Skip to content

A self-hostable CDN for databases. Spice provides a unified SQL query interface and portable runtime to locally materialize, accelerate, and query datasets from any database, data warehouse, or data lake.


Notifications You must be signed in to change notification settings

spiceai/spiceai OSS

CodeQL License: Apache-2.0 Discord Follow on X

What is Spice?

Spice is a portable runtime offering developers a unified SQL interface to materialize, accelerate, and query data from any database, data warehouse, or data lake.

📣 Read the OSS announcement blog post.

Spice connects, fuses, and delivers data to applications, machine-learning models, and AI-backends, functioning as an application-specific, tier-optimized Database CDN.

The Spice runtime, written in Rust, is built-with industry leading technologies such as Apache DataFusion, Apache Arrow, Apache Arrow Flight, SQLite, and DuckDB.

How Spice works.

Why Spice?

Spice makes it easy and fast to query data from one or more sources using SQL. You can co-locate a managed dataset with your application or machine learning model, and accelerate it with Arrow in-memory, SQLite/DuckDB, or with attached PostgreSQL for fast, high-concurrency, low-latency queries. Accelerated engines give you flexibility and control over query cost and performance.

How is Spice different?

  1. Application-focused: Spice is designed to integrate at the application level; 1:1 or 1:N application to Spice mapping, whereas most other data systems are designed for multiple applications to share a single database or data warehouse. It's not uncommon to have many Spice instances, even down to one for each tenant or customer.

  2. Dual-Engine Acceleration: Spice supports both OLAP (Arrow/DuckDB) and OLTP (SQLite/PostgreSQL) databases at the dataset level, unlike other systems that only support one type.

  3. Separation of Materialization and Storage/Compute: Spice separates storage and compute, allowing you to keep data close to its source and bring a materialized working set next to your application, dashboard, or data/ML pipeline.

  4. Edge to Cloud Native. Spice is designed to be deployed anywhere, from a standalone instance to a Kubernetes container sidecar, microservice, or cluster at the Edge/POP, On-Prem, or in public clouds. You can also chain Spice instances and deploy them across multiple infrastructure tiers.

How does Spice compare?

Spice Trino/Presto Dremio Clickhouse
Primary Use-Case Data & AI Applications Big Data Analytics Interative Analytics Real-Time Analytics
Typical Deployment Colocated with application Cloud Cluster Cloud Cluster On-Prem/Cloud Cluster
Application-to-Data System One-to-One/Many Many-to-One Many-to-One Many-to-One
Query Federation Native with query push-down Supported with push-down Supported with limited push-down Limited
Materialization Arrow/SQLite/DuckDB/PostgreSQL Intermediate Storage Reflections (Iceberg) Views & MergeTree
Query Result Caching Supported Supported Supported Supported
Typical Configuration Single-Binary/Sidecar/Microservice Coodinator+Executor w/ Zookeeper Coodinator+Executor w/ Zookeeper Clickhouse Keeper+Nodes

Example Use-Cases

1. Faster applications and frontends. Accelerate and co-locate datasets with applications and frontends, to serve more concurrent queries and users with faster page loads and data updates. Try the CQRS sample app

2. Faster dashboards, analytics, and BI. Faster, more responsive dashboards without massive compute costs. Watch the Apache Superset demo

3. Faster data pipelines, machine learning training and inferencing. Co-locate datasets in pipelines where the data is needed to minimize data-movement and improve query performance. Predict hard drive failure with the SMART data demo

4. Easily query many data sources. Federated SQL query across databases, data warehouses, and data lakes using Data Connectors.


  • Is Spice a cache? No, however you can think of Spice data materialization like an active cache or data prefetcher. A cache would fetch data on a cache-miss while Spice prefetches and materializes filtered data on an interval or as new data becomes available. In addition to materialization Spice supports results caching.

  • Is Spice a CDN for databases? Yes, you can think of Spice like a CDN for different data sources. Using CDN concepts, Spice enables you to ship (load) a working set of your database (or data lake, or data warehouse) where it's most frequently accessed, like from a data application or for AI-inference.

