superagent-ai
Do not fetch embeddings on every query to a Document Agent
Please check that this issue hasn't been reported before.
- [X] I searched previous Bug Reports didn't find any similar reports.
Expected Behavior
In order to optimize the the request times we need to store embeddings after it's created the first time.
Current behaviour
Currrently embeddings are created for each query.
Steps to reproduce
Call a DoumentAgent.
Possible solution
No response
Which Operating Systems are you using?
- [ ] Android
- [ ] iPhone/iPad
- [ ] Linux
- [X] macOS
- [ ] Windows
Acknowledgements
- [X] My issue title is concise, descriptive.
- [X] I have searched the existing issues to make sure this bug has not been reported yet.
- [X] I am using the latest version of superagent.
- [X] I have provided enough information for the maintainers to reproduce and diagnose the issue.
@homanp ...hmm. Trying to understand this. Are you talking about at a question level or a document level? At a document level, we can do a similarity search based on the question already being embedded when asked.
@homanp ...hmm. Trying to understand this. Are you talking about at a question level or a document level? At a document level, we can do a similarity search based on the question already being embedded when asked.
I'm specifically thinking about this row: https://github.com/homanp/superagent/blob/main/app/lib/agents/base.py#L258
Every time a request via the API is processed this code is run. Just to fetch the Vectorstore. I'm thinking this should be optimized by making the vectorstore persistant.