Qdrant-client

Changelog

---
Features 🪄
- exposed _gRPC_ options 401
- support setting custom shard key in `upload_collection` and `upload_points` 396
- unlocked python dependency upper cup 419
- unlocked `urllib3` v2 418
- ease access to `fastembed` models 412
- extend access for `fastembed` configuration 416

---
Fixes 🪛
- replaced `httpx` _data_ with _content_ 426
- forbid having _NaN_ in payload in `local mode` 397
- fix python warning which was sometimes emitted on garbage collection in `local mode` 398 421
- fix division by zero error in cosine similarity in `local mode` 425
- add `wait` parameter to `upload_records` in `migrate` to avoid false-positive asserts 428
- fix operations with _sparse vectors_ with unsorted indices in _local mode_ 442
- restrict inserting points with non-configured vectors in _local mode_ 432
- fix reading collections with unnamed dense and named sparse vectors in _local mode_ 433
- forbid to `migrate` from collections with custom shard keys 447
- align `timeout` param types in `QdrantClient` and its methods


Deprecation ⏳
- `upload_records` has been deprecated in favour of `upload_points` 447

Thanks to the devs who contributed to this release

generall shivas1516 praveen-palanisamy joein

Changelog

---
Features

* Add Discovery API ([docs](https://qdrant.tech/documentation/concepts/explore/#discovery-api)) - https://github.com/qdrant/qdrant-client/pull/368 by coszio
- **Discovery search**: Uses a target and context pairs to limit the search space during the search
- **Context search**: Using only the context pairs, get the points that live in the best zone
* Add Sparse vectors support ([docs](https://qdrant.tech/documentation/concepts/collections/#collection-with-sparse-vectors)) - https://github.com/qdrant/qdrant-client/pull/378 by agourlay
- Combine the dense and sparse vectors to build hybrid search
* Add Manhattan distance support - https://github.com/qdrant/qdrant-client/pull/391 by generall
* Add custom shard key API ([docs](https://qdrant.tech/documentation/guides/distributed_deployment/#user-defined-sharding)) - https://github.com/qdrant/qdrant-client/pull/391 by generall
- Configure how to distribute points among shards

---
Fixes

* https://github.com/qdrant/qdrant-client/pull/337 - Add `wait` parameter to snapshots API by joein

Changelog

* Missing fix form previous release: https://github.com/qdrant/qdrant-client/pull/338 - add missing parameters in recommend batch API

Changelog

Features

* Proper support for Async client - https://github.com/qdrant/qdrant-client/pull/319
* Now all methods of regular qdrant client are available in async version.
* All method names, parameters and return types are preserved.
* Both gRPC and REST version are available.

* Improvements of [fastembed](https://github.com/qdrant/fastembed) integration:
* Support the latest v0.1.x version of fastembed: https://github.com/qdrant/fastembed/releases/tag/0.1.0
* Support for encoding of iterators in `add` API, useful for lazy reading of large dataset
* Support for data-level parallelism of the encoding operations


Fixes

* https://github.com/qdrant/qdrant-client/pull/337 - fix: convert score to float in local mode for pydantic

---

Async Qdrant client usage example:

python
from qdrant_client import AsyncQdrantClient, models
import numpy as np
import asyncio

async def main():
client = AsyncQdrantClient(url="http://localhost:6333")

res = await client.search(
collection_name="my_collection",
query_vector=np.random.rand(10).tolist(), type: ignore
limit=10,
)

print(res)

asyncio.run(main())

Changelog

Imptovements

* Support for Qdrant 1.6.x features
* https://github.com/qdrant/qdrant-client/pull/325 - geo polygons filtering
* https://github.com/qdrant/qdrant-client/pull/314 - recommendation API
* Support for python 3.12
* Integration of FastEmbed v0.0.5 - https://github.com/qdrant/fastembed/releases/tag/0.0.5

Bug fixes

* https://github.com/qdrant/qdrant-client/pull/330 - fix usage of write ordering parameter
* https://github.com/qdrant/qdrant-client/pull/332 - `wait` param to the upload records method
* https://github.com/qdrant/qdrant-client/pull/317 - fix for `init_from` parameter

Changelog

Features

* https://github.com/qdrant/qdrant-client/pull/280, https://github.com/qdrant/qdrant-client/commit/6f8c5172d77a1775474d6ae145e7552543803806 - Compatibility updates for Qdrant v1.5.x
* https://github.com/qdrant/qdrant-client/pull/210 - [fastembed](https://github.com/qdrant/fastembed) integration. Enables lightweight, fast, Python library built for retrieval embedding generation.
* https://github.com/qdrant/qdrant-client/pull/243 - Migration tool, allows easy data migration from one instance to another

Bugfix

* https://github.com/qdrant/qdrant-client/pull/258 - disable forcing of http2 for cloud connections
* https://github.com/qdrant/qdrant-client/pull/268 - fix values count & is_empty & is_null conditions for local mode


Important Notes

* Python 3.7 is no longer supported

---

Use fastembed library to easily encode & index documents into qdrant


pip install fastembed qdrant-client


python
from qdrant_client import QdrantClient

Initialize the client
client = QdrantClient(":memory:") or QdrantClient(path="path/to/db")

Prepare your documents, metadata, and IDs
docs = ["Qdrant has Langchain integrations", "Qdrant also has Llama Index integrations"]
metadata = [
{"source": "Langchain-docs"},
{"source": "Linkedin-docs"},
]
ids = [42, 2]

Use the new add method
client.add(
collection_name="demo_collection",
documents=docs,
metadata=metadata,
ids=ids
)

search_result = client.query(
collection_name="demo_collection",
query_text="This is a query document"
)
print(search_result)


More in [Notebook](https://github.com/qdrant/fastembed/blob/main/docs/examples/Usage_With_Qdrant.ipynb)