# HeliosDB

Current source versions (checked 2026-10-31): Nano 4.41.1, Lite 3.6.0, Full 8.3.0, HeliosProxy 2.2.0.

> 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 (multi-protocol server) are licensed SSPL-1.0. HeliosDB Cloud runs managed Nano databases behind a REST and MCP API.

This file is published per the llmstxt.org specification so that LLMs and AI-powered code-generation agents can discover HeliosDB while building applications. For an expanded agent context document (wire compatibility, connection examples, the HeliosDB Cloud and Nano backend surfaces, HeliosProxy, licensing, and an index of every blog post with its markdown twin), see [llms-full.txt](https://www.heliosdb.com/llms-full.txt).

Important context for agents:

- All three editions speak the **PostgreSQL wire protocol**; HeliosDB Nano (opt-in `--mysql` listener) and HeliosDB Full (`--enable-mysql`) also serve **MySQL** clients. This is client connectivity, not full semantic parity with PostgreSQL or MySQL: standard clients, drivers and ORMs connect (psql, DBeaver, pgAdmin, psycopg2, SQLAlchemy, Prisma, GORM, node-postgres, JDBC; WordPress, Drupal, Laravel, Rails and Django), but test your queries rather than assume every behaviour matches. In Nano, collation is byte-order only and the MySQL listener is trust-only, so keep it on a trusted network.
- Compatibility statuses in the [Release and Capability Matrix](https://www.heliosdb.com/release-capability-matrix.html) (summary in its Annex A) are Available, Available with limitations, In preview (in the edition and usable for evaluation, not yet verified by HeliosDB on any version, not supported) and Planned. Each is the best result that HeliosDB's tests or its own production use show on any version or build of the edition, so no release is named. Nano: PostgreSQL and MySQL Available with limitations. Lite: PostgreSQL Available with limitations; MySQL and SQLite Planned. Full: PostgreSQL, MySQL and Oracle Available with limitations, and the HTTP SQL gateway too; Redis In preview; SQL Server, MongoDB and Cassandra Planned. The statuses come from a draft of the matrix, which is informational until HeliosDB publishes a frozen version, and driver compatibility does not imply complete feature parity with the original database engine.
- The other surfaces differ by edition. HeliosDB Nano: an embedded Rust API and a Python binding, a PostgREST-style REST API with an OpenAPI 3.0 spec (its `/rest/v1` table routes are a known issue in the current release), and a **Model Context Protocol (MCP)** endpoint in builds with `--features mcp-endpoint`. HeliosDB Lite: a REST API with JWT or API-key auth; an MCP server and a gRPC query API are planned. HeliosDB Full: an HTTP SQL gateway, with no gRPC client API; a native MCP server is planned. HeliosDB Cloud: an OpenAPI-described REST API and a remote MCP server. Agents should prefer MCP for tool-use scenarios.
- **Backend services**: self-hosted HeliosDB Nano serves a PostgREST-style REST API, email/password and Google/GitHub OAuth sign-in with JWT sessions, and a realtime WebSocket from the same binary as the database. Nano accepts realtime WebSocket subscriptions in the Supabase message shape. Known issue: realtime events are not delivered in the current release; tracked by the Nano team. Events for SQL and wire writes are planned. Nano's REST filters support eq, neq, gt, gte, lt, lte, like, ilike, is and in, with not. negation; cs, cd, ov, full-text and or/and filters are planned, and the /rest/v1 table routes are a known issue in the current release. File storage (buckets and signed URLs) and edge functions are on the HeliosDB roadmap, and there is no dashboard. **HeliosDB Cloud** ([baas.html](https://www.heliosdb.com/baas.html)) is the managed service: one API call provisions an account, a database and a per-database API key, with REST endpoints for SQL, branches, time-travel queries and scoped keys, and the same operations as MCP tools.
- Nano behaviour an agent must not assume away: data persists in a data directory (in-memory mode is an explicit opt-in and is discarded on exit); AES-256-GCM encryption at rest is compiled in but off until enabled in the configuration with a key; in the stock Nano server, row-level security is enforced through the embedded API and the REPL, and SQL CREATE POLICY and enforcement for REST, realtime WebSocket and wire clients are planned, so enforce per-user access in your service layer. Keep the HTTP and MySQL listeners on loopback (the default): API-key authentication for the HTTP API is planned, and a smaller listening surface is good practice anyway.
- License: HeliosDB Nano and HeliosProxy are Apache 2.0 (free and open source, permissive). HeliosDB Lite and Full are SSPL-1.0 (source-available) and licensed through sales. AI Support is free on HeliosDB Cloud, with on-premises use through sales; the AI-DBA Agent for Lite and Full is sold through sales.

