v1.0.0

First stable release of querywise-mcp — an MCP server and CLI for querying databases in natural language through a business semantic layer.

Highlights

• 25 MCP tools with full parameter descriptions and behavioral annotations (read-only / destructive hints), plus a schema resource and a text_to_sql prompt.
• Two query paths: the lightweight get_semantic_context + run_sql (no LLM key needed) and the automated ask / generate_sql server-side NL→SQL pipeline.
• Semantic layer: business glossary, metrics, value dictionaries, sample queries, and importable knowledge documents.
• Embedded metadata store: SQLite + sqlite-vec (no external Postgres/pgvector).
• Connectors: SQLite (read-only), PostgreSQL, BigQuery, Databricks. Read-only enforced via SQL sanitizer + connector-level guards.
• Provider-agnostic LLM (Anthropic / OpenAI / Ollama), optional — keyword-only operation works with no key.
• Zero-config startup, stdio + Streamable HTTP transports, Docker image, and a Glama container release.

Install

pip install -e ".[llm]"
querywise init
querywise serve