fail with "No agent run found" even when agent runs exist for other schemas.	1 month ago
Rene Cannao	5b502c0864	feat: Add question learning capability to demo agent ... Add ability for the demo agent to learn new questions and add them to the catalog, making it smarter over time. Changes: - Added get_last_agent_run_id() function to Discovery_Schema: - Queries agent_runs table for the most recent agent_run_id for a run_id - Returns 0 if no agent runs exist for the schema - Updated llm.question_template_add handler: - Made agent_run_id optional (defaults to 0 when not provided) - When agent_run_id <= 0, auto-fetches last agent_run_id for the schema - Returns helpful error if no agent run exists for the schema - Returns agent_run_id in response for visibility - Updated llm.question_template_add tool schema: - Moved agent_run_id from required to optional parameters - Updated description to explain auto-fetch behavior - Updated demo_agent_claude.sh prompt: - Added llm.question_template_add to available tools - Added Step 4: "Learn from Success" to workflow - Added explicit instruction to ALWAYS LEARN new questions - Added example showing learning workflow - Expanded from 4 steps to 5 steps to include learning Now the demo agent can: 1. Search for existing questions 2. Reuse SQL if a good match exists 3. Generate new SQL if no good match 4. LEARN new questions by adding them to the catalog 5. Present results This enables continuous learning - the more users interact with it, the smarter it becomes.	1 month ago
Rene Cannao	ee74384c79	fix: Prevent llm.search from returning huge object lists in list mode ... When llm.search is called with an empty query (list mode) to retrieve all available questions, include_objects=true was returning full object schemas for all related objects, resulting in massive responses that could fill the LLM's context and cause rejections. Fix: include_objects now only works when query is non-empty (search mode). When query is empty (list mode), only question templates are returned without object details, regardless of include_objects setting. This makes semantic sense: - Empty query = "list all questions" → just titles/bodies (compact) - Non-empty query = "search for specific questions" → full details including object schemas (for answering the question) Changes: - Modified fts_search_llm() to check !query.empty() before fetching objects - Updated tool schema description to clarify this behavior	1 month ago
Rene Cannao	7e522aa2c0	feat: Add schema parameter to run_sql_readonly with per-connection tracking ... Add optional schema parameter to run_sql_readonly tool that allows queries to be executed against a specific schema, independent of the default schema configured in mcp-mysql_schema. Changes: - Added current_schema field to MySQLConnection structure to track the currently selected schema for each connection in the pool - Added find_connection() helper to find connection wrapper by mysql pointer - Added execute_query_with_schema() function that: - Uses mysql_select_db() instead of 'USE schema' SQL statement - Only calls mysql_select_db() if the requested schema differs from the current schema (optimization to avoid unnecessary switches) - Updates current_schema after successful schema switch - Updated run_sql_readonly handler: - Extracts optional 'schema' parameter - Calls execute_query_with_schema() instead of execute_query() - Returns error response when query fails (instead of success) - Updated tool schema to document the new 'schema' parameter This fixes the issue where queries would run against the default schema (configured in mcp-mysql_schema) instead of the schema being queried, causing "Table doesn't exist" errors when the default schema differs from the discovered schema.	1 month ago