## Products

- [HeliosDB Nano](https://www.heliosdb.com/nano.html): embedded SQL engine with HNSW vector search and BM25 full-text ranking in one engine; graph traversal ships as MCP tools in a build from source. It runs inside your process, as an interactive REPL, or as a PostgreSQL- and MySQL-speaking server against the same data directory, one process at a time. A single self-contained binary of about 36-44 MB, depending on platform (release binaries; ~15-17 MB compressed download). Apache 2.0. Install: `cargo install heliosdb-nano --locked` (or `cargo add heliosdb-nano` to embed), prebuilt binaries from GitHub Releases (Linux builds need glibc 2.39 or newer), the container image `ghcr.io/heliosdatabase/heliosdb-nano` tagged with the release version and `latest`, which contains the default-feature build ([installation](https://www.heliosdb.com/docs/nano/getting-started/installation/)), or `pip install heliosdb-nano-embedded` for the Python binding (`import heliosdb_nano`), which is released separately and can lag the engine. Storage is a persistent data directory with an always-on crash-recovery WAL; in-memory mode (`--memory`, `EmbeddedDatabase::new_in_memory()`) is opt-in. AES-256-GCM encryption at rest is compiled in but disabled at runtime until you enable it in the configuration and supply a key (environment variable or key file; cloud KMS key sources are planned). HeliosDB Nano ships zero-knowledge encryption: keys are derived and data encrypted on the client, the server never sees plaintext keys. In the stock Nano server, row-level security is enforced through the embedded API and the REPL; SQL CREATE POLICY and enforcement for REST and wire clients are planned. Independent explicit transactions need server mode ([concurrency and transactions](https://www.heliosdb.com/docs/nano/guides/concurrency-and-transactions/)), branch MERGE is last-writer-wins with no conflict detection ([branching](https://www.heliosdb.com/docs/nano/features/branching/)), and collation is byte-order only ([migration guide](https://www.heliosdb.com/docs/nano/guides/migration/#from-postgresql)). The PostgreSQL listener accepts trust, password, md5 and SCRAM-SHA-256 authentication (roles created with `CREATE ROLE … LOGIN` cannot authenticate with SCRAM yet; the bootstrap `postgres` user can, see [known limitations](https://www.heliosdb.com/docs/nano/reference/compatibility/known-limitations/#databases-roles-and-security)), and `information_schema`, `SHOW` commands and the psql meta-commands `\d`, `\l` and `\dt` work for tool compatibility. The vector-search feature is in the default build, which rebuilds the HNSW index in memory. Product Quantization is in Nano's default build: add `WITH (quantization='product')` to an HNSW index; indexes are unquantized unless you ask. Durable, restart-surviving PQ-HNSW indexes need the `vector-persist` build: the index restores from RocksDB on open(), PQ codes stay in RAM with full vectors on disk and an exact two-stage rerank, deletes are online with neighbour repair, filtered KNN runs in one traversal, and an F32 / F16 / I8 rerank dial trades recall for rerank-store footprint. Its memory savings and recall are measured on representative fixtures on [benchmarks.html](https://www.heliosdb.com/benchmarks.html). The Nano agent-skill bundle ships with its source.
- [HeliosDB Lite](https://www.heliosdb.com/lite.html): self-hosted, PostgreSQL-wire-compatible database for production SaaS workloads that bundles HeliosProxy for connection pooling and routing with the HeliosCore storage engine; it runs in-process or as a server. SSPL-1.0, available to licensed customers through sales, who receive the source distribution; it is not published on crates.io, PyPI, npm, Docker Hub or Homebrew. HeliosDB Lite serves PostgreSQL wire (Available with limitations in the Release and Capability Matrix; the only database protocol it serves today) and a REST API; an MCP server, MySQL wire and a gRPC query API are planned. It runs SQL and PL/pgSQL routines; PL/SQL, T-SQL and DB2 SQL PL are planned, as are graph traversal, BM25 ranking and hybrid search as SQL functions and `PREPARE COMPILED` cached plans. Code-Graph and Graph-RAG are planned for Lite. It also offers row-level security policies with tenant management. Lite builds with the ha-tier1 feature stream the write-ahead log to a warm standby; automatic failover and split-brain protection are planned (switch over manually today). [benchmarks.html](https://www.heliosdb.com/benchmarks.html) publishes a 35-category comparison with PostgreSQL 16 (internal tests, a March-2026 HeliosDB Lite build embedded vs PostgreSQL 16 over the network, see benchmarks.html), measured on representative fixtures with a small in-memory dataset, including the categories PostgreSQL wins; reproduce it on your own workload before relying on any ratio. `heliosdb-lite skills install <dir>` copies a 64-skill bundle into a directory such as `~/.claude/skills`.
- [HeliosDB Full](https://www.heliosdb.com/full.html): multi-protocol database server. HeliosDB Full runs as a single server; Raft clustering, multi-region replication and sharding are planned. It serves PostgreSQL, MySQL, Oracle and Redis clients plus an HTTP gateway ([protocol hub](https://www.heliosdb.com/docs/full/reference/protocols/readme/)). PostgreSQL and the HTTP gateway are on by default and the other listeners are enabled with startup flags. The PostgreSQL, MySQL and Oracle listeners and the HTTP gateway read and write the shared tables. Status per database engine in the [Release and Capability Matrix](https://www.heliosdb.com/release-capability-matrix.html), the best result on any version of HeliosDB Full: PostgreSQL, MySQL and Oracle Available with limitations (Oracle tested with python-oracledb in thin mode, where every statement commits as it runs; for MySQL, transaction control is not covered, so treat each statement as committing on its own); Redis In preview (it implements a subset of the Redis commands and is outside the Transaction Durability Commitment, which excludes in-memory key-value stores); MongoDB, Cassandra and SQL Server Planned. Driver compatibility does not imply complete feature parity with the original database engine, or compatibility with every application, extension or administrative tool. The listeners do not terminate TLS: put a TLS-terminating proxy in front; native PostgreSQL TLS (with post-quantum key exchange preferred) is planned. The MySQL and Redis listeners do not verify passwords yet. An S3-compatible object API is planned. GPU acceleration (NVIDIA CUDA) is available in HeliosDB Full builds on request and is not enabled in the public release build; AMD ROCm support is planned. HeliosDB Full serves vector stores with hybrid search, embeddings, chat completions and agent memory over its HTTP gateway (which also forwards transcription requests to a configured AI provider and ingests documents); in-database model training in SQL, the neural query planner, anomaly detection and forecasting, NL-to-SQL, quantum-inspired optimization, and SQL vector indexes with Product Quantization are planned, and so are a native MCP server, Code-Graph and Graph-RAG for Full. Nano and Lite run time-travel queries (AS OF) and Git-like branches; for HeliosDB Full both are planned (the current release accepts the syntax but does not read historical snapshots or create branches). HeliosDB Full authenticates PostgreSQL clients with SCRAM-SHA-256; SQL GRANT/REVOKE enforcement, row-level security, encryption at rest with key rotation and cloud KMS, ABAC, SAML/OIDC single sign-on and LDAP are planned. Writes are fsync-durable when acknowledged by default, and under a commercial licence the [Transaction Durability Commitment](https://www.heliosdb.com/transaction-durability.html) covers acknowledged commits against loss caused by HeliosProxy in the path or by the Generally Released Full version in use, including documented upgrades. SSPL-1.0, custom pricing on an annual contract through sales under the [Customer Terms](https://www.heliosdb.com/customer-terms.html); binaries, container images and Helm charts are provided during onboarding. A 116-skill agent bundle ships with the source and installs with its bundled install script.
- [HeliosProxy](https://www.heliosdb.com/heliosproxy.html): programmable PostgreSQL data-plane, a wire-protocol connection router that pools connections in session, transaction and statement modes (as PgBouncer does), splits reads from writes, and keeps client sessions alive through a primary failure with Transaction Replay. Transaction Replay ships in the default build (in-session recovery, the recovery journal and the operator-driven `POST /api/replay` and `POST /api/shadow` endpoints); `tr_enabled` is the runtime switch and the old `ha-tr` Cargo feature is a deprecated no-op ([Transaction Replay](https://www.heliosdb.com/docs/proxy/transaction-replay/)). `tr_mode` is none, session (default), select or transaction (opt-in). Replay verifies itself: each statement's response frames are digested as the client saw them and re-checked against the replacement backend, and divergence rolls the replay back and returns 40001. An interrupted read is re-executed only when every function it calls is provably side-effect-free (a PostgreSQL built-in on the proxy's allowlist, or a name listed in `tr_read_functions`); anything else returns 08007 and is never run twice. A COMMIT with unknown outcome is never retried, and SERIALIZABLE or REPEATABLE READ transactions are never replayed. The recovery journal records committed transactions in memory, or also in CRC-checked segment files when `[journal] dir` is set. Connection pooling is the only default Cargo feature and Transaction Replay needs none; the query cache (off by default), rate limiting, lag-aware routing, circuit breaking, WASM plugins, the GraphQL gateway, anomaly detection, edge mode and the PostgreSQL topology provider are opt-in Cargo features ([feature flags](https://www.heliosdb.com/docs/proxy/feature-flags/)). Also: Ed25519-signed WASM plugins distributed as OCI artefacts, a zero-downtime PostgreSQL major-version upgrade orchestrator with shadow execution, and a built-in admin web UI. Works in front of any PostgreSQL-wire backend, including HeliosDB, PostgreSQL 12+, Amazon RDS, Aurora, Cloud SQL and Azure Database. Apache 2.0. Install with `cargo install heliosdb-proxy` (Rust 1.86 or newer) or the container image `ghcr.io/heliosdatabase/hdb-heliosdb-proxy`; `heliosdb-proxy install skills` deploys 22 operator skills into `~/.claude/skills/` and `~/.codex/skills/`.
- [HeliosCore](https://www.heliosdb.com/helioscore.html): the HTAP storage engine with O_DIRECT I/O and hybrid columnar+row layout that powers HeliosDB Lite and Full.
- [Code-Graph & Graph-RAG](https://www.heliosdb.com/code-graph.html): in HeliosDB Nano builds with `--features code-graph`, query a codebase with SQL via the `hdb_code` extension and SQL-callable LSP functions; with `--features graph-rag`, build graph-augmented RAG pipelines with the `WITH CONTEXT` SQL clause.

