Make intent identification in SPARQL QA chain more resilient

[felixocker]

Feature request

Make the intent identification (SELECT vs UPDATE) within the SPARQL QA chain more resilient
As originally described in #7758 and further discussed in #8521, some models struggle with providing an unambiguous response to the intent identification prompt, i.e., they return a sentence that contains both keywords instead of returning exactly one keyword.
This should be resolvable with a loop that reprompts the model if its response is ambiguous.
The loop should probably be limited to one or two retries.

Motivation

Make the intent identification within the SPARQL QA chain more resilient, so that it can more easily be used with models other than the OpenAI ones, e.g., StarCoder and NeoGPT3 mentioned in #7758, which are still able to generate sensible SPARQL

Your contribution

I should be able to create a PR soonish, but if somebody else wants to give this a go I would also be happy to review

[dosubot[bot]]

Hi, @felixocker! I'm Dosu, and I'm here to help the LangChain team manage their backlog. I wanted to let you know that we are marking this issue as stale.

From what I understand, you created this issue as a feature request to enhance the intent identification in the SPARQL QA chain. The goal is to improve the models' ability to provide clear responses to intent identification prompts, making it easier to use with models other than OpenAI's. You mentioned that you plan to submit a pull request soon and welcome contributions from others.

Before we close this issue, we wanted to check if it is still relevant to the latest version of the LangChain repository. If it is, please let us know by commenting on the issue. Otherwise, feel free to close the issue yourself, or it will be automatically closed in 7 days.

Thank you for your contribution and we appreciate your understanding. Let us know if you have any further questions or concerns.

[felixocker]

yes, this is still relevant

[dosubot[bot]]

@baskaryan Could you please help @felixocker with this issue? They have indicated that it is still relevant. Thank you!

[dosubot[bot]]

Hi, @felixocker

I'm helping the LangChain team manage their backlog and am marking this issue as stale. The issue you opened aims to improve the intent identification process in the SPARQL QA chain by making it more resilient. It involves implementing a loop that reprompts the model if its response is ambiguous, with a suggested limit of one or two retries. The goal is to enable the use of models other than OpenAI's, such as StarCoder and NeoGPT3, for generating sensible SPARQL.

Could you please confirm if this issue is still relevant to the latest version of the LangChain repository? If it is, please let the LangChain team know by commenting on the issue. Otherwise, feel free to close the issue yourself, or it will be automatically closed in 7 days. Thank you!

[baskaryan]

@felixocker seems like there's still some implementation details to be figured out here so moving to discussion for now