feat(loaders): add Microsoft SQL Server support

.github/wordlist.txt

@@ -124,6 +124,14 @@ SDK
 Dependabot
 PyPI
 pypi
+pymssql
+sqlserver
+SQLServerLoader
+dbo
+tsql
+hostname
+TLS
+sqlglot
 signup
 SMTP
 outbox

README.md

@@ -335,6 +335,10 @@ pip install queryweaver[server]
 pip install queryweaver[dev]
 ```
 
+> **SQL Server and Snowflake need `queryweaver[server]`.** Their drivers
+> (`pymssql` and `snowflake-connector-python`) ship in the `server` extra, so the
+> minimal SDK install can only connect to PostgreSQL and MySQL.
+
 ### Quick Start
 
 ```python
@@ -345,7 +349,7 @@ async def main():
     # Initialize with FalkorDB connection
     qw = QueryWeaver(falkordb_url="redis://localhost:6379")
 
-    # Connect a PostgreSQL or MySQL database
+    # Connect a PostgreSQL, MySQL, SQL Server or Snowflake database
     conn = await qw.connect_database("postgresql://user:pass@host:5432/mydb")
     print(f"Connected: {conn.database_id}")  # "mydb"
 
@@ -389,7 +393,7 @@ async with QueryWeaver(falkordb_url="redis://host-a:6379", user_id="tenant_a") a
 
 | Method | Description |
 |--------|-------------|
-| `connect_database(db_url)` | Connect PostgreSQL/MySQL and load schema |
+| `connect_database(db_url)` | Connect PostgreSQL/MySQL/SQL Server/Snowflake and load schema (SQL Server and Snowflake require `queryweaver[server]`) |
 | `query(database, question)` | Convert natural language to SQL and execute |
 | `get_schema(database)` | Retrieve database schema (tables and relationships) |
 | `list_databases()` | List all connected databases |
@@ -435,7 +439,8 @@ if result.requires_confirmation:
 - Python 3.12+
 - FalkorDB instance (local or remote)
 - OpenAI or Azure OpenAI API key (for LLM)
-- Target SQL database (PostgreSQL or MySQL)
+- Target SQL database (PostgreSQL, MySQL, SQL Server or Snowflake — the last two
+  require the `queryweaver[server]` extra)
 
 ## Development
 

api/core/pipeline.py

@@ -115,8 +115,9 @@ def get_database_type_and_loader(
     PostgreSQL for backward compatibility on the server path.
 
     When ``sdk_only`` is True, raises ``InvalidArgumentError`` for vendors
-    that need the ``[server]`` extra (snowflake) or for unknown URL schemes,
-    so SDK callers get a clean error instead of a deferred ``ImportError``.
+    that need the ``[server]`` extra (snowflake, sqlserver) or for unknown URL
+    schemes, so SDK callers get a clean error instead of a deferred
+    ``ImportError``.
     """
     if not db_url or db_url == "No URL available for this database.":
         return None, None
@@ -138,6 +139,17 @@ def get_database_type_and_loader(
         # pylint: disable=import-outside-toplevel
         from api.loaders.snowflake_loader import SnowflakeLoader
         return 'snowflake', SnowflakeLoader
+    if db_url_lower.startswith('sqlserver://'):
+        if sdk_only:
+            raise InvalidArgumentError(
+                "SQL Server requires the [server] extra: "
+                "pip install queryweaver[server]"
+            )
+        # Lazy-import: pymssql is in the [server] extra, not in the core SDK
+        # install.
+        # pylint: disable=import-outside-toplevel
+        from api.loaders.sqlserver_loader import SQLServerLoader
+        return 'sqlserver', SQLServerLoader
 
     if sdk_only:
         raise InvalidArgumentError(
@@ -205,6 +217,7 @@ def truncate_for_log(query: str, max_length: int = 200) -> str:
     "postgres": "postgres",
     "mysql": "mysql",
     "snowflake": "snowflake",
+    "sqlserver": "tsql",
 }
 
 # sqlglot expression class names that represent a write, DDL, privilege change,

api/core/schema_loader.py

@@ -32,7 +32,9 @@ def _step_start(steps_counter: int) -> dict[str, str]:
         "message": f"Step {steps_counter}: Starting database connection",
     }
 
-_KNOWN_DB_SCHEMES = ("postgresql://", "postgres://", "mysql://", "snowflake://")
+_KNOWN_DB_SCHEMES = (
+    "postgresql://", "postgres://", "mysql://", "snowflake://", "sqlserver://",
+)
 
 
 def _step_detect_db_type(steps_counter: int, url: str) -> tuple[type[BaseLoader], dict[str, str]]: