AI integration (Andreas)

CrateDB is not just a real-time analytics database, it's a powerful platform to feed and interact with machine learning models, thanks to its ability to store, query, and transform structured, unstructured, and vectorized data at scale using standard SQL.

Whether you're training models, running batch or real-time inference, or integrating with AI pipelines, CrateDB offers:

• High-ingestion performance for time-series or sensor data

• Real-time queries across structured and semi-structured data

• SQL-powered transformations and filtering

• Native support for embeddings via FLOAT_VECTOR

For more details on how CrateDB handles similarity search and embeddings, see the Vector Search use case.

---

Why CrateDB for ML Use Cases?

Feature	Benefit
Real-time ingestion	Ingest millions of records per second from IoT, logs, or user behavior
Unified data	Mix structured, full-text, vector, and JSON data
FLOAT_VECTOR	Store and query high-dimensional embeddings
SQL transforms	Use SQL for preprocessing, feature extraction, and filtering
ML integration	Use CrateDB as a feature store or inference backend
Python & LangChain support	Easily connect to training and inference pipelines

---

Common Machine Learning Patterns

Feature Engineering

Use SQL to build features dynamically from raw data.

SELECT
  user_id,
  AVG(duration) AS avg_session,
  COUNT(DISTINCT page) AS page_diversity
FROM sessions
GROUP BY user_id;

Training Dataset Extraction

Efficiently extract and filter relevant training data.

SELECT *
FROM telemetry
WHERE temperature > 80
  AND error_code IS NOT NULL
  AND ts BETWEEN NOW() - INTERVAL '7 days' AND NOW();

Store Embeddings

Save vector representations for documents or entities.

CREATE TABLE article_embeddings (
  id UUID PRIMARY KEY,
  content TEXT,
  embedding FLOAT_VECTOR(384)
);

CrateDB supports high-dimensional vectors with FLOAT_VECTOR. To query these vectors for similarity-based inference, see Vector Search.

Use CrateDB as a Feature Store

Centralize your features and use them in production models.

SELECT *
FROM user_features
WHERE last_active > NOW() - INTERVAL '1 day';

---

When to Use CrateDB in ML Pipelines

Use Case	CrateDB Role
Feature Store	Store pre-computed features with SQL access
Real-Time Inference	Serve vector-based results with KNN_MATCH
Experimentation	Use SQL for fast slicing, filtering, and aggregations
Monitoring	Track model performance, drift, or input quality
Data Collection	Capture telemetry, events, logs, and raw user data

---

Architecture Examples

Model Training Pipeline

[ Ingestion (sensors, APIs) ]
           ↓
     [ CrateDB ]
  (Real-time data lake)
           ↓
[ Python / Spark / Pandas ]
  (Feature engineering, training)
           ↓
[ Model Registry / Serving ]

Real-Time Inference with Hybrid Queries

[ CrateDB ]
 - JSON filters
 - Full-text search
 - FLOAT_VECTOR support
         ↓
  [ SQL + KNN_MATCH ]
         ↓
[ Application Response ]

---

Performance Tips for ML Scenarios

Tip	Benefit
Use FLOAT_VECTOR for embeddings	Store and access high-dimensional vectors efficiently
Normalize vectors before inference	Improves ANN accuracy
Index only what you need	Reduces overhead
Use hybrid filters + vector search	Improves performance and precision
Train offline, infer online	CrateDB is ideal for live inference from pre-trained models

---

Ecosystem & Integration

Tool	Integration
Python	crate-python for pandas, scikit-learn
LangChain	Native CrateDB vector store
Jupyter	Ideal for experimentation, model development
OpenAI, Cohere, etc.	Store and search their embeddings via SQL
Kafka	Connect for real-time ingestion and prediction

---

Further Reading & Resources

• Vector Search use case

• LangChain Integration with CrateDB

• CrateDB ML Guide

• CrateDB + Embeddings Blog

• CrateDB Examples – ML