Merge branch 'sinaptik-ai:main' into template_improvements

Author: matteocacciola

pandasai/agent/base.py

@@ -85,7 +85,10 @@ def chat(self, query: str, output_type: Optional[str] = None):
         Start a new chat interaction with the assistant on Dataframe.
         """
         if self._state.config.llm is None:
-            raise ValueError("No LLM provided. Please provide an LLM to the agent.")
+            raise ValueError(
+                "PandasAI API key does not include LLM credits. Please configure an OpenAI or LiteLLM key. "
+                "Learn more at: https://docs.pandas-ai.com/v3/large-language-models#how-to-set-up-any-llm%3F"
+            )
 
         self.start_new_conversation()
         return self._process_query(query, output_type)

pandasai/core/code_generation/code_cleaning.py

@@ -55,7 +55,8 @@ def _clean_sql_query(self, sql_query: str) -> str:
         Clean the SQL query by trimming semicolons and validating table names.
         """
         sql_query = sql_query.rstrip(";")
-        table_names = SQLParser.extract_table_names(sql_query)
+        dialect = self.context.dfs[0].get_dialect()
+        table_names = SQLParser.extract_table_names(sql_query, dialect)
         allowed_table_names = {
             df.schema.name: df.schema.name for df in self.context.dfs
         } | {f'"{df.schema.name}"': df.schema.name for df in self.context.dfs}

pandasai/dataframe/base.py

@@ -132,20 +132,23 @@ def rows_count(self) -> int:
     def columns_count(self) -> int:
         return len(self.columns)
 
+    def get_dialect(self):
+        source = self.schema.source or None
+        if source:
+            dialect = "duckdb" if source.type in LOCAL_SOURCE_TYPES else source.type
+        else:
+            dialect = "postgres"
+
+        return dialect
+
     def serialize_dataframe(self) -> str:
         """
         Serialize DataFrame to string representation.
 
         Returns:
             str: Serialized string representation of the DataFrame
         """
-        source = self.schema.source or None
-
-        if source:
-            dialect = "duckdb" if source.type in LOCAL_SOURCE_TYPES else source.type
-        else:
-            dialect = "postgres"
-
+        dialect = self.get_dialect()
         return DataframeSerializer.serialize(self, dialect)
 
     def get_head(self):

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "pandasai"
-version = "3.0.0-beta.18"
+version = "3.0.0-beta.19"
 description = "Chat with your database (SQL, CSV, pandas, mongodb, noSQL, etc). PandasAI makes data analysis conversational using LLMs (GPT 3.5 / 4, Anthropic, VertexAI) and RAG."
 authors = ["Gabriele Venturi"]
 license = "MIT"

tests/unit_tests/core/code_generation/test_code_cleaning.py

@@ -72,6 +72,7 @@ def test_clean_sql_query(self):
         mock_dataframe.schema = MagicMock()
         mock_dataframe.schema.name = "my_table"
         self.cleaner.context.dfs = [mock_dataframe]
+        mock_dataframe.get_dialect = MagicMock(return_value="duckdb")
         result = self.cleaner._clean_sql_query(sql_query)
         self.assertEqual(result, "SELECT * FROM my_table")
 
@@ -85,6 +86,7 @@ def test_validate_and_make_table_name_case_sensitive(self):
         self.cleaner.context.dfs = [mock_dataframe]
         mock_dataframe.schema = MagicMock()
         mock_dataframe.schema.name = "my_table"
+        mock_dataframe.get_dialect = MagicMock(return_value="duckdb")
         updated_node = self.cleaner._validate_and_make_table_name_case_sensitive(node)
         self.assertEqual(updated_node.value.value, "SELECT * FROM my_table")