metadata attribute of the response object. Here's an example of how you can do this:response = agent.chat("Your query here")
print(response.metadata)PandasQueryEngine, the metadata might include the pandas python instructions used to generate the response:# get pandas python instructions print(response.metadata["pandas_instruction_str"])
QueryEngineTool objects, which is different from the metadata of the response. The metadata of the QueryEngineTool objects is static and set when you create the QueryEngineTool objects, as you've done in your code. This metadata is not automatically included in the response metadata.response.metadata, it means that the specific tool or query engine you're using is not providing any response metadata.QueryEngineTool objects in the response metadata, you would need to modify the QueryEngineTool or the query engine to include this information in the response metadata. This would likely involve modifying the query method of the query engine to include the desired metadata in the response.QueryEngineTool and the query engines, I can't provide specific instructions on how to do this. You might want to refer to the LlamaIndex documentation () or the source code of the QueryEngineTool and the query engines for more information.QueryEngineTool that was used to generate a response, you can modify the call method of the QueryEngineTool to include the tool's metadata in the response. Here's an example of how you can do this:class MyQueryEngineTool(QueryEngineTool):
def call(self, *args: Any, **kwargs: Any) -> ToolOutput:
query_str = self._get_query_str(*args, **kwargs)
response = self._query_engine.query(query_str)
return ToolOutput(
content=str(response),
tool_name=self.metadata.name,
raw_input={"input": query_str},
raw_output=response,
tool_metadata=self.metadata.dict(), # include tool metadata in the response
)MyQueryEngineTool is a subclass of QueryEngineTool that overrides the call method to include the tool's metadata in the response. The tool's metadata is included in the tool_metadata field of the ToolOutput object.MyQueryEngineTool instead of QueryEngineTool when defining your query engine tools:query_engine_tools = [
MyQueryEngineTool(
query_engine=ppm_query_engine,
metadata=ToolMetadata(
name="ppm_plan_view",
description=(
"Provides information about Ppm plan view data."
"Don't use your prior knowledge in answering a question"
),
),
),
# ...
]response = agent.chat("Your query here")
print(response.tool_metadata)QueryEngineTool, which might not be possible or desirable in all situations. Also, this solution assumes that you're using a version of LlamaIndex where the ToolOutput class has a tool_metadata field. If this is not the case, you might need to modify the ToolOutput class as well.PandasQueryEngine was used to generate a response, you can modify the query method of the PandasQueryEngine to include this information in the response metadata. Here's an example of how you can do this:class MyPandasQueryEngine(PandasQueryEngine):
def __init__(self, df, name, **kwargs):
super().__init__(df, **kwargs)
self.name = name
def query(self, query_str):
response = super().query(query_str)
# Add the name of the query engine to the response metadata
response.metadata["query_engine_name"] = self.name
return responseMyPandasQueryEngine is a subclass of PandasQueryEngine that adds a name attribute to the query engine and includes this name in the response metadata. You can then use MyPandasQueryEngine instead of PandasQueryEngine when defining your query engines:ppm_query_engine = MyPandasQueryEngine(df=ppm_df, name="ppm_plan_view", verbose=True) # ...
response = agent.chat("Your query here")
print(response.metadata["query_engine_name"])PandasQueryEngine, which might not be possible or desirable in all situations. Also, this solution assumes that you're using a version of LlamaIndex where the Response class has a metadata field. If this is not the case, you might need to modify the Response class as well.