Pandas Query Engines

[mraguth]

I am new to llamaindex and I'm curious about the distinction between Query Pipeline (https://docs.llamaindex.ai/en/stable/examples/pipeline/query_pipeline_pandas/) and the pandas query engine (https://docs.llamaindex.ai/en/stable/examples/query_engine/pandas_query_engine/). I noticed that Query Pipeline provides higher accuracy compared to the pandas query engine.

Additionally, I am interested in understanding the importance of response_synthesis prompt and how I can tailor it for more precise answers. Thank you for all your advise.

[dosubot[bot]]

The LlamaIndex Query Pipeline and the pandas query engine serve different purposes and operate at different levels of abstraction. The Query Pipeline is a more advanced component that leverages Language Models (LLMs) for interpreting user queries and synthesizing responses, aiming for higher accuracy and user-friendly outputs by understanding the context of the query. It's designed to infer the necessary operations to satisfy a query and then synthesize a response based on the output, making it more flexible and capable of providing tailored responses.

In contrast, the pandas query engine is focused on directly executing pandas operations without the intermediate step of natural language understanding or response synthesis. This means it's more straightforward but lacks the ability to interpret queries or synthesize responses in a user-friendly manner.

The response_synthesis_prompt is crucial in the Query Pipeline for guiding the LLM on how to craft responses based on the results of executed operations. Customizing this prompt allows for more precise answers by providing specific instructions or context to the LLM. You can tailor the response_synthesis_prompt by modifying the BasePromptTemplate instance or creating a new one with a customized template string, which then needs to be assigned to the response_synthesis_prompt attribute of the JSONalyzeQueryEngine class. This customization can lead to more accurate or relevant synthesized responses by including more specific instructions or context for the LLM.

For more detailed customization and understanding of how these components work, including the response_synthesis_prompt, you might want to explore the source code and documentation provided in the LlamaIndex GitHub repository, specifically within the query_pipeline_pandas.ipynb, pandas_query_engine.py, and jsonalyze_query_engine.py files.

Details

Sources

• llama-index-core/llama_index/core/query_engine/jsonalyze_query_engine.py
• llama-index-core/llama_index/core/response_synthesizers/factory.py
• docs/docs/examples/low_level/response_synthesis.ipynb
• docs/docs/examples/pipeline/query_pipeline_pandas.ipynb
• llama-index-experimental/llama_index/experimental/query_engine/pandas/pandas_query_engine.py
• docs/docs/module_guides/querying/response_synthesizers/index.md
• docs/docs/module_guides/models/prompts/usage_pattern.md

About Dosu
This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

To continue the conversation, mention @dosu.