Is that possible to add SubQuestionQueryEngine to Pandas Query Pipeline?

[mraguth]

Is it possible to add a subquestion query engine to a query pipeline? In my example, I'm using a pandas query pipeline. I know how to add SubQuestionQueryEngine to a regular query by adding it to the query engine tool. However, I'm unsure how to do this for the pandas query pipeline I posted below. Please help.

My pandas query pipeline sample:

qp = QP(
    modules={
        "input": InputComponent(),
        "pandas_prompt": pandas_prompt,
        "llm1": llm,
        "pandas_output_parser": pandas_output_parser,
        "response_synthesis_prompt": response_synthesis_prompt,
        "llm2": llm,
    },
    verbose=True,
)
qp.add_chain(["input", "pandas_prompt", "llm1", "pandas_output_parser"])
qp.add_links(
    [
        Link("input", "response_synthesis_prompt", dest_key="query_str"),
        Link(
            "llm1", "response_synthesis_prompt", dest_key="pandas_instructions"
        ),
        Link(
            "pandas_output_parser",
            "response_synthesis_prompt",
            dest_key="pandas_output",
        ),
    ]
)
# add link from response synthesis prompt to llm2
qp.add_link("response_synthesis_prompt", "llm2")

This is the way I used to add my subquestionqueryengine to regular query engines:

 # setup base query engine as tool
query_engine_tools = [
    QueryEngineTool(
        query_engine=vector_query_engine,
        metadata=ToolMetadata(
            name="pg_essay",
            description="Paul Graham essay on What I Worked On",
        ),
    ),
]

query_engine = SubQuestionQueryEngine.from_defaults(
    query_engine_tools=query_engine_tools,
    use_async=True,
)

[dosubot[bot]]

Yes, it is possible to add a SubQuestionQueryEngine to your Pandas query pipeline. You can integrate the SubQuestionQueryEngine by creating it with the appropriate parameters and then adding it to your pipeline. Here's how you can modify your existing Pandas query pipeline to include the SubQuestionQueryEngine:

1. Instantiate the SubQuestionQueryEngine with the necessary components.
2. Add the SubQuestionQueryEngine to your pipeline modules.
3. Link the SubQuestionQueryEngine appropriately within the pipeline.

Here's an example of how you can do this:

from llama_index.core.query_pipeline import QueryPipeline as QP, InputComponent, Link
from llama_index.query_engine.sub_question_query_engine import SubQuestionQueryEngine
from llama_index.query_engine.query_engine_tool import QueryEngineTool
from llama_index.query_engine.base_question_generator import BaseQuestionGenerator
from llama_index.synthesizer.base_synthesizer import BaseSynthesizer
from llama_index.service_context import ServiceContext

# Setup base query engine as tool
query_engine_tools = [
    QueryEngineTool(
        query_engine=vector_query_engine,
        metadata=ToolMetadata(
            name="pg_essay",
            description="Paul Graham essay on What I Worked On",
        ),
    ),
]

# Instantiate the SubQuestionQueryEngine
sub_question_query_engine = SubQuestionQueryEngine.from_defaults(
    query_engine_tools=query_engine_tools,
    question_gen=BaseQuestionGenerator(),
    response_synthesizer=BaseSynthesizer(),
    service_context=ServiceContext(),
    use_async=True,
)

# Define your Pandas query pipeline
qp = QP(
    modules={
        "input": InputComponent(),
        "pandas_prompt": pandas_prompt,
        "llm1": llm,
        "pandas_output_parser": pandas_output_parser,
        "response_synthesis_prompt": response_synthesis_prompt,
        "llm2": llm,
        "sub_question_query_engine": sub_question_query_engine,  # Add the SubQuestionQueryEngine
    },
    verbose=True,
)

