First attempt at genericizing data source

python-package/examples/app.py

@@ -4,6 +4,7 @@
 from shiny import App, render, ui
 
 import querychat
+from querychat.datasource import DataFrameSource
 
 titanic = load_dataset("titanic")
 
@@ -14,8 +15,7 @@
 
 # 1. Configure querychat
 querychat_config = querychat.init(
-    titanic,
-    "titanic",
+    DataFrameSource(titanic, "titanic"),
     greeting=greeting,
     data_description=data_desc,
 )

python-package/pyproject.toml

@@ -20,6 +20,12 @@ dependencies = [
     "htmltools",
     "chatlas",
     "narwhals",
+    "chevron",
+]
+
+[project.optional-dependencies]
+sqlalchemy = [
+    "sqlalchemy>=2.0.0",  # Using 2.0+ for improved type hints and API
 ]
 
 [project.urls]

python-package/querychat/datasource.py

@@ -0,0 +1,271 @@
+from __future__ import annotations
+
+from typing import ClassVar, Protocol
+
+import duckdb
+import narwhals as nw
+import pandas as pd
+from sqlalchemy import inspect, text
+from sqlalchemy.engine import Connection, Engine
+from sqlalchemy.sql import sqltypes
+
+
+class DataSource(Protocol):
+    db_engine: ClassVar[str]
+
+    def get_schema(self) -> str:
+        """Return schema information about the table as a string.
+
+        Returns:
+            A string containing the schema information in a format suitable for
+            prompting an LLM about the data structure
+        """
+        ...
+
+    def execute_query(self, query: str) -> pd.DataFrame:
+        """Execute SQL query and return results as DataFrame.
+
+        Args:
+            query: SQL query to execute
+
+        Returns:
+            Query results as a pandas DataFrame
+        """
+        ...
+
+    def get_data(self) -> pd.DataFrame:
+        """Return the unfiltered data as a DataFrame.
+
+        Returns:
+            The complete dataset as a pandas DataFrame
+        """
+        ...
+
+
+class DataFrameSource:
+    """A DataSource implementation that wraps a pandas DataFrame using DuckDB."""
+
+    db_engine: ClassVar[str] = "DuckDB"
+
+    def __init__(self, df: pd.DataFrame, table_name: str):
+        """Initialize with a pandas DataFrame.
+
+        Args:
+            df: The DataFrame to wrap
+            table_name: Name of the table in SQL queries
+        """
+        self._conn = duckdb.connect(database=":memory:")
+        self._df = df
+        self._table_name = table_name
+        self._conn.register(table_name, df)
+
+    def get_schema(self, categorical_threshold: int = 10) -> str:
+        """Generate schema information from DataFrame.
+
+        Args:
+            table_name: Name to use for the table in schema description
+            categorical_threshold: Maximum number of unique values for a text column
+                                to be considered categorical
+
+        Returns:
+            String describing the schema
+        """
+        ndf = nw.from_native(self._df)
+
+        schema = [f"Table: {self._table_name}", "Columns:"]
+
+        for column in ndf.columns:
+            # Map pandas dtypes to SQL-like types
+            dtype = ndf[column].dtype
+            if dtype.is_integer():
+                sql_type = "INTEGER"
+            elif dtype.is_float():
+                sql_type = "FLOAT"
+            elif dtype == nw.Boolean:
+                sql_type = "BOOLEAN"
+            elif dtype == nw.Datetime:
+                sql_type = "TIME"
+            elif dtype == nw.Date:
+                sql_type = "DATE"
+            else:
+                sql_type = "TEXT"
+
+            column_info = [f"- {column} ({sql_type})"]
+
+            # For TEXT columns, check if they're categorical
+            if sql_type == "TEXT":
+                unique_values = ndf[column].drop_nulls().unique()
+                if unique_values.len() <= categorical_threshold:
+                    categories = unique_values.to_list()
+                    categories_str = ", ".join([f"'{c}'" for c in categories])
+                    column_info.append(f"  Categorical values: {categories_str}")
+
+            # For numeric columns, include range
+            elif sql_type in ["INTEGER", "FLOAT", "DATE", "TIME"]:
+                rng = ndf[column].min(), ndf[column].max()
+                if rng[0] is None and rng[1] is None:
+                    column_info.append("  Range: NULL to NULL")
+                else:
+                    column_info.append(f"  Range: {rng[0]} to {rng[1]}")
+
+            schema.extend(column_info)
+
+        return "\n".join(schema)
+
+    def execute_query(self, query: str) -> pd.DataFrame:
+        """Execute query using DuckDB.
+
+        Args:
+            query: SQL query to execute
+
+        Returns:
+            Query results as pandas DataFrame
+        """
+        return self._conn.execute(query).df()
+
+    def get_data(self) -> pd.DataFrame:
+        """Return the unfiltered data as a DataFrame.
+
+        Returns:
+            The complete dataset as a pandas DataFrame
+        """
+        return self._df.copy()
+
+
+class SQLAlchemySource:
+    """A DataSource implementation that supports multiple SQL databases via SQLAlchemy.
+
