Redis 8.0 Unveiled: AGPLv3, Vector Search, JSON, and 16× Query Boost

On May 1, 2025 Redis 8.0 was officially released, bringing a leap in performance and major breakthroughs in licensing, data structures, and the query engine. The founder Salvatore Sanfilippo (antirez) returned, re‑embracing the open‑source community and adding an OSI‑approved AGPLv3 option.

1. Open‑source License Innovation: AGPLv3 Restores Open‑source Principles

Redis 8.0 adds AGPLv3 as an open‑source license option alongside the existing RSALv2/SSPLv1, marking a return to the open‑source ecosystem after previous licensing controversies.

License background In March 2024 Redis 7.4 switched from BSD to a dual RSALv2/SSPLv1 license to limit free commercial use by cloud providers, sparking community division and forks like Valkey.

AGPLv3 significance AGPLv3 enforces “network use as distribution,” requiring cloud services that modify the code to contribute back, balancing commercial interests with open‑source spirit.

User impact Developers can freely choose a license; AGPLv3 reduces compliance risk and attracts more contributors.

2. Data Structure Explosion: 8 New Types Covering AI to Time‑Series

2.1 Vector Set: Similarity Search for the AI Era

Core functionality Designed for high‑dimensional vector embeddings, supporting storage and queries for recommendation systems, semantic search, and image retrieval.

Technical highlights Borrowed Sorted Set design, enabling range queries. Integrated Redis Query Engine for real‑time vector search. Benchmarks show 66k–160k vector inserts per second depending on precision.

Application case E‑commerce recommendation calculates user preference similarity for precise product push.

Below is a complete example of an e‑commerce recommendation system using Redis 8.0 Vector Set.

# Create index "product_index" with 768‑dim FLOAT32 vectors, cosine similarity
FT.CREATE product_index
  ON HASH
  PREFIX 1 "product:"
  SCHEMA
    id NUMERIC
    title TEXT
    category TEXT
    embedding VECTOR HNSW 6 TYPE FLOAT32 DIM 768 DISTANCE_METRIC COSINE

Step 2: Insert product data

# Insert product 1 (wireless Bluetooth headphones)
HSET product:1 id 1 title "无线蓝牙耳机" category "electronics" embedding "\x00\x01\x02..."
FT.ADD product_index product:1 VECTOR 0.1 0.2 0.3 ...
# Insert product 2 (smartwatch)
HSET product:2 id 2 title "运动手表" category "electronics" embedding "\x03\x04\x05..."
FT.ADD product_index product:2 VECTOR 0.4 0.5 0.6 ...
# Insert product 3 (coffee machine)
HSET product:3 id 3 title "咖啡机" category "home" embedding "\x06\x07\x08..."
FT.ADD product_index product:3 VECTOR 0.7 0.8 0.9 ...

Step 4: Execute similarity search

# Simulated user query vector (e.g., smartwatch behavior)
FT.NEARESTNEIGHBORS product_index VECTOR 0.4 0.5 0.6 0.7 ... K 2 FILTER @category:electronics

Result (simulated):

1) (integer) 2  # product ID 2 (smartwatch)
"score" -> 0.9876  # similarity score
2) (integer) 1  # product ID 1 (headphones)
"score" -> 0.8765

2.2 Native JSON Support: Structured Data Operations Revolution

Core functionality Store and manipulate JSON documents directly without serialization.

Technical highlights JSONPath syntax for precise nested field targeting. Atomic commands like JSON.SET allow partial updates, avoiding full document transfer. Secondary indexing via the query engine for JSON fields.

Application case IoT platform stores device metadata and quickly filters specific attributes using JSONPath.

# Store device metadata with nested structure
JSON.SET device:1001 $ '{"id":1001,"type":"sensor","location":{"room":"A1","floor":3},"metrics":{"temperature":25.6,"humidity":60}}'
# Query device temperature
JSON.GET device:1001 $.metrics.temperature
# Batch get room info for multiple devices
JSON.MGET device:1001 device:1002 $.location.room
# Atomically update humidity value
JSON.SET device:1001 $.metrics.humidity 65

2.3 Time Series: A Tool for IoT and Financial Data

Core functionality Optimized storage for time‑series data with automatic compression and down‑sampling.

