Langchain Agent Toolkits - Pandas DataFrame, JSON and SQL Database Agents

In this video, we look at different agent toolkits for Langchain, including:

1. Pandas DataFrame agent - for interrogating Pandas DataFrames
2. JSON Agent - for querying against JSON data
3. SQL Database Agents - for interrogating SQL Databases

We'll look at the concept of a "tool" in langchain, and how agent toolkits combine different tools and make smart decisions at each step of an iterative process.

☕️ 𝗕𝘂𝘆 𝗺𝗲 𝗮 𝗰𝗼𝗳𝗳𝗲𝗲:
To support the channel and encourage new videos, please consider buying me a coffee here:

📌 𝗖𝗵𝗮𝗽𝘁𝗲𝗿𝘀:
00:00 Intro
02:28 Pandas DataFrame Agent
09:31 JSON Agent
17:15 SQL Database Agent

𝗦𝗼𝗰𝗶𝗮𝗹 𝗠𝗲𝗱𝗶𝗮:

📚 𝗙𝘂𝗿𝘁𝗵𝗲𝗿 𝗿𝗲𝗮𝗱𝗶𝗻𝗴 𝗮𝗻𝗱 𝗶𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻:

#python #langchain #datascience #llm

[silkogelman]

These LangChain agents and tools are very interesting, thank you! 😀🙏
Defining a custom tool also looks cool, as interface to your own API backend in Django / FastAPI or so.

[MichaelSchwabTX]

Thanks for the informative video. I did find it amusing that the AI said females were MORE likely to survive even though they had a LOWER survival score of 2.159 vs the males as 2.389. I suppose we should take this as a cautionary discussion that you really cannot assume it will always give the right answer.

[jhanicoetzee]

Thanks for the great video!
I am running into issues with big datasets and the JSON Agent, that it doesn't consider all the data, just basically that fits into the context window. From my understanding, it is supposed to overcome the context window issue. Any advice on this?

[siddharthchoudhury]

Please make a video on how to have chat continuation in pandas agent

[flfbeqk]

Thank you for this video! I am having trouble using pandas agent: When I ask for more than one answer in the same question it throws errors at me.

If I ask for the name or I ask for the age of the oldest passenger, no problem sending me the result.
But if I ask for the name and age of the oldest passenger I get a JSONDecodeError, Thoughts?

[CMAZZONI]

Have you tried using a open source model for this? Most times OpenAi will not be usable, specially if your data contains sensitive or personal information.

[fasedbikes]

have you checked openSSM?seems interesting

[nitinnegi]

i am using sqldatabase toolkit at an enterprise, data is huge.When i am using this toolkit i am getting max tokens error and query is totally depend upon the toolkit.Can you tell me how to modify sqldatabase toolkit and is this possible to get 100% accuracy?

[theanigos]

I am getting this error

1 validation error for JsonToolkit
spec
value is not a valid dict (type=type_error.dict)

[sudhanshurathore]

Hey I am usign Azureopenai LLM model but constantly parsing error. Can you help?