HeliosDB is a family of databases for AI applications that speak the PostgreSQL wire protocol and build vector search, Git-like branching and time-travel queries into the engine. HeliosDB Nano is the embedded edition and HeliosProxy is a PostgreSQL data-plane, both licensed Apache 2.0; HeliosDB Lite (self-hosted, for production SaaS) and HeliosDB Full (enterprise, multi-protocol) are licensed SSPL-1.0.
Last updated
| Feature | SQLite | PostgreSQL | MySQL | HeliosDB |
|---|---|---|---|---|
| Distribution | Library | Server packages and images | Server packages and images | One binary of about 36-44 MB (by platform), or the official GHCR image |
| PostgreSQL + MySQL protocol | PG only | MySQL only | Both native | |
| Vector + BM25 + Graph search | Extensions | Vector + BM25 in SQL; graph via MCP tools (source build) | ||
| Database branching | ||||
| Time-travel queries | ||||
| Encryption (AES-256) | Extension | Enterprise $ | Built-in, off until enabled with a key | |
| Embedded / zero-config | ||||
| Multi-tenant RLS | Embedded API and REPL only (Nano) |
HeliosDB combines the best features from embedded databases, enterprise databases, and vector databases—all in a single, lightweight package.
HNSW indexing and cosine, L2 and inner-product distance operators built into the engine, with Product Quantization, which is in Nano's default build. Works with any embedding model.
Git-like branching at the storage layer: CREATE DATABASE BRANCH, MERGE and DROP in SQL, with copy-on-write branches for schema testing and A/B experiments. Merges are last-writer-wins.
Query any point in your database's history with AS OF syntax. Audit compliance, debug production issues, and recover data without restoring backups.
AES-256-GCM encryption at rest (in Nano, off until you enable it in the configuration and supply a key), zero-knowledge encryption in Nano (keys are derived and data encrypted on the client; the server never sees plaintext keys), an opt-in FIPS 140-3 build mode in Nano and Lite, and row-level security (in the stock Nano server, enforced through the embedded API and the REPL; enforcement for wire and REST clients is planned).
Handle transactional and analytical workloads in one database. Columnar storage with Apache Arrow, SIMD acceleration, and auto-refreshing materialized views.
PostgreSQL and MySQL wire compatibility plus a documented SQL subset, not full semantic parity. Clients such as psql, mysql, DBeaver and pgAdmin connect, as do ORMs such as SQLAlchemy, Diesel, Prisma and SQLx. Recent releases add SCRAM-SHA-256 (RFC 5802 GS2) auth, full information_schema, CREATE / DROP DATABASE per-tenant selection, and real \d / \l / \dt psql meta-commands.
HeliosDB Nano serves PostgREST-style REST, auth and a realtime WebSocket from the same binary as the database and stores data in its data directory. The /rest/v1 table routes are a known issue in the current release. Known issue: realtime events are not delivered in the current release; tracked by the Nano team. File storage (buckets and signed URLs) and edge functions are on the HeliosDB roadmap.
Run SQL against a live HeliosDB Nano instance — no signup required.
From embedded prototyping to multi-protocol enterprise servers. Choose the edition that fits your scale.
Lightweight embedded database for AI applications, edge devices, and single-server deployments. SQLite simplicity with PostgreSQL & MySQL power.
Edge database with HeliosProxy, the connection router with multi-tier caching, a GraphQL gateway and query routing.
Enterprise database server with seven database wire protocols plus an HTTP gateway; Raft clustering and research features such as quantum-inspired optimization are planned.
From AI/ML infrastructure to genomics research, HeliosDB powers applications across every sector. Now with 14 HeliosProxy use cases covering the platform layer.
Current workarounds involve full database backups, application-level change tracking, or expensive cloud database dependencies that violate edge computing requirements.
HeliosDB Nano's Database Branching feature brings Git-like version control directly to your database, enabling zero-downtime schema migrations, A/B testing, and parallel development workflows.
HeliosDB Nano's Materialized Views with Auto-Refresh deliver pre-computed query results with intelligent, CPU-aware automatic refresh capabilities.