Rene Cannao	ee13e4bf13	feat: Add include_objects parameter to llm_search for complete object retrieval ... Enhance the llm_search MCP tool to return complete question template data and optionally include full object schemas, reducing the need for additional MCP calls when answering questions. Changes: - Added related_objects column to llm_question_templates table - Updated add_question_template() to accept and store related_objects JSON array - Enhanced fts_search_llm() with include_objects parameter: - LEFT JOIN with llm_question_templates to return example_sql, related_objects, template_json, and confidence - When include_objects=true, fetches full object schemas (columns, indexes) for all related objects in a single batch operation - Added error checking for SQL execution failures - Fixed fts_search_llm() get_object() call to pass schema_name and object_name separately instead of combined object_key - Updated Query_Tool_Handler: - Added is_boolean() handling to json_int() helper to properly convert JSON boolean true/false to int 1/0 - Updated llm.search handler to extract and pass include_objects parameter - Updated llm.question_template_add to extract and pass related_objects - Updated tool schemas to document new parameters This change allows agents to get all necessary schema information in a single llm_search call instead of making multiple catalog_get_object calls, significantly reducing MCP call overhead.	1 month ago
Rene Cannao	1b42cfbd27	feat: Add empty query support to llm_search for listing all artifacts ... Changes: - fts_search_llm(): Empty query now returns all artifacts (list mode) - Update llm.search tool: query parameter is now optional - Tool description mentions empty query lists all artifacts - Add body field to llm_search results - Update demo script: Add special case for "What questions can I ask?" This enables agents to retrieve all pre-defined question templates when users ask what questions are available, instead of inferring questions from schema.	1 month ago
Rene Cannao	2250b762a3	feat: Add query_tool_calls table to log MCP tool invocations ... Add query_tool_calls table to Discovery Schema to track all MCP tool invocations via the /mcp/query/ endpoint. Logs: - tool_name: Name of the tool that was called - schema: Schema name (nullable, empty if not applicable) - run_id: Run ID from discovery (nullable, 0 if not applicable) - start_time: Start monotonic time in microseconds - execution_time: Execution duration in microseconds - error: Error message (null if success) Modified files: - Discovery_Schema.cpp: Added table creation and log_query_tool_call function - Discovery_Schema.h: Added function declaration - Query_Tool_Handler.cpp: Added logging after each tool execution	1 month ago
Rene Cannao	77643859e3	feat: Add timing columns to stats_mcp_query_tools_counters ... Extend the stats_mcp_query_tools_counters table with timing statistics (first_seen, last_seen, sum_time, min_time, max_time) following the same pattern as stats_mysql_query_digest. All timing values are in microseconds using monotonic_time(). New schema: - tool VARCHAR - schema VARCHAR - count INT - first_seen INTEGER (microseconds) - last_seen INTEGER (microseconds) - sum_time INTEGER (microseconds - total execution time) - min_time INTEGER (microseconds - minimum execution time) - max_time INTEGER (microseconds - maximum execution time)	1 month ago
Rene Cannao	a816a756d4	feat: Add MCP query tool usage counters to stats schema ... Add stats_mcp_query_tools_counters and stats_mcp_query_tools_counters_reset tables to track MCP query tool usage statistics. - Added get_tool_usage_stats_resultset() method to Query_Tool_Handler - Defined table schemas in ProxySQL_Admin_Tables_Definitions.h - Registered tables in Admin_Bootstrap.cpp - Added pattern matching in ProxySQL_Admin.cpp - Added stats___mcp_query_tools_counters() in ProxySQL_Admin_Stats.cpp - Fixed friend declaration for track_tool_invocation() - Fixed Discovery_Schema.cpp log_llm_search() to use prepare_v2/finalize	1 month ago
Rene Cannao	393967f511	fix: Use row->cnt instead of row->fields_count	1 month ago
Rene Cannao	df0527c044	refactor: list_schemas to use catalog instead of live database ... - Query schemas from catalog's schemas table - Maintains same output format for compatibility - Removes dependency on live MySQL connection	1 month ago
Rene Cannao	527a748d16	refactor: Remove describe_table tool completely ... Tool was deprecated; users should use catalog.get_object instead.	1 month ago
