SQL agents with rows that creates long tokens

[lasagna0]

Hi to all,

I'm creating a Langchain that helps students of Law ask questions about resolutions and different topics, but this SQL databases have large texts that sometimes the agent executor makes prompts that exceed the openai max tokens giving this error.

InvalidRequestError: This model's maximum context length is 4097 tokens, however you requested 12538 tokens (11538 in your prompt; 1000 for the completion). Please reduce your prompt; or completion length.

So, theres is any way that langchain creates embeddings of the text to avoid this error during the creation of the token? or something that you could recommend to avoid this :3.

[hunterchristian]

Bumping this question, I've hit the same

[jzluo]

Oh wow. If the data MUST be in a SQL database, have your SQL chain just return the query result directly instead of sending it back through the llm to interpret it. Then play around with something like this on the result https://python.langchain.com/en/latest/modules/chains/index_examples/vector_db_qa.html