HeliosDB Nano provides AES-256-GCM transparent data encryption at rest, with the key read from an environment variable or a file; cloud KMS key sources are planned.
Encryption is compiled in but disabled at runtime by default and needs a configured key. Read the documentation.
HeliosDB Nano delivers an offline-first, embedded database solution purpose-built for IoT and edge computing deployments where cloud connectivity is intermittent, expensive, or unreliable.
HeliosDB Nano implements the PostgreSQL wire protocol and a large subset of PostgreSQL SQL, so psql, pgAdmin, DBeaver and most PostgreSQL drivers connect to it.
Wire compatibility is not semantic parity: Nano implements the PostgreSQL wire protocol and a large subset of PostgreSQL SQL, collation is byte-order (C) only, and row-level security is not enforced over the wire protocols. Read the documentation.
Row-level security is reachable only through the embedded API and the REPL, not over the PostgreSQL wire, MySQL wire or REST. Read the documentation.
HeliosDB Nano serves a REST API over HTTP, so applications without a PostgreSQL driver can run queries and manage the database.
Row-level security is reachable only through the embedded API and the REPL, not over the PostgreSQL wire, MySQL wire or REST. Read the documentation.
AI agents require persistent, semantically-searchable memory to maintain context across conversations, learn from interactions, and provide personalized experiences.
Retrieval-Augmented Generation (RAG) systems require fast, accurate semantic search to retrieve relevant context before LLM inference.
The Model Context Protocol (MCP) enables AI assistants like Claude to interact with external tools and data sources through a standardized interface.
Row-level security is reachable only through the embedded API and the REPL, not over the PostgreSQL wire, MySQL wire or REST. Read the documentation.
HeliosDB Nano enables e-commerce platforms to deliver personalized product discovery combining search + recommendations + inventory in a single embedded database.
HeliosDB Lite runs vector similarity search in SQL, with VECTOR columns, HNSW indexes and distance operators next to ordinary data.
HeliosDB Lite queries earlier states of a table with AS OF TIMESTAMP, AS OF TRANSACTION or AS OF SCN.
HeliosDB Lite creates copy-on-write database branches for development, testing and migration rehearsals, then merges or drops them in SQL.
HeliosDB Lite stores pre-computed query results in materialized views refreshed with SQL; automatic background refresh is planned.
Planned: automatic background refresh of materialized views in the Lite server is on the roadmap; refresh them with REFRESH MATERIALIZED VIEW today. Read the documentation.
HeliosDB Lite runs as a local SQL server on IoT gateways, collecting and querying sensor data on site.
Built-in audit logging for HeliosDB Lite, an append-only SQL-queryable record of schema, data and login events, is planned.
Planned: built-in audit logging in the Lite server is on the roadmap. Read the documentation.
HeliosDB Lite shrinks large tables with per-column storage modes: dictionary encoding, columnar batches and content-addressed deduplication.
HeliosDB Lite speaks the PostgreSQL wire protocol, so standard PostgreSQL drivers, ORMs and psql connect to it directly.
HeliosDB Lite stores JSON documents in JSONB columns and queries them with PostgreSQL operators next to relational data.
HeliosDB Lite isolates tenants in shared tables with row-level security enforced for PostgreSQL wire sessions and REST requests.
Superusers bypass row-level security; give each tenant its own non-superuser role and start the server with SCRAM authentication so clients cannot claim another role. Read the documentation.
HeliosDB Lite supports PostgreSQL LISTEN, UNLISTEN and NOTIFY over the wire protocol for event notifications without a broker.
HeliosDB Lite explains query plans with EXPLAIN and EXPLAIN ANALYZE, including cost estimates and an OLTP or OLAP classification.
Automatic index recommendations for HeliosDB Lite are planned; EXPLAIN and EXPLAIN ANALYZE show where an index would help today.
Planned: automatic index recommendations are on the roadmap. Read the documentation.
HeliosDB Lite serves a JWT-authenticated REST API for SQL queries, statements, branches and schema inference over HTTP.
