feat: Adds Qdrant vector search

[Anush008]

Description

This PR adds support for using Qdrant as a vector search provider in UltraRAG.

Testing

I've unit tested this integration against a local Qdrant instance.

Setup

You can run Qdrant with

docker run -p 6333:6333 drant/qdrant

The Qdrant URL then will be http://localhost:6333.

The dashboard is accessible at http://localhost:6333/dashboard.

[Anush008]

Hi @xhd0728. Just bumping this PR.

Please take a look when possible.

[xhd0728]

Thanks for adding Qdrant support. I found a blocker in the non-demo indexing path.

QdrantIndexBackend.build_index() requires contents, but the non-demo Retriever.retriever_index() call only passes embeddings, ids, and overwrite. In addition, retriever_init() currently loads the corpus into self.contents only for FAISS/BM25, not Qdrant.

I reproduced it through the CLI on Windows with qdrant-client 1.18.0 and a local Qdrant path:

• ultrarag build examples/experiments/qdrant_index.yaml succeeds.
• After configuring index_backend: qdrant, a JSONL corpus, and a matching precomputed .npy embedding file, ultrarag run examples/experiments/qdrant_index.yaml --param examples/experiments/parameter/qdrant_index_parameter.yaml fails in retriever_index with:

ToolError: [qdrant] 'contents' is required.

Could you load the corpus for Qdrant in the non-demo path and pass the contents aligned with the embedding rows into build_index()? A CLI-level regression test for ultrarag build + ultrarag run would also help cover the actual integration path.

[Anush008]

Resolved with f768da2.

[xhd0728]

Thanks for the pr!