Technical highlights Designed for high‑frequency changing data such as sensor readings or stock prices. Integrated probabilistic data structures for fast aggregation queries. Compatible with monitoring systems like Prometheus.

Application case Financial risk system analyzes transaction anomalies via time‑series analysis.

# Create a time‑series collection (retain 7 days, compressed encoding)
TS.CREATE stock:AAPL RETENTION 604800 ENCODING COMPRESSED
# Add stock price data
TS.ADD stock:AAPL 1728000000000 185.2  # timestamp (ms), price
TS.ADD stock:AAPL 1728003600000 186.1
# Query last hour data
TS.RANGE stock:AAPL -3600000 0
# Compute 1‑minute price volatility
TS.MRANGE 1728000000000 1728003600000 FILTER sensor_id=2 WITHLABELS AGGREGATION STDDEV 60000

2.4 Five Probabilistic Data Structures: Efficiency Tools for Big Data

Bloom Filter Fast existence checks with allowed false positives.

Cuckoo Filter Bloom‑filter variant supporting deletions.

Count‑min Sketch Estimates element frequencies for Top‑K queries.

Top‑K Tracks the most frequent K elements in real time.

t‑Digest Accurately computes quantiles such as median or P99.

3. Query Engine Refactor: 16× Performance Boost, Horizontal Scaling

The Redis Query Engine in 8.0 introduces over 30 performance improvements, becoming the core for complex queries.

Performance leap Multithreaded query execution: vertical scaling, up to 16× throughput. Horizontal scaling: supports massive datasets in cluster mode with linear read/write growth. Command latency reduction: 90 commands see 5.4%‑87.4% latency drops, max 87.4%.

Feature extensions Secondary indexes for precise Hash and JSON field matching and range queries. Full‑text search with stemming, synonym expansion, fuzzy matching. Vector search tightly integrated with Vector Set for AI‑driven similarity retrieval.

4. Performance & Scalability: Throughput Doubled, Replication Lag Cut

Redis 8.0 pushes performance to new heights through low‑level optimizations.

I/O multithreading Configure io-threads to achieve up to 112% throughput increase on multi‑core CPUs; 8‑thread setup doubles OPS.

Replication optimization Parallel data‑change streams reduce master‑slave sync time by 18%; master write rate up 7.5%, peak buffer usage down 35%.

Memory efficiency Improved compression reduces fragmentation; hybrid storage (DRAM+SSD) via Redis Flex lowers cost.

5. Security & Ecosystem: Fine‑grained Controls, AI Assistant

Redis 8.0 also strengthens security and ecosystem integration.

Security enhancements ACL fine‑grained control with new @search and @json permission categories. Vulnerability fixes (e.g., CVE‑2025‑21605) prevent buffer overflows.

Ecosystem integration Redis Copilot: AI‑driven assistant that auto‑generates queries and code. Data integration: seamless connectors to MySQL, PostgreSQL, CDC support. Cloud‑native: Alpine/Debian Docker images and Kubernetes Operator.

6. Upgrade Recommendations: When to Move to 8.0?

AI/ML workloads Vector Set and JSON boost model inference efficiency.

Real‑time analytics Time Series and probabilistic structures accelerate streaming data processing.

High‑concurrency systems I/O multithreading and replication improvements lower latency.

Security‑sensitive environments Enhanced ACL and vulnerability patches improve compliance.

7. Future Outlook: Redis 8.0 Roadmap

Vector search Distributed indexes and ANN algorithms.

SQL compatibility Extended query syntax with JOIN support.

Cloud‑native features Better Serverless support and multi‑region replication.

Redis 8.0’s open‑source return marks a new balance between innovation and community for this “15‑year‑old” database. Whether you are an AI developer, data engineer, or traditional cache user, it’s time to experience this database revolution.