HeliosDB Lite keeps embeddings next to business data, so AI applications combine similarity search and SQL filters in one query.
HeliosDB Lite runs multi-tenant SaaS data in shared tables, with one role per tenant and row-level security on wire and REST.
Superusers bypass row-level security; give each tenant its own non-superuser role and start the server with SCRAM authentication so clients cannot claim another role. Read the documentation.
HeliosDB Lite runs dashboard aggregations on live operational tables, with columnar storage for the analytical columns.
HeliosDB Lite keeps sensor time series and pattern embeddings in one table, combining columnar aggregation with vector similarity.
HeliosDB Lite combines PostgreSQL-style full-text predicates and vector similarity in one SQL query for documentation search.
HeliosDB Lite powers product search by meaning, with embeddings, an HNSW index and price or category filters in one SQL query.
HeliosProxy isolates tenants that share one HeliosDB Lite database by injecting a tenant predicate into their queries.
HeliosProxy injects the tenant filter only into queries on the tables listed in tenant_tables, and it identifies the tenant from a connection attribute (application_name by default). Read the documentation.
HeliosProxy pools many SaaS client connections onto fewer HeliosDB Lite backend connections in session, transaction or statement mode.
HeliosDB Lite limits each role's request rate with CREATE ROLE ... RATE_LIMIT, so tenants with their own roles get separate budgets.
HeliosDB Lite runs as one self-contained server process at edge sites, serving SQL locally over the PostgreSQL wire and REST.
HeliosDB Lite runs vector similarity search on the edge device itself, with HNSW indexes and SIMD-accelerated distance computation.
HeliosDB Lite runs as one database container per microservice, with its data on a volume and a built-in health check.
HeliosProxy keeps client sessions alive across a primary failover by reconnecting, restoring session settings and optionally replaying the open transaction.
HeliosDB Lite replication needs a binary built with the ha-tier1 feature; promoting the standby is a manual or external step. Read the documentation.
HeliosProxy takes failing HeliosDB Lite nodes out of rotation with a per-node circuit breaker and probes them before returning them.
Semantic caching of RAG retrievals in HeliosProxy is planned; its query cache serves repeated read queries today.
Planned: semantic (similarity-matched) result caching in HeliosProxy is on the roadmap. Read the documentation.
Conversation-context caching for chatbots is planned; HeliosDB Lite keeps conversation history in ordinary tables today.
Planned: conversation-context caching is on the roadmap. Read the documentation.
Tool-result caching for AI agents is planned; agents can persist their sessions and tool outputs in HeliosDB Lite tables today.
Planned: tool-result caching for AI agents is on the roadmap. Read the documentation.
HeliosDB Lite takes over PostgreSQL workloads: load a cleaned pg_dump with psql and keep the application's PostgreSQL driver.
HeliosDB Lite traces query timing per session and reports percentile latencies and the slowest statements in SQL.
HeliosProxy serves repeated read queries for HeliosDB Lite from in-process cache tiers with commit-aware invalidation.
HeliosProxy rewrites SQL on its way to HeliosDB Lite with ordered rules that add predicates, limits or table redirects.
HeliosProxy routes writes to the primary, reads to standbys and analytical queries to a dedicated HeliosDB Lite node.
HeliosProxy gives a HeliosDB Lite primary and warm standby one endpoint: writes and transactions run on the primary, side-effect-free reads on healthy standbys.
HeliosDB Lite replication needs a binary built with the ha-tier1 feature; standby write forwarding covers simple-protocol statements only. Read the documentation.
HeliosDB Lite removes duplicates in SQL with SELECT DISTINCT and counts unique values with COUNT(DISTINCT).
HeliosDB Lite computes rankings, running totals and row-to-row comparisons with SQL window functions over PARTITION BY and ORDER BY.
HeliosDB Full serves PostgreSQL, MySQL and Oracle clients from one server over shared tables, with more wire-protocol listeners and an HTTP gateway in the same process.
