feat: Index key generation fix

[chandrasekharan-zipstack]

What

• Index key generation fixed - hash a dict of values instead of combining UUIDs
• File hash function is optimized
• Minor: Dead code removed in wrappers of adapters
• Version bumped to 0.23.0, lockfile updated
• MINOR: Some error handling code around vector DB / embedding was removed
• MINOR: Increased line length limit to precommit hooks to 88

Why

• File hashing was incorrect - hashes were appended while in reality they should be XOR-ed
• Previous index key gen logic did not allow adapter updates to be reflected during reindex of a document, since they were based on UUIDs
• Error handling code removal reason here

How

• New index key gen logic uses the adapter configurations and hash them altogether

Can this PR break any existing features. If yes please list of possible items. If no please explain why. (PS: Admins do not merge the PR without this section filled)

Yes, since there's an update in the index key generated - all old index keys would be invalid. In order to perform a clean up - a migration has been added in Zipstack/unstract#289

Related PRs

• Unstract PR that contains the necessary migrations

Notes on Testing

• Tested them along with other changes in unstract, index key generated as sha 256 hash.
• Adapter changes causes index to be generated again

Screenshots

Checklist

I have read and understood the Contribution Guidelines.

[CLAassistant]

All committers have signed the CLA.

[ritwik-g]

@chandrasekharan-zipstack I have looked at the code from a very high level since the relation of same with existing code is not something that I know very well.

One major pattern I observed is that we have removed some of the instance variables. I would like you to double check and make sure those are not used in other repos and if they are we have made required changes

[hari-kuriakose]

@chandrasekharan-zipstack Doc ID generation looks good.

Please double check the impact on other areas.