Web Search tool

[sainisanjay]

🚀 The feature

Can we have Web Search tool support in the agent?

Motivation, pitch

We are utilizing PandasAI for our analytics chatbot but are exploring ways to incorporate web-sourced information alongside our own data analytics. I’ve observed that this functionality is already supported by various frameworks such as LangChain, CrewAI, and others.

Alternatives

No response

Additional context

No response

[gventuri]

Hi @sainisanjay currently pandasAI is focused on the data analysis. Could you elaborate further how such an addition would benefit the analysis?

This would help us and the community to properly assess the feature request.

Thanks a lot!

[sainisanjay]

Hi @gventuri Thanks for your respose.

The integration of a web search tool into PandasAI would be beneficial in several ways:

Enriching Data Context: Often, during data analysis, there are situations where our internal datasets lack certain contextual or auxiliary information (e.g., market trends, industry benchmarks, or live updates). A web search tool would enable the agent to fetch relevant, real-time data to supplement and enhance the analysis.

Cross-Referencing: For validation purposes, being able to reference external data directly from the web alongside analytics insights would add more credibility and accuracy to the analysis. For instance, comparing internal sales data with real-time industry statistics sourced from the web.

Broader Use Cases: Incorporating web search functionality can expand PandasAI’s usability beyond traditional analytics to include tasks like competitor analysis, sentiment analysis from online sources, or fetching dynamic financial/economic data.

Seamless Integration: Frameworks like LangChain and CrewAI already support this functionality, and having it in PandasAI would make it more versatile and competitive for end-users, enabling a one-stop solution for analysis and intelligence generation.

[gdcsinaptik]

Hey @sainisanjay closing this issue as it will not be in scope for a while in the core library. But thanks anyway for sharing the thought. The idea of adding a data enrichment agent is indeed very vaulable