MongoDB, Cassandra and SQL Server clients can connect and authenticate, and reading and writing HeliosDB tables through them is planned, as are the GraphQL, Snowflake and DB2 interfaces. Read the documentation.
HeliosDB Full serves RAG retrieval through its HTTP gateway (documents, vector stores, hybrid search, embeddings and chat); SQL vector indexes and SQL search functions are planned.
Planned: SQL vector indexes, SQL search and RAG functions and natural-language-to-SQL are on the roadmap; use the HTTP gateway's document, vector-store and chat routes today. Read the documentation.
Serverless scale-to-zero and autoscaling for HeliosDB Full are planned; the current release runs as a single server.
Planned: serverless scale-to-zero and autoscaling are on the roadmap. Read the documentation.
Homomorphic encryption, differential privacy and federated learning for HeliosDB Full are planned.
Planned: homomorphic encryption, differential privacy and federated learning are on the roadmap. Read the documentation.
Apache Iceberg and Delta Lake tables with a unified catalog for HeliosDB Full are planned.
Planned: Apache Iceberg and Delta Lake tables and a unified catalog are on the roadmap. Read the documentation.
Quantum-inspired query optimization is a research item for HeliosDB Full; it is planned and not in the current release.
Planned: quantum-inspired query optimization is a research item on the roadmap. Read the documentation.
Autonomous self-healing agents for HeliosDB Full are planned.
Planned: autonomous self-healing agents are on the roadmap. Read the documentation.
AI-driven data-growth forecasting and capacity planning for HeliosDB Full are planned.
Planned: data-growth forecasting and capacity recommendations are on the roadmap. Read the documentation.
Offline-first edge sync for HeliosDB Full is planned.
Planned: offline-first edge sync is on the roadmap. Read the documentation.
Natural-language schema design for HeliosDB Full is planned.
Planned: natural-language schema design is on the roadmap.
In-database time-series forecasting for HeliosDB Full is planned.
Planned: in-database time-series forecasting is on the roadmap. Read the documentation.
Continuous streaming queries over Kafka topics for HeliosDB Full are planned.
Planned: continuous streaming queries and Kafka integration are on the roadmap.
Energy-aware and carbon-aware workload scheduling for HeliosDB Full is a planned research item.
Planned: energy-aware and carbon-aware scheduling is a research item on the roadmap. Read the documentation.
Conversational BI with multi-turn natural-language questions for HeliosDB Full is planned.
Planned: conversational BI and natural-language-to-SQL are on the roadmap. Read the documentation.
A blockchain-backed, tamper-evident audit trail for HeliosDB Full is planned.
Planned: the blockchain-backed audit trail and data lineage are on the roadmap. Read the documentation.
A no-code, natural-language database builder for HeliosDB Full is planned.
Planned: a natural-language database builder and natural-language-to-SQL are on the roadmap. Read the documentation.
Approximate query processing with confidence bounds for HeliosDB Full is planned.
Planned: approximate and probabilistic queries are on the roadmap. Read the documentation.
OpenTelemetry trace and metric export for HeliosDB Full is planned.
Planned: OpenTelemetry trace and metric export is on the roadmap. Read the documentation.
A FinOps cost optimizer for HeliosDB Full is planned.
Planned: cost analytics and optimization recommendations are on the roadmap.
Automated migration with schema mapping, query translation and live cutover for HeliosDB Full is planned; per-database migration guides cover the steps today.
Planned: automated schema mapping, query translation and live-replication cutover are on the roadmap. Read the documentation.
Workload-driven materialized view selection and maintenance for HeliosDB Full is planned.
Planned: automatic materialized view selection, incremental refresh and query rewriting are on the roadmap.
Distributed deadlock detection and prevention for HeliosDB Full is planned together with clustering.
Planned: distributed deadlock detection is on the roadmap together with clustering. Read the documentation.
Change data capture with Kafka and Kinesis delivery for HeliosDB Full is planned.
Planned: CDC connectors for Kafka and Kinesis are on the roadmap.
Automated zero-downtime schema evolution for HeliosDB Full is planned; the schema evolution runbook covers additive changes today.