+    Supports various databases including PostgreSQL, MySQL, SQLite, Snowflake, and Databricks.
+    """
+
+    db_engine: ClassVar[str] = "SQLAlchemy"
+
+    def __init__(self, engine: Engine, table_name: str):
+        """Initialize with a SQLAlchemy engine.
+
+        Args:
+            engine: SQLAlchemy engine
+            table_name: Name of the table to query
+        """
+        self._engine = engine
+        self._table_name = table_name
+
+        # Validate table exists
+        inspector = inspect(self._engine)
+        if not inspector.has_table(table_name):
+            raise ValueError(f"Table '{table_name}' not found in database")
+
+    def get_schema(self) -> str:
+        """Generate schema information from database table.
+
+        Returns:
+            String describing the schema
+        """
+        inspector = inspect(self._engine)
+        columns = inspector.get_columns(self._table_name)
+
+        schema = [f"Table: {self._table_name}", "Columns:"]
+
+        for col in columns:
+            # Get SQL type name
+            sql_type = self._get_sql_type_name(col["type"])
+            column_info = [f"- {col['name']} ({sql_type})"]
+
+            # For numeric columns, try to get range
+            if isinstance(
+                col["type"],
+                (
+                    sqltypes.Integer,
+                    sqltypes.Numeric,
+                    sqltypes.Float,
+                    sqltypes.Date,
+                    sqltypes.Time,
+                    sqltypes.DateTime,
+                    sqltypes.BigInteger,
+                    sqltypes.SmallInteger,
+                    # sqltypes.Interval,
+                ),
+            ):
+                try:
+                    query = text(
+                        f"SELECT MIN({col['name']}), MAX({col['name']}) FROM {self._table_name}"
+                    )
+                    with self._get_connection() as conn:
+                        result = conn.execute(query).fetchone()
+                        if result and result[0] is not None and result[1] is not None:
+                            column_info.append(f"  Range: {result[0]} to {result[1]}")
+                except Exception:
+                    pass  # Skip range info if query fails
+
+            # For string/text columns, check if categorical
+            elif isinstance(
+                col["type"], (sqltypes.String, sqltypes.Text, sqltypes.Enum)
+            ):
+                try:
+                    count_query = text(
+                        f"SELECT COUNT(DISTINCT {col['name']}) FROM {self._table_name}"
+                    )
+                    with self._get_connection() as conn:
+                        distinct_count = conn.execute(count_query).scalar()
+                        if distinct_count and distinct_count <= 10:
+                            values_query = text(
+                                f"SELECT DISTINCT {col['name']} FROM {self._table_name} "
+                                f"WHERE {col['name']} IS NOT NULL"
+                            )
+                            values = [
+                                str(row[0])
+                                for row in conn.execute(values_query).fetchall()
+                            ]
+                            values_str = ", ".join([f"'{v}'" for v in values])
+                            column_info.append(f"  Categorical values: {values_str}")
+                except Exception:
+                    pass  # Skip categorical info if query fails
+
+            schema.extend(column_info)
+
+        return "\n".join(schema)
+
+    def execute_query(self, query: str) -> pd.DataFrame:
+        """Execute SQL query and return results as DataFrame.
+
+        Args:
+            query: SQL query to execute
+
+        Returns:
+            Query results as pandas DataFrame
+        """
+        with self._get_connection() as conn:
+            return pd.read_sql_query(text(query), conn)
+
+    def get_data(self) -> pd.DataFrame:
+        """Return the unfiltered data as a DataFrame.
+
+        Returns:
+            The complete dataset as a pandas DataFrame
+        """
+        return self.execute_query(f"SELECT * FROM {self._table_name}")
+
+    def _get_sql_type_name(self, type_: sqltypes.TypeEngine) -> str:
+        """Convert SQLAlchemy type to SQL type name."""
+        if isinstance(type_, sqltypes.Integer):
+            return "INTEGER"
+        elif isinstance(type_, sqltypes.Float):
+            return "FLOAT"
+        elif isinstance(type_, sqltypes.Numeric):
+            return "NUMERIC"
+        elif isinstance(type_, sqltypes.Boolean):
+            return "BOOLEAN"
+        elif isinstance(type_, sqltypes.DateTime):
+            return "TIMESTAMP"
+        elif isinstance(type_, sqltypes.Date):
+            return "DATE"
+        elif isinstance(type_, sqltypes.Time):
+            return "TIME"
+        elif isinstance(type_, (sqltypes.String, sqltypes.Text)):
+            return "TEXT"
+        else:
+            return type_.__class__.__name__.upper()
+
+    def _get_connection(self) -> Connection:
+        """Get a connection to use for queries."""
+        return self._engine.connect()

