ofermend
commented
11 months ago Hi there! Thank you for offering the contribution. However, the vectara-ingest project is not intended to be a general-purpose ETL tool for arbitrary destinations. There are a variety of generic ETL tools on the market, and the reason vectara-ingest is separate is in the name: to optimize data loading into Vectara. A couple thoughts:
- If you want to submit a PR to a generic ETL tool to load data into your product, you could do so.
- To maintain the spirit of this repository, we would consider a PR that exports data from Arguflow and ingests it into Vectara if such an API exists on the Arguflow side.
eskibars
This is the Arguflow team’s submission to the “best contribution to vectara-ingest” part of the hackathon. We really admire the work that has gone into this repository, and want to start a trend of making it compatible with more varied services.
This PR adds support for Arguflow to the crawler such that users are able to add documents/chunks to Arguflow, Vectara, or both.
Internally, we were motivated to add this support so that we can stand up more Arguflow demos by using the crawlers, however, we are PR’ing it because we think it can also offer Vectara users value.
Arguflow has support for a few things that Vectara does not which users may desire:
enhanced duplicate detection beyond UPSERT
Arguflow also ships with a default search and chat UI in addition to its OpenAPI spec which makes it a bit easier to get going with a deployment you can share and reverse-engineer to build your own applications.
This repository is great, and we really admire @ofermend’s work on it especially. Excited to enhance it and bring it more into the lens of the open source AI world! We also edited the documentation in all the right places, let us know if may have missed a few spots.
Happy to address any review comments or change requests in a timely manner.