Planned: online DDL with shadow tables, traffic validation and automatic rollback is on the roadmap. Read the documentation.
Statistics-driven, adaptive cardinality estimation for HeliosDB Full is planned.
Planned: automatic statistics collection and adaptive cardinality estimation are on the roadmap.
Optimizer hints in SQL comments (index, join order, parallelism, cache) for HeliosDB Full are planned.
Planned: optimizer hints are on the roadmap.
HeliosProxy's query-cache feature serves repeated read results from two in-process tiers, an LRU hot cache and a warm cache keyed by normalized query fingerprints, with TTL expiry and commit-aware invalidation; a semantic tier is planned.
HeliosProxy's WASM plugin runtime runs sandboxed WebAssembly plugins on the pre_query, post_query and route hooks; it is built with the wasm-plugins feature.
HeliosProxy's POST /api/replay re-executes the committed transactions its recovery journal holds for a time window, either best effort or in commit order stopping at the first failure.
HeliosProxy includes a zero-downtime PostgreSQL major-version upgrade orchestrator whose validation stage checks row-count parity.
HeliosProxy's anomaly-detection feature flags SQL injection payloads, credential-stuffing bursts and novel query shapes at the proxy and lists the events at GET /anomalies.
Every query record embeds the SHA-256 of the previous record's bytes; auditors recompute the chain and any modified record breaks the link.
Plugins ship as .tar.gz OCI artefacts with versioned manifest.json + signed plugin.wasm.
HeliosProxy's POST /api/shadow runs a query against the primary and a shadow backend in parallel and compares order-independent result digests, bounded by max_rows and max_bytes.
Per-tenant minute / hour / day query-cost budgets.
Three HeliosProxy plugins for AI traffic: ai-classifier detects LLM-generated SQL, token-budget gates cost per agent and model, and llm-guardrail refuses dangerous SQL such as DROP or a missing WHERE.
Per-role substring-based column masking with whole-word matcher and idempotence detection (skips already-rewritten patterns).
HeliosProxy's residency-router plugin routes each user's queries only to replicas tagged with the user's region and answers a cross-region read with a PostgreSQL error.
HeliosProxy's pgvector-router plugin routes pgvector top-K queries (<->, <#> or <=> in ORDER BY) to a vector-tagged replica; the same query without ORDER BY goes to the default route.
Install, connect, and run your first query — three steps, any language.
curl -sSf https://install.heliosdb.com | shAlso via: cargo install heliosdb-nano, pip install heliosdb-nano-embedded
# In-memory REPL
heliosdb-nano repl --memory
# Or run a server and use any PostgreSQL client
heliosdb-nano start --data-dir ./mydb &
psql -h localhost -p 5432 -d postgresCREATE TABLE docs (
id SERIAL PRIMARY KEY,
content TEXT,
embedding VECTOR(768)
);// Embedded: open a data directory in-process
use heliosdb_nano::EmbeddedDatabase;
let db = EmbeddedDatabase::new("./myapp")?;
db.execute("CREATE TABLE users (
id SERIAL PRIMARY KEY,
name TEXT,
embedding VECTOR(768)
)")?;
// Semantic search in one query
let results = db.query_params("
SELECT name, embedding <=> $1::vector(768) AS distance
FROM users ORDER BY distance LIMIT 10
", &[query_vec])?;# pip install heliosdb-nano-embedded (in-process, no server)
from heliosdb_nano import EmbeddedDatabase
db = EmbeddedDatabase("./myapp.db")
db.execute("""
CREATE TABLE documents (
id SERIAL PRIMARY KEY,
content TEXT,
embedding VECTOR(3)
)
""")
db.execute("INSERT INTO documents (content, embedding) VALUES "
"('apples', '[1,0,0]'), ('pears', '[0.9,0.1,0]'), ('cars', '[0,0,1]')")
# Nearest neighbours by cosine distance; rows come back as dicts
rows = db.query(
"SELECT content, embedding <=> $1 AS distance "
"FROM documents ORDER BY distance LIMIT 2",
["[1,0,0]"],
)// npm install pg — Nano speaks the PostgreSQL wire protocol