Rene Cannao	623675b369	feat: Add schema name resolver and deprecate direct DB tools ... - Add resolve_run_id() to map schema names to latest run_id - Update all catalog and LLM tools to accept schema names - Deprecate describe_table, table_profile, column_profile - Deprecate get_constraints, suggest_joins, find_reference_candidates - Keep sample_rows, sample_distinct for data preview	1 month ago
Rene Cannao	757cdaff15	fix: Improve error logging and fix llm.domain_set_members ... 1. Fix error logging to catch ALL tool failures, not just those with both success and result fields. Previously, error responses like {"success": false, "error": "..."} without a result field were silently ignored. 2. Fix llm.domain_set_members to accept both array and JSON string formats for the members parameter. Some clients send it as a JSON string, others as a native array. 3. Add detailed error logging for llm.domain_set_members failures, including what was actually received.	1 month ago
Rene Cannao	53ecda7730	fix: Add comprehensive error handling and logging for MCP tools ... - Add try-catch around handle_jsonrpc_request to catch unexpected exceptions - Add detailed logging for tool execution success/failure - Add proper SQLite error checking in create_agent_run with error messages - Fix json_int/json_double to handle both numbers and numeric strings The json_int function was throwing exceptions when receiving numeric strings (e.g., "14" instead of 14) from clients, causing 500 errors. Now it handles both formats gracefully. Also added logging so tool failures are visible in logs instead of being silent 500 errors.	1 month ago
Rene Cannao	1b7335acfe	Fix two-phase discovery documentation and scripts ... - Add mcp_config.example.json for Claude Code MCP configuration - Fix MCP bridge path in example config (../../proxysql_mcp_stdio_bridge.py) - Update Two_Phase_Discovery_Implementation.md with correct Phase 1/Phase 2 usage - Fix Two_Phase_Discovery_Implementation.md DELETE FROM fts_objects to scope to run_id - Update README.md with two-phase discovery section and multi-agent legacy note - Create static_harvest.sh bash wrapper for Phase 1 - Create two_phase_discovery.py orchestration script with prompts - Add --run-id parameter to skip auto-fetch - Fix RUN_ID placeholder mismatch (<USE_THE_PROVIDED_RUN_ID>) - Fix catalog path default to mcp_catalog.db - Add test_catalog.sh to verify catalog tools work - Fix Discovery_Schema.cpp FTS5 syntax (missing space) - Remove invalid CREATE INDEX on FTS virtual tables - Add MCP tool call logging to track tool usage - Fix Static_Harvester::get_harvest_stats() to accept run_id parameter - Fix DELETE FROM fts_objects to only delete for specific run_id - Update system prompts to say DO NOT call discovery.run_static - Update user prompts to say Phase 1 is already complete - Add --mcp-only flag to restrict Claude Code to MCP tools only - Make FTS table failures non-fatal (check if table exists first) - Add comprehensive documentation for both discovery approaches	1 month ago
Rene Cannao	6f23d5bcd0	feat: Implement two-phase schema discovery architecture ... Phase 1 (Static/Deterministic): - Add Discovery_Schema: SQLite catalog with deterministic and LLM tables - Add Static_Harvester: MySQL INFORMATION_SCHEMA metadata extraction - Harvest schemas, objects, columns, indexes, foreign keys, view definitions - Compute derived hints: is_time, is_id_like, has_pk, has_fks, has_time - Build quick profiles and FTS5 indexes Phase 2 (LLM Agent): - Add 19 new MCP tools for two-phase discovery - discovery.run_static: Trigger ProxySQL's static harvest - Catalog tools: init, search, get_object, list_objects, get_relationships - Agent tools: run_start, run_finish, event_append - LLM tools: summary_upsert, relationship_upsert, domain_upsert, etc. Files: - include/Discovery_Schema.h, lib/Discovery_Schema.cpp - include/Static_Harvester.h, lib/Static_Harvester.cpp - include/Query_Tool_Handler.h, lib/Query_Tool_Handler.cpp (updated) - lib/Makefile (updated) - scripts/mcp/DiscoveryAgent/ClaudeCode_Headless/prompts/ - scripts/mcp/DiscoveryAgent/ClaudeCode_Headless/two_phase_discovery.py	1 month ago
