Download the PHP package hkulekci/qdrant without Composer

Qdrant PHP Client

This library is a PHP Client for Qdrant.

Qdrant is a vector similarity engine & vector database. It deploys as an API service providing search for the nearest high-dimensional vectors. With Qdrant, embeddings or neural network encoders can be turned into full-fledged applications for matching, searching, recommending, and much more!

Installation

You can install the client in your PHP project using composer:

Versioning & Compatibility

Starting with v1.19.0, the version of this client follows the Qdrant server version: 1.19.x releases target and are tested against Qdrant 1.19, and new client releases are published together with new Qdrant minor versions. Patch releases (1.19.1, 1.19.2, ...) contain client-side fixes only.

[!IMPORTANT] The jump from v1.0.0 to v1.19.0 only aligns the version numbers with Qdrant, it does not remove or rename any public API. Still, please review the following before upgrading:

• Features such as TurboQuant, memory tiers, named vector management or quotas require a Qdrant server that supports them. Older servers may reject or silently ignore the new fields.
• false and 0 values set on HnswConfig, WalConfig and CollectionParams (for example on_disk: false) were previously dropped and are now sent to the server.
• CreateShardKey now sends shards_number; the shard number was previously ignored by the server.
• on_disk, always_ram, on_disk_payload, memmap_threshold and init_from are deprecated in Qdrant and marked @deprecated in the client. Prefer the memory options and snapshots.
• The legacy search() and recommend() endpoints are deprecated; use the Query API.

Connecting to Qdrant

Creating a Collection

Inserting Points Into Collection

Wait for Acknowledges

While upsert data, if you want to wait for upsert to actually happen, you can use query parameters:

You can check for more parameters : https://qdrant.github.io/qdrant/redoc/index.html#tag/points/operation/upsert_points

Search on Points

Use the universal Query API (points()->query()). The legacy search() and recommend() endpoints are deprecated and have been removed from the Qdrant OpenAPI specification since Qdrant 1.19.

Hybrid search with prefetch and fusion (weighted RRF, MMR, formula and relevance feedback queries are passed the same way):

Quantization (TurboQuant, Scalar, Product, Binary)

TurboQuant (Qdrant 1.18+) gives up to 8x vector compression with high recall:

Since Qdrant 1.19 vectors can be stored only as TurboQuant 4-bit, without keeping the original vectors:

Other quantization methods are available as ScalarQuantization, ProductQuantization and BinaryQuantization (with encoding / query_encoding for 1.5-bit, 2-bit and asymmetric binary quantization). Use DisabledQuantization with UpdateCollection to turn quantization off.

Memory Tiers

Since Qdrant 1.19 the memory option (cold, cached, pinned) replaces the on_disk and always_ram flags for each collection component:

Sparse Vectors and Named Vectors

Update Modes and Conditional Updates

Filters

Besides the classic conditions (MatchString, MatchInt, MatchAny, Range, GeoRadius, ...) the client supports MatchPrefix (1.19, requires a keyword index created with prefix: true), MatchTextAny, MatchPhrase, IsNull, HasVector and Slice (1.19, deterministic partitioning for parallel scroll or sampling):

Operations

AI-Assisted Development

This library ships with an AI context file that helps AI assistants (Claude Code, etc.) understand the full API surface. To enable it in your project:

This gives AI tools a complete reference of all endpoints, request models, filters, and usage examples.