import pg from 'pg';
// a local `heliosdb-nano start --data-dir ./mydb` server
const db = new pg.Client({ host: 'localhost', port: 5432, user: 'postgres', database: 'postgres' });
await db.connect();
await db.query(`
CREATE TABLE documents (
id SERIAL PRIMARY KEY,
content TEXT,
embedding VECTOR(3)
)
`);
await db.query(`INSERT INTO documents (content, embedding) VALUES
('apples', '[1,0,0]'), ('pears', '[0.9,0.1,0]'), ('cars', '[0,0,1]')`);
// Nearest neighbours by cosine distance
const { rows } = await db.query(
'SELECT content, embedding <=> $1 AS distance FROM documents ORDER BY distance LIMIT 2',
['[1,0,0]']
);# Install and start in seconds
curl -sSf https://install.heliosdb.com | sh
# Interactive REPL
heliosdb-nano repl --memory
# Or start as a server
heliosdb-nano start --data-dir ./mydb --port 5432
# Connect with any PostgreSQL client
psql -h localhost -p 5432 -d postgres
# Branch your database like Git (SQL, from psql or the REPL)
psql -h localhost -d postgres -c "CREATE BRANCH feature_auth"
psql -h localhost -d postgres -c "SHOW BRANCHES"Rust, Python, TypeScript, Go, Java, C#, Ruby
Axum, FastAPI, Express, Spring Boot, Rails
LangChain, LlamaIndex, OpenAI, AutoGen, CrewAI
Embedded, Server, In-Memory, Hybrid
The database market is $100B+ and growing. AI applications demand new capabilities — native vector search, semantic caching, and intelligent query optimization — that legacy databases can't deliver without bolted-on extensions.
Source-available core (SSPL-1.0 for Lite/Full, Apache 2.0 for Nano and HeliosProxy) with enterprise features driving revenue. Three-tier product architecture enables land-and-expand from developer adoption to enterprise contracts.
Database engineers, systems programmers, and developer advocates building the database developers deserve. Our aim is to make advanced database capabilities accessible to every developer.
Contact Us →HeliosDB Nano on HeliosDB Cloud, the managed service. Cloud accounts use a business email address and start at the Verified level; Starter and Pro have no trial period. Compare all editions →
BaaS API signup for agents and scripts, no verification needed.
Verify your email to unlock larger quotas. Cloud accounts (business email) start here.
Complete a product feedback session.
One quick-start API call provisions a database instantly.
POST /api/v1/quick-startConfirm your email to lift your quotas.
Complete a feedback session for the highest free quotas.
For small teams and growing projects.
For production workloads and power users.
For dedicated isolation and custom scale.
Instant database provisioning for lowcode, nocode, and AI agent workflows. Free tier with branching, time travel and vector search, no credit card.
curl -X POST https://cloud.heliosdb.com/api/v1/quick-start \
-H "Content-Type: application/json" \
-d '{
"email": "dev@myapp.com",
"password": "secure123",
"username": "dev",
"org_name": "MyApp",
"db_name": "production"
}'
// Response: token, org_id, db_id, api_key, endpoint URLsCreate branches from any point in time. Test schema changes, A/B test data, merge when ready.
Per-plan quotaQuery data as it existed at any point in the past. Undo mistakes, audit changes, debug issues.
Free: 7-day windowNative HNSW indexing and BM25-ranked full-text search. Built for RAG pipelines.
Included freeScoped hdb_ keys for each database. Embed safely in frontend apps and agent configs.
AI agents can discover all available tools with a single request. We expose both MCP (Model Context Protocol) and OpenAPI 3.1 specs — any framework can integrate automatically. MCP calls authenticate with your database API key.
POST /api/v1/mcp
Authorization: Bearer hdb_...
{"jsonrpc":"2.0",
"method":"tools/list",
"id":1}
// execute_sql, create_branch,
// query_at, vector_search, ...
// API key from quick-start or
// the console; no key: 401Five products. Start with Nano for free or explore the full platform.