Skip to content

Commit

Permalink
update comparision to pinecone
Browse files Browse the repository at this point in the history
  • Loading branch information
@generall
generall committed Feb 17, 2024
1 parent a9a2cad commit 9f7d10d
Showing 1 changed file with 7 additions and 7 deletions.
14 changes: 7 additions & 7 deletions qdrant-landing/content/documentation/overview/qdrant-alternatives.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -30,13 +30,13 @@ Although both tools serve similar purposes, there are some differences in the te
in handy during the transition.

| Pinecone | Qdrant | Comments |
|----------------|-----------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **Index** | [**Collection**](../../concepts/collections/) | Pinecone's index is an organizational unit for storing and managing vectors of the same size. The index is tightly coupled with hardware (pods). Qdrant uses the collection to describe a similar concept, however, a single instance may handle multiple collections at once. |
| **Collection** | N/A | A collection in Pinecone is a static copy of an *index* that you cannot query, mostly used as some sort of backup. There is no such concept in Qdrant, but if you want to back your collection up, you may always create a [snapshot](../../concepts/snapshots/). |
| **Namespace** | [**User-defined sharding**](../../guides/distributed_deployment/#user-defined-sharding) | Namespaces allow the partitioning of the vectors in an index into subsets. Qdrant features user-defined sharding for this, which is more versatile as operations can be over multiple shards. |
| **Metadata** | [**Payload**](../../concepts/payload/) | Additional attributes describing a particular object, other than the embedding vector. Both engines support various data types, but Pinecone metadata is key-value, while Qdrant supports any JSON-like objects. |
| **Query** | [**Search**](../../concepts/search/) | Name of the method used to find the nearest neighbors for a given vector, possibly with some additional filters applied on top. |
| N/A | [**Scroll**](../../concepts/points/#scroll-points) | Pinecone does not offer a way to iterate through all the vectors in a particular index. Qdrant has a `scroll` method to get them all without using search. |
|----------------|------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **Index** | [**Collection**](../../concepts/collections/) | Pinecone's index is an organizational unit for storing and managing vectors of the same size. The index is tightly coupled with hardware (pods). Qdrant uses the collection to describe a similar concept, however, a single instance may handle multiple collections at once. |
| **Collection** | [**Snapshots**](../../concepts/snapshots/) | A collection in Pinecone is a static copy of an *index* that you cannot query, mostly used as some sort of backup. There is no direct analogy in Qdrant, but if you want to back your collection up, you may always create a more flexible [snapshot](../../concepts/snapshots/). |
| **Namespace** | [**Payload-based isolation**](../../guides/multiple-partitions/) / [**User-defined sharding**](../../guides/distributed_deployment/#user-defined-sharding) | Namespaces allow the partitioning of the vectors in an index into subsets. Qdrant provides multiple tools to ensure efficient data isolation within a collection. For fine-grained data segreation you can use payload-based approach to multitenancy, and use custom sharding at bigger scale |
| **Metadata** | [**Payload**](../../concepts/payload/) | Additional attributes describing a particular object, other than the embedding vector. Both engines support various data types, but Pinecone metadata is key-value, while Qdrant supports any JSON-like objects. |
| **Query** | [**Search**](../../concepts/search/) | Name of the method used to find the nearest neighbors for a given vector, possibly with some additional filters applied on top. |
| N/A | [**Scroll**](../../concepts/points/#scroll-points) | Pinecone does not offer a way to iterate through all the vectors in a particular index. Qdrant has a `scroll` method to get them all without using search. |

## Known limitations

Expand Down

0 comments on commit 9f7d10d

Please sign in to comment.