  • Where is the AI? Spice provides a unified API for both data and AI/ML with a high-performance bus between the two. However, because the first step in AI-readiness is data-readiness, the Getting Started content is focused on data. Spice has endpoints and APIs for model deployment and inference including LLMs, accelerated embeddings, and an AI-gateway for providers like OpenAI and Anthropic. Read more about the vision to enable development of intelligent AI-driven applications.

Watch a 30-sec BI dashboard acceleration demo


Supported Data Connectors

Currently supported data connectors for upstream datasets. More coming soon.

Name Description Status Protocol/Format
databricks Databricks Alpha Spark Connect
S3/Delta Lake
delta_lake Delta Lake Alpha Delta Lake
postgres PostgreSQL Alpha
spiceai Alpha Arrow Flight
s3 S3 Alpha Parquet, CSV
dremio Dremio Alpha Arrow Flight
mysql MySQL Alpha
duckdb DuckDB Alpha
clickhouse Clickhouse Alpha
odbc ODBC Alpha ODBC
spark Spark Alpha Spark Connect
flightsql Apache Arrow Flight SQL Alpha Arrow Flight SQL
snowflake Snowflake Alpha Arrow
ftp, sftp FTP/SFTP Alpha Parquet, CSV
graphql GraphQL Alpha JSON
debezium Debezium CDC Alpha Kafka + JSON

Supported Data Stores/Accelerators

Currently supported data stores for local materialization/acceleration. More coming soon.

Name Description Status Engine Modes
arrow In-Memory Arrow Records Alpha memory
duckdb Embedded DuckDB Alpha memory, file
sqlite Embedded SQLite Alpha memory, file
postgres Attached PostgreSQL Alpha file

⚠️ DEVELOPER PREVIEW Spice is under active alpha stage development and is not intended to be used in production until its 1.0-stable release. If you are interested in running Spice in production, please get in touch so we can support you (See Connect with us below).

⚡️ Quickstart (Local Machine)


Step 1. Install the Spice CLI:

On macOS, Linux, and WSL:

curl | /bin/bash

Or using brew:

brew install spiceai/spiceai/spice

On Windows:

curl -L "" -o Install.ps1 && PowerShell -ExecutionPolicy Bypass -File ./Install.ps1

Step 2. Initialize a new Spice app with the spice init command:

spice init spice_qs

A spicepod.yaml file is created in the spice_qs directory. Change to that directory:

cd spice_qs

Step 3. Start the Spice runtime:

spice run

Example output will be shown as follows: runtime starting...
Using latest 'local' runtime version.
2024-06-03T23:21:26.819978Z  INFO spiced: Metrics listening on
2024-06-03T23:21:26.821863Z  INFO runtime::http: Spice Runtime HTTP listening on
2024-06-03T23:21:26.821898Z  INFO runtime::flight: Spice Runtime Flight listening on
2024-06-03T23:21:26.821958Z  INFO runtime::opentelemetry: Spice Runtime OpenTelemetry listening on
2024-06-03T23:21:26.822128Z  INFO runtime: Initialized results cache; max size: 128.00 MiB, item ttl: 1s

The runtime is now started and ready for queries.

Step 4. In a new terminal window, add the spiceai/quickstart Spicepod. A Spicepod is a package of configuration defining datasets and ML models.

spice add spiceai/quickstart

The spicepod.yaml file will be updated with the spiceai/quickstart dependency.

version: v1beta1
kind: Spicepod
  - spiceai/quickstart

The spiceai/quickstart Spicepod will add a taxi_trips data table to the runtime which is now available to query by SQL.

2024-06-03T23:21:29.721705Z  INFO runtime: Dataset taxi_trips registered (s3://spiceai-demo-datasets/taxi_trips/2024/), acceleration (arrow, 10s refresh), results cache enabled.
2024-06-03T23:21:29.722839Z  INFO runtime::accelerated_table::refresh: Loading data for dataset taxi_trips
2024-06-03T23:21:50.813510Z  INFO runtime::accelerated_table::refresh: Loaded 2,964,624 rows (421.71 MiB) for dataset taxi_trips in 21s 90ms.