Rene Cannao	7de3f0c510	feat: Add schema separation to MCP catalog and discovery scope constraint ... This commit addresses two issues: 1. MCP Catalog Schema Separation: - Add 'schema' column to catalog table for proper isolation - Update all catalog methods (upsert, get, search, list, remove) to accept schema parameter - Update MCP tool handlers and JSON-RPC parameter parsing - Unique constraint changed from (kind, key) to (schema, kind, key) - FTS table updated to include schema column 2. Discovery Prompt Scope Constraint: - Add explicit SCOPE CONSTRAINT section to multi_agent_discovery_prompt.md - Agents now respect Target Schema and skip list_schemas when specified - Prevents analyzing all schemas when only one is targeted Files modified: - include/MySQL_Catalog.h: Add schema parameter to all catalog methods - include/MySQL_Tool_Handler.h: Update wrapper method signatures - lib/MySQL_Catalog.cpp: Implement schema filtering in all operations - lib/MySQL_Tool_Handler.cpp: Update wrapper implementations - lib/Query_Tool_Handler.cpp: Extract schema from JSON-RPC params, update tool descriptions - scripts/mcp/DiscoveryAgent/ClaudeCode_Headless/prompts/multi_agent_discovery_prompt.md: Add scope constraint	1 month ago
Rene Cannao	ef5b99edbf	Fix MCP tool bugs: NULL value handling and query validation ... - Fixed NULL value handling in execute_query: use empty string instead of nullptr to avoid "basic_string: construction from null" errors - Fixed validate_readonly_query: corrected substring length check from substr(0,6)!="SELECT " to substr(0,6)!="SELECT" - Fixed test script: added proper variable_name parameter for get_config/set_config tools Query endpoint tools now pass all tests.	1 month ago
Rene Cannao	22db1a5fdd	Fix JSON value extraction in Query_Tool_Handler::execute_tool ... The nlohmann::json value() method can throw "basic_string: construction from null is not valid" when trying to convert a JSON null value to std::string. Added helper functions get_json_string() and get_json_int() that: - Check if key exists before accessing - Check if value is not null - Check if value has correct type - Return default value if any check fails This prevents crashes when: 1. Arguments are missing (returns default) 2. Arguments are explicitly null (returns default) 3. Arguments have wrong type (returns default)	1 month ago
Rene Cannao	c86a048d9c	Implement MCP multi-endpoint architecture with dedicated tool handlers ... This commit implements Option 1 (Multiple Tool Handlers) for the MCP module, where each of the 5 endpoints has its own dedicated tool handler with specific tools. ## Architecture Changes - Created MCP_Tool_Handler base class interface for all tool handlers - Each endpoint now has its own dedicated tool handler: - /mcp/config → Config_Tool_Handler (configuration management) - /mcp/query → Query_Tool_Handler (database exploration) - /mcp/admin → Admin_Tool_Handler (administrative operations) - /mcp/cache → Cache_Tool_Handler (cache management) - /mcp/observe → Observe_Tool_Handler (monitoring & metrics) ## New Files Base Interface: - include/MCP_Tool_Handler.h - Base class for all tool handlers Tool Handlers: - include/Config_Tool_Handler.h, lib/Config_Tool_Handler.cpp - include/Query_Tool_Handler.h, lib/Query_Tool_Handler.cpp - include/Admin_Tool_Handler.h, lib/Admin_Tool_Handler.cpp - include/Cache_Tool_Handler.h, lib/Cache_Tool_Handler.cpp - include/Observe_Tool_Handler.h, lib/Observe_Tool_Handler.cpp Documentation: - doc/MCP/Architecture.md - Comprehensive architecture documentation ## Modified Files - include/MCP_Thread.h, lib/MCP_Thread.cpp - Added 5 tool handler pointers - include/MCP_Endpoint.h, lib/MCP_Endpoint.cpp - Use tool_handler base class - lib/ProxySQL_MCP_Server.cpp - Create and pass handlers to endpoints - lib/Makefile - Added new source files ## Implementation Status - Config_Tool_Handler: Functional (get_config, set_config, list_variables, get_status) - Query_Tool_Handler: Functional (wraps MySQL_Tool_Handler, all 18 tools) - Admin_Tool_Handler: Stub implementations (TODO: implement) - Cache_Tool_Handler: Stub implementations (TODO: implement) - Observe_Tool_Handler: Stub implementations (TODO: implement) See GitHub Issue #8 for detailed TODO list. Co-authored-by: Claude <claude@anthropic.com>	1 month ago