Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

feature: add hybrid search retriever using Qdrant in-memory vector store #1176

Open
wants to merge 6 commits into
base: main
Choose a base branch
from
Open

feature: add hybrid search retriever using Qdrant in-memory vector store #1176

wants to merge 6 commits into from

Conversation

asajatovic
Copy link

@asajatovic asajatovic commented Dec 5, 2024 •
edited
Loading

Description

Adds a new hybrid search retriever that combines dense and sparse embedding models to score documents and return the most relevant results. It uses Qdrant's in-memory vector store and supports filtering, with results fused using the Reciprocal Rank Fusion algorithm.
Implementation wise, the new class HybridQdrantInMemoryRetriever inherits QdrantInMemoryRetriever and acts as a drop in replacement, with the addition of also doing sparse vector search, using the reference BM25 algorithm. By using this hybrid approach, we can improve search relevance.

Before Merging

  • [ x] Review the code changes
    • Unused print / comments / TODOs
    • Missing docstrings for functions that should have them
    • Consistent variable names
    • ...
  • [x ] Update changelog.md if necessary
  • [x ] Commit messages should contain a semantic label and the ticket number
    • Consider squashing if this is not the case

@asajatovic asajatovic changed the title feature: add hybrid search retriever using Qdrant in-memory vector st… feature: add hybrid search retriever using Qdrant in-memory vector store Dec 5, 2024
@asajatovic asajatovic marked this pull request as ready for review December 5, 2024 14:17
@asajatovic asajatovic force-pushed the feature/in-memory-hybrid-search branch from 9953869 to f0d8972 Compare December 5, 2024 14:22
NiklasKoehneckeAA
NiklasKoehneckeAA requested changes Dec 9, 2024
View reviewed changes
Copy link
Contributor

@NiklasKoehneckeAA NiklasKoehneckeAA left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I reviewed this, it seems some test code is still missing. Currently IL pipelines are still not fixed after the cluster move, which is something we have are working on, so this is blocked until then.

src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
src/intelligence_layer/connectors/retrievers/hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
tests/connectors/retrievers/test_hybrid_qdrant_in_memory_retriever.py Outdated Show resolved Hide resolved
@asajatovic asajatovic force-pushed the feature/in-memory-hybrid-search branch from f0d8972 to 2d6d24d Compare December 9, 2024 10:37
@asajatovic
feature: add hybrid search retriever using Qdrant in-memory vector st…
a279b5d 
…ore and Reciprocal Rank Fusion
@asajatovic
fix(tests): define fixtures properly
ef4395b
@asajatovic
refactor(retrievers): set default value of max_workers through constr…
b396572 
…uctor

Set default value of `max_workers` in `HybridQdrantInMemoryRetriever` in `__init__`.
In addition, use `min(len(documents), self._max_workers)` as number of concurrent calls to embedding service.
Finally, updates constructor docs to use Google style default values.
@asajatovic
fix(retrievers): fix threshold logic in HybridQdrantInMemoryRetriever
41afc0c
@asajatovic
refactor(tests): reorganize fixtures for hybrid retriever
51638e1
@asajatovic
fix: fix sparse embedding errors in hybrid retriever
b91e180
@asajatovic asajatovic force-pushed the feature/in-memory-hybrid-search branch from e6b6f8c to b91e180 Compare December 18, 2024 14:55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@SebastianNiehusAA SebastianNiehusAA Awaiting requested review from SebastianNiehusAA

@NiklasKoehneckeAA NiklasKoehneckeAA Awaiting requested review from NiklasKoehneckeAA

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@asajatovic @NiklasKoehneckeAA