# Add chains and links
qp.add_chain(["input", "pandas_prompt", "llm1", "pandas_output_parser"])
qp.add_links(
    [
        Link("input", "response_synthesis_prompt", dest_key="query_str"),
        Link("llm1", "response_synthesis_prompt", dest_key="pandas_instructions"),
        Link("pandas_output_parser", "response_synthesis_prompt", dest_key="pandas_output"),
    ]
)
# Add link from response synthesis prompt to llm2
qp.add_link("response_synthesis_prompt", "llm2")

# Add link to integrate SubQuestionQueryEngine
qp.add_link("response_synthesis_prompt", "sub_question_query_engine")

In this example, the SubQuestionQueryEngine is instantiated and added to the pipeline modules. The add_link method is used to link the response_synthesis_prompt to the sub_question_query_engine, ensuring that the sub-question query engine is part of the pipeline execution flow.

This approach allows you to leverage the capabilities of the SubQuestionQueryEngine within your existing Pandas query pipeline ^[1][2].

To continue talking to Dosu, mention @dosu.

[mraguth]

@dosu in your example, where you got 'query_engine=vector_query_engine'? I don't have that. All I have here is query pipeline and I like to add sub_question_query_engine to pandas query pipe line. Please help.

[dosubot[bot]]

To get the vector_query_engine in the example, you can create it from a VectorStoreIndex using the following code snippet:

# build index and query engine
vector_query_engine = VectorStoreIndex.from_documents(
    documents=documents,
    use_async=True,
).as_query_engine()

This vector_query_engine can then be used to add sub_question_query_engine to a Pandas query pipeline as shown in the example:

from llama_index.core.query_pipeline import QueryPipeline as QP, InputComponent, Link
from llama_index.query_engine.sub_question_query_engine import SubQuestionQueryEngine
from llama_index.query_engine.query_engine_tool import QueryEngineTool
from llama_index.query_engine.base_question_generator import BaseQuestionGenerator
from llama_index.synthesizer.base_synthesizer import BaseSynthesizer
from llama_index.service_context import ServiceContext

# Assuming you have your documents ready
documents = [...]  # Your documents here

# Build the vector query engine
vector_query_engine = VectorStoreIndex.from_documents(
    documents=documents,
    use_async=True,
).as_query_engine()

# Setup base query engine as tool
query_engine_tools = [
    QueryEngineTool(
        query_engine=vector_query_engine,
        metadata=ToolMetadata(
            name="pg_essay",
            description="Paul Graham essay on What I Worked On",
        ),
    ),
]

# Instantiate the SubQuestionQueryEngine
sub_question_query_engine = SubQuestionQueryEngine.from_defaults(
    query_engine_tools=query_engine_tools,
    question_gen=BaseQuestionGenerator(),
    response_synthesizer=BaseSynthesizer(),
    service_context=ServiceContext(),
    use_async=True,
)

# Define your Pandas query pipeline
qp = QP(
    modules={
        "input": InputComponent(),
        "pandas_prompt": pandas_prompt,
        "llm1": llm,
        "pandas_output_parser": pandas_output_parser,
        "response_synthesis_prompt": response_synthesis_prompt,
        "llm2": llm,
        "sub_question_query_engine": sub_question_query_engine,  # Add the SubQuestionQueryEngine
    },
    verbose=True,
)

# Add chains and links
qp.add_chain(["input", "pandas_prompt", "llm1", "pandas_output_parser"])
qp.add_links(
    [
        Link("input", "response_synthesis_prompt", dest_key="query_str"),
        Link("llm1", "response_synthesis_prompt", dest_key="pandas_instructions"),
        Link("pandas_output_parser", "response_synthesis_prompt", dest_key="pandas_output"),
    ]
)
# Add link from response synthesis prompt to llm2
qp.add_link("response_synthesis_prompt", "llm2")

# Add link to integrate SubQuestionQueryEngine
qp.add_link("response_synthesis_prompt", "sub_question_query_engine")

This code demonstrates how to initialize the vector_query_engine and use it within a SubQuestionQueryEngine to integrate it into your Pandas query pipeline ^[1][2].

To continue talking to Dosu, mention @dosu.