Chat with a CSV - LangChain CSV Agents Tutorial For Beginners (OpenAI API)

[@RyanAndMattDataScience]

Hey guys I hope you enjoyed the video! If you did please subscribe to the channel! Join our Data Science Discord Here: discord.com/invite/F7dxbvHUhg If you want to watch a full course on Langchain check out Datacamp: datacamp.pxf.io/XYD7Qg Want to solve Python data interview questions: stratascratch.com/?via=ryan I'm also open to freelance data projects. Hit me up at [e-mail] *Both Datacamp and Stratascratch are affiliate links.

[@evanklingle8842]

Ryan- This is awesome! I’ve been looking for tutorials to get into LangChain. It’s been super intimidating, but you broke it down into simple steps. Thank you!

[@RyanAndMattDataScience]

No problem I have another agent type video out next week with csvs

[@RyanAndMattDataScience]

Looking for the code? I just uploaded an article based on this video: ryannolandata.com/chat-with-a-csv-using-langchain/ Also, I'm taking on Data and AI projects. Have something for me to work on? Fill out the contact form on my site or email me.

[@ta_helado]

how to solve the outputparsererror you got? im having the same

[@devalmodi141]

How do I take the file name from the user prompt and then ask questions on it?

[@Abdulrhman146.]

Thanks a lot Ryan, that was a great tutorial. do you know how can i add memory to the agent? i want it to remember all interactions with user.

[@FIRZAICHLASULAMALARIANSYAH]

Hello, i want to ask. It possible if result/output of the prompt outside the .csv?

[@vispinet]

I have the same question. Also it would be nice to know how to handle exceptions

[@usamasajid2600]

Ryan I follow your tutorail install all libraries but having a problem in creating csv agent when i run this command agent = create_csv_agent(ChatOpenAI(temperature=0), "/content/drive/MyDrive/data.csv", verbose=True) It gives me this error below ValueError: Prompt missing required variables: {'tool_names', 'tools'} I am an absolute beginner and can't able to fix this error please help.... Please Reply

[@mikedowns1842]

armchair comment, but I believe agents must have tools

[@GamerGDS]

How is this different from a simple OpenAI Assistant that has file_search?

[@tonygan1132]

Good tutorial, everything is clear, learned a lot, thx for the efforts.

[@RyanAndMattDataScience]

Thank you for checking it out

[@stanTrX]

What s the difference with directly parsing csv to llm vs using RAG? Per this example?

[@rohansingh1057]

RAG just gives you the context you require to answer the question. While nice for paragraph style text. It is awful for CSV or tabular data. Simple operations like summing up the table will be quite confusing for the LLM. Also it's nearly impossible to use it with large tables (or even the one shown here). Assume you have 100K rows, and you ask for the mean of a column. Pretty simple, but would break RAG. Here, langchain is using a technique known as "chain of thought" where it has a problem and it tries to solve it using "tools". These tools are defined as prompts and in this case these tools are Pandas functions and the LLM output is then run as Pandas code over the actual CSV. So essentially when Ryan asked the total wins, the LLM did not do any calculations. It simply deduced the csv['W'].sum() should give the correct answer and that generated code was executed by the CSV. Also in RAG, there is only 1 API call (generally speaking in most basic RAG implementations), while in Chain of Thought agents, there can be several API calls to the LLM.

[@stanTrX]

@@rohansingh1057 thanks a lot

[@NileshKumar_NK]

@@rohansingh1057 bro I am facing the same problem, not able to get simple answers from a csv file, like avg, sum etc. How can I do it. I am a beginner in this. Can u guide me a lil.

[@rohansingh1057]

@@NileshKumar_NK Follow the video as it is. It is the best there is. Alternatively you can use OpenAI assistants API which allows you to upload a few files and get answers but I think that will be pricing for GPT-4 which is very expensive.

[@sherifalaa5]

@@rohansingh1057 amazing explaination, thank you

[@ranati2000]

{'input': 'How many total FIPS are there?', 'output': 'Agent stopped due to iteration limit or time limit.'} Any Idea please let me know ?

[@ihsanbp7786]

can you share the colab notebook?

[@RyanAndMattDataScience]

I need to upload all the vids into GitHub. I’m so far behind but n doing that

[@akshatjain5586]

is the openAi api key free to use ?

[@RyanAndMattDataScience]

They give you credit when you create an account

[@akath22]

Would be nice to get the code snippets :D Thanks for the tutorial

[@RyanAndMattDataScience]

I can get the Notebook in the description later today.

[@akath22]

Tried this and it has too many bugs unfortunately

[@RyanAndMattDataScience]

It worked for me? Maybe there is a typo in your code? Uploading the notebook later today

[@FrozenValorantYT]

Doesn’t work accurately for large dataset like 55K rows

[@RyanAndMattDataScience]

Haven’t tested it with a large dataset yet. Maybe another project down the line.

[@nishadkhare7164]

chunk that dataset and test

[@loveutube04]

Bro looks like the younger version of seriel k!ller guy from No country for Old Men