Step 5. Start the Spice SQL REPL:

spice sql

The SQL REPL inferface will be shown:

Welcome to the SQL REPL! Type 'help' for help.

show tables; -- list available tables

Enter show tables; to display the available tables for query:

sql> show tables
| table_catalog | table_schema | table_name    | table_type |
| spice         | public       | taxi_trips    | BASE TABLE |
| spice         | runtime      | metrics       | BASE TABLE |
| spice         | runtime      | query_history | BASE TABLE |

Time: 0.007505084 seconds. 3 rows.

Enter a query to display the longest taxi trips:

sql> SELECT trip_distance, total_amount FROM taxi_trips ORDER BY trip_distance DESC LIMIT 10;


| trip_distance | total_amount |
| 312722.3      | 22.15        |
| 97793.92      | 36.31        |
| 82015.45      | 21.56        |
| 72975.97      | 20.04        |
| 71752.26      | 49.57        |
| 59282.45      | 33.52        |
| 59076.43      | 23.17        |
| 58298.51      | 18.63        |
| 51619.36      | 24.2         |
| 44018.64      | 52.43        |

Time: 0.015596458 seconds. 10 rows.

⚙️ Runtime Container Deployment

Using the Docker image locally:

docker pull spiceai/spiceai

In a Dockerfile:

from spiceai/spiceai:latest

Using Helm:

helm repo add spiceai
helm install spiceai spiceai/spiceai

🏎️ Next Steps

You can use any number of predefined datasets available from the Cloud Platform in the Spice runtime.

A list of publicly available datasets from can be found here:

In order to access public datasets from, you will first need to create an account with by selecting the free tier membership.

Navigate to and create a new account by clicking on Try for Free.


After creating an account, you will need to create an app in order to create to an API key.


You will now be able to access datasets from For this demonstration, we will be using the dataset.

Step 1. Log in and authenticate from the command line using the spice login command. A pop up browser window will prompt you to authenticate:

spice login

Step 2. Initialize a new project and start the runtime:

# Initialize a new Spice app
spice init spice_app

# Change to app directory
cd spice_app

# Start the runtime
spice run

Step 3. Configure the dataset:

In a new terminal window, configure a new dataset using the spice dataset configure command:

spice dataset configure

You will be prompted to enter a name. Enter a name that represents the contents of the dataset

dataset name: (spice_app) eth_recent_blocks

Enter the description of the dataset:

description: eth recent blocks

Enter the location of the dataset:


Select y when prompted whether to accelerate the data:

Locally accelerate (y/n)? y

You should see the following output from your runtime terminal:

2024-06-03T23:25:59.514395Z  INFO runtime: Dataset eth_recent_blocks registered (, acceleration (arrow, 10s refresh), results cache enabled.
2024-06-03T23:25:59.514624Z  INFO runtime::accelerated_table::refresh: Loading data for dataset eth_recent_blocks
2024-06-03T23:26:00.758813Z  INFO runtime::accelerated_table::refresh: Loaded 143 rows (6.22 MiB) for dataset eth_recent_blocks in 1s 244ms.

Step 4. In a new terminal window, use the Spice SQL REPL to query the dataset

spice sql
SELECT number, size, gas_used from eth_recent_blocks LIMIT 10;

The output displays the results of the query along with the query execution time:

| number   | size   | gas_used |
| 19281345 | 400378 | 16150051 |
| 19281344 | 200501 | 16480224 |
| 19281343 | 97758  | 12605531 |
| 19281342 | 89629  | 12035385 |
| 19281341 | 133649 | 13335719 |
| 19281340 | 307584 | 18389159 |
| 19281339 | 89233  | 13391332 |
| 19281338 | 75250  | 12806684 |
| 19281337 | 100721 | 11823522 |
| 19281336 | 150137 | 13418403 |

Time: 0.004057791 seconds. 10 rows.

You can experiment with the time it takes to generate queries when using non-accelerated datasets. You can change the acceleration setting from true to false in the datasets.yaml file.

📄 Documentation

Comprehensive documentation is available at

🔨 Upcoming Features

🚀 See the Roadmap to v1.0-stable for upcoming features.

🤝 Connect with us

We greatly appreciate and value your support! You can help Spice in a number of ways:

⭐️ star this repo! Thank you for your support! 🙏