You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Redis 8.0 introduces powerful new capabilities, including the beta release of the Vector Set data structure, designed for AI use cases such as semantic search and recommendation systems.
17
+
18
+
This release also delivers over 30 performance improvements, including:
19
+
- Up to 87% faster command execution
20
+
- Up to 2x higher throughput
21
+
- Up to 18% faster replication
22
+
23
+
Additionally, the Redis Community Edition has been renamed to Redis Open Source, marking a shift toward a unified, modern distribution. This version offers the full power of Redis — no modules, no fragmentation — just Redis.
24
+
25
+
Below is a detailed breakdown of these updates.
26
+
27
+
## New features
28
+
### Vector set [beta]
29
+
The new `Vector set` data structure supports high-dimensional vector similarity search, which is ideal for AI use cases such as semantic search and recommendation systems. Vector set complements Redis’s existing vector search capabilities via the Query Engine and is currently available in beta. APIs and behaviors may change in future releases.
30
+
31
+
### New hash commands
32
+
Redis 8 builds on Redis 7.4’s introduction of field-level expiration in hashes and adds three new commands for working with hashes:
33
+
-[HGETEX](https://redis.io/docs/latest/commands/hgetex/): Fetch a hash field and optionally set an expiration
34
+
-[HSETEX](https://redis.io/docs/latest/commands/hsetex/): Set a hash field and optionally set an expiration
35
+
-[HGETDEL](https://redis.io/docs/latest/commands/hgetdel/): Fetch and delete a hash field
36
+
37
+
These commands simplify common caching and session management usage patterns.
38
+
39
+
## Improvements
40
+
### Redis Query Engine Improvements
41
+
The Redis Query Engine allows users to use Redis as a document database, a vector database, a secondary index, and a search engine. With Redis Query Engine, users can define indexes for hash and JSON documents, and use a rich query language for vector search, full-text search, geospatial queries, and aggregations.
42
+
43
+
Use cases include:
44
+
45
+
- Vector Search: Search based on semantic similarity using vector embeddings stored in `JSON` or `hashes`
46
+
- Exact Matching: Lookups using filters, tags, and ranges
47
+
- Full-Text Search: Supports stemming, synonyms, and fuzzy matching
48
+
49
+
The Redis Query Engine powers a wide range of applications, from AI retrieval-augmented generation (RAG) systems to full-featured search engines.
50
+
51
+
### Access Control Improvements
52
+
Access Control Lists (ACLs) have been updated to support the new data structures introduced in Redis 8. Existing ACL categories such as `@read` and `@write` now include commands for `JSON`, `time series`, `VECTOR`, and `probabilistic` data structures.
53
+
54
+
These updates allow for more precise control over which operations users can perform on each data structure.
55
+
56
+
### Performance improvements
57
+
Redis 8 delivers the largest performance leap in Redis history with over 30 optimizations, including:
58
+
- Up to 87% lower command latency
59
+
- 2x throughput in multi-threaded workloads
60
+
- 35% memory savings for replica nodes
61
+
- 16x more query processing capacity with horizontal and vertical scaling
62
+
63
+
These improvements benefit both single-node and clustered deployments. More details are available in the [Redis 8 GA](https://redis.io/blog/redis-8-ga/) blog post.
64
+
65
+
## Changes
66
+
67
+
### Breaking changes
68
+
#### ACL behavior
69
+
Commands from included modules are now covered under standard categories (e.g., `+@read`, `+@write`). A user with `+@all``-@write` will no longer be able to execute `FT.SEARCH` as they could before.
70
+
Explicit inclusion of new command categories is required to maintain access.
71
+
72
+
#### Redis Query Engine
73
+
The following changes affect behavior and validation in the Redis Query Engine:
74
+
- Enforces validation for `LIMIT` arguments (offset must be 0 if limit is 0).
75
+
- Enforces parsing rules for `FT.CURSOR READ` and `FT.ALIASADD`.
76
+
- Parentheses are now required for exponentiation precedence in `APPLY` expressions.
77
+
- Invalid input now returns errors instead of empty results.
78
+
- Default values revisited for reducers like `AVG`, `COUNT`, `SUM`, `STDDEV`, `QUANTILE`, and others.
79
+
- Updates to scoring (`BM25` is now the default instead of `TF-IDF`).
80
+
- Improved handling of expired records, memory constraints, and malformed fields.
81
+
82
+
For a full list of the Redis Query Engine-related changes, see the [release notes](https://github.com/redis/redis/releases).
83
+
84
+
### Other changes
85
+
#### One Redis
86
+
Over time, Redis introduced several modules to address new use cases such as search, vector similarity, time series analysis, and probabilistic modeling. While these modules extended Redis’s functionality, managing version compatibility and module installation introduced friction.
87
+
88
+
Redis Stack partially solved this by bundling modules together, but it also created fragmentation across the ecosystem.
89
+
90
+
Redis 8 merges Redis Stack and Redis Community Edition into a single unified distribution: Redis Open Source. All previously modular functionality is now built into the Redis Open Source package, eliminating the need to manage separate modules. You now get a single, consistent feature set across all deployments.
91
+
92
+
## Component versions
93
+
Redis 8.0 delivers just Redis — no modules, no fragmentation.
0 commit comments