AxonBase

Multi-model database written in Java 21+, combining document, graph, relational, key-value, full-text, vector, and geospatial models in a single engine with a unified query language (AxonQL), HTTP and WebSocket server, and SDKs in 8 languages.

• Multi-model engine: documents, graph edges, relational tables, key-value, time series, geospatial (GeoJSON), and vectors in one ACID-compliant engine
• AxonQL: SQL-like query language with graph traversal (->/<-), subqueries, aggregations, 130+ built-in functions, and EXPLAIN/EXPLAIN ANALYZE
• Native sagas: BEGIN SAGA, COMMIT SAGA, CANCEL SAGA with correlation IDs, automatic compensation, remote rollback via Database Link, and ledger tracking
• AI Audit: LLM-generated security rules classify every query as SAFE, WARNING, or DANGER before execution — no production data reaches the AI provider
• Data Rules: row-level security with predicate ANDing and automatic field masking, enforced across all queries without application changes
• Search engines: BM25 full-text with configurable analyzers, HNSW vector search (cosine/Euclidean/Manhattan), geo spatial with R-Tree, graph traversal, and columnar indexes for direct aggregation (count/sum/avg/min/max)
• Multi-model indexes: unique, count, full-text (SEARCH ANALYZER), HNSW vector, GEO spatial, and COLUMNAR analytical
• MVCC transactions: snapshot isolation, savepoints, optimistic retry, phantom protection via read ranges, and VersionConflictException
• Raft consensus: TCP Raft with leader election, quorum, dynamic membership (join/leave via joint consensus), catch-up, snapshots, and consistent hash sharding
• Authentication: JWT (HS256) with root/namespace/database scope, certificate-based (ICP-Brasil compatible), mTLS, and Data Rules from JWT claims
• Protocols: JSON-RPC over WebSocket and HTTP, REST (/sql, /table), GraphQL, MCP (Model Context Protocol), and JDBC
• Observability: Prometheus metrics (/metrics), health/ready/status endpoints, structured JSON logging, S3 backup with AES-256-GCM encryption
• Configuration: file-based (axonbase.conf), environment variables (AXON_*), and CLI flags with precedence chain

All 8 SDKs expose the same public API with feature parity, tested independently:

Every SDK implements: connect, use, signin, authenticate, query, select, create, insert, update, upsert, delete, relate, begin/commit/cancel, kv_get/kv_set/kv_del/kv_scan, live/kill, certificateBegin/certificateComplete, mTLS, typed RPC errors, Migrator (.axql), and SagaTransaction/SagaParticipantTransaction.

• Only database with native sagas — distributed transactions with correlation, compensation, and remote rollback, no Kafka or outbox pattern needed
• Only database with AI Audit — LLM-generated rules govern every query, no separate proxy or middleware required
• Only multi-model database with columnar indexes — direct aggregation without loading documents, traditionally a separate ClickHouse workload
• MCP with read-only defaults — agents discover tables safely via axon_select and axon_info; arbitrary query (axon_query) is not exposed
• 8 SDKs with verified parity — every SDK passes the same feature matrix across all 8 languages

# Full build
mvn install -DskipTests

# Run all Java tests
mvn test

# Run SDK-specific tests
cd axonbase-sdk-nodejs && npm test
cd axonbase-sdk-rust && cargo test
cd axonbase-sdk-go && go test ./...
cd axonbase-sdk-ruby && ruby -Ilib test/client_test.rb
cd axonbase-sdk-python && pip install -e ".[test]" && pytest
docker build --target php-test --file Dockerfile.sdk-tests .
docker build --target dotnet-test --file Dockerfile.sdk-tests .

# Run with Docker
docker run -d --name axonbase -p 8000:8000 axonbase/axonbase:latest

# Execute AxonQL via HTTP
curl -X POST http://127.0.0.1:8000/sql \
  -H 'Axon-Ns: test' -H 'Axon-Db: dev' \
  --data "CREATE person CONTENT {name: 'Ana', age: 30}"

curl -X POST http://127.0.0.1:8000/sql \
  -H 'Axon-Ns: test' -H 'Axon-Db: dev' \
  --data "SELECT * FROM person"

// Node.js SDK
import { Axon } from '@axonbase/sdk';
const axon = await Axon.connect('ws://127.0.0.1:8000/rpc/ws');
await axon.use('test', 'dev');
await axon.create('person', { name: 'Ana', age: 30 });
const rows = await axon.select('person');

Use DESCRIBE <table> to inspect a table definition. The result is a row set that JDBC clients can display directly. Every row contains the table name, its schema mode, and field metadata.

DESCRIBE person;

For a SCHEMAFULL table, DESCRIBE returns fields declared with DEFINE FIELD. For a SCHEMALESS table, it also infers fields and types from stored documents. Declared fields take precedence over inferred metadata.

name    schema      field         type
person  SCHEMALESS  age           int
person  SCHEMALESS  id            string
person  SCHEMALESS  municipio_id  string
person  SCHEMALESS  name          string

• Financial services: sagas for distributed ledger transactions, Data Rules for PII masking, AI Audit for Bacen/LGPD compliance, ICP-Brasil certificates
• AI agents: MCP for agent connectivity, AI Audit for query governance, Data Rules for row-level access control, vector + full-text hybrid search for RAG
• E-commerce: BM25 product search, HNSW recommendations, graph traversal for related items, geo for local inventory, live queries for real-time inventory updates
• Healthcare: Data Rules for patient PII protection, AI Audit for access governance, full-text search over clinical notes, sagas for distributed care workflows
• IoT: columnar indexes for real-time sensor aggregation, geo for fleet tracking, live queries for telemetry streaming, Raft cluster for high availability

Álvaro Brito is a Backend & AI Engineer with more than 10 years of Java experience, architecting secure, scalable microservices and delivering modern LLM and RAG solutions. He combines deep proficiency in Node.js, TypeScript, Python, Express, and PostgreSQL with infrastructure-grade execution on AWS, Docker, Terraform, and event-driven architectures. He builds observable AI workflows with LangChain and Langfuse, instrumenting LLM evaluations and operational metrics with Prometheus and Grafana to improve quality and reliability. His work turns complex workflows into production-ready platforms through CI/CD automation, real-time orchestration, and high-throughput data flows with FastAPI and Kafka.

LinkedIn: https://www.linkedin.com/in/alvarogomes/

MIT