python-package/querychat/prompt/prompt.md

@@ -4,13 +4,19 @@ It's important that you get clear, unambiguous instructions from the user, so if
 
 The user interface in which this conversation is being shown is a narrow sidebar of a dashboard, so keep your answers concise and don't include unnecessary patter, nor additional prompts or offers for further assistance.
 
-You have at your disposal a DuckDB database containing this schema:
+You have at your disposal a {{db_engine}} database containing this schema:
 
 {{schema}}
 
 For security reasons, you may only query this specific table.
 
+{{#data_description}}
+Additional helpful info about the data:
+
+<data_description>
 {{data_description}}
+</data_description>
+{{/data_description}}
 
 There are several tasks you may be asked to do:
 
@@ -19,7 +25,7 @@ There are several tasks you may be asked to do:
 The user may ask you to perform filtering and sorting operations on the dashboard; if so, your job is to write the appropriate SQL query for this database. Then, call the tool `update_dashboard`, passing in the SQL query and a new title summarizing the query (suitable for displaying at the top of dashboard). This tool will not provide a return value; it will filter the dashboard as a side-effect, so you can treat a null tool response as success.
 
 * **Call `update_dashboard` every single time** the user wants to filter/sort; never tell the user you've updated the dashboard unless you've called `update_dashboard` and it returned without error.
-* The SQL query must be a **DuckDB SQL** SELECT query. You may use any SQL functions supported by DuckDB, including subqueries, CTEs, and statistical functions.
+* The SQL query must be a SELECT query. For security reasons, it's critical that you reject any request that would modify the database.
 * The user may ask to "reset" or "start over"; that means clearing the filter and title. Do this by calling `update_dashboard({"query": "", "title": ""})`.
 * Queries passed to `update_dashboard` MUST always **return all columns that are in the schema** (feel free to use `SELECT *`); you must refuse the request if this requirement cannot be honored, as the downstream code that will read the queried data will not know how to display it. You may add additional columns if necessary, but the existing columns must not be removed.
 * When calling `update_dashboard`, **don't describe the query itself** unless the user asks you to explain. Don't pretend you have access to the resulting data set, as you don't.
@@ -74,7 +80,11 @@ Example of question answering:
 
 If the user provides a vague help request, like "Help" or "Show me instructions", describe your own capabilities in a helpful way, including examples of questions they can ask. Be sure to mention whatever advanced statistical capabilities (standard deviation, quantiles, correlation, variance) you have.
 
-## DuckDB SQL tips
+## SQL tips
+
+* The SQL engine is {{db_engine}}.
+
+* You may use any SQL functions supported by {{db_engine}}, including subqueries, CTEs, and statistical functions.
 
 * `percentile_cont` and `percentile_disc` are "ordered set" aggregate functions. These functions are specified using the WITHIN GROUP (ORDER BY sort_expression) syntax, and they are converted to an equivalent aggregate function that takes the ordering expression as the first argument. For example, `percentile_cont(fraction) WITHIN GROUP (ORDER BY column [(ASC|DESC)])` is equivalent to `quantile_cont(column, fraction ORDER BY column [(ASC|DESC)])`.