## BaaS — HeliosDB Cloud and the Nano backend

- [HeliosDB Cloud BaaS](https://www.heliosdb.com/baas.html): managed database API. `POST https://cloud.heliosdb.com/api/v1/quick-start` creates an account, a database and a per-database API key (the request carries an email address, which may be a personal mailbox at the Instant level (the account is then API-only), and `"accept_terms": true`, which accepts the Cloud Terms, the Data Processing Agreement and the Transaction Durability Commitment; without it the answer is 422 `TERMS_NOT_ACCEPTED`); REST endpoints execute SQL, create, switch and merge branches, run time-travel queries (`query-at`) and create scoped API keys. For accounts created through the API the free tier grows in three levels (Instant: API signup; Verified: email verified; Feedback given); console accounts and every account at the Verified level or above are for a business email address, accounts created through the API may use any email address at the Instant level (API-only with a personal mailbox), and console accounts start at the Verified level. HeliosDB Cloud is in an Experimental Period: every plan is free for every customer, limits are indicative, and there is no trial of any kind; it gives no uptime commitment, and the [Transaction Durability Commitment](https://www.heliosdb.com/transaction-durability.html) applies instead. Plans and quotas are on [pricing.html](https://www.heliosdb.com/pricing.html); terms on [cloud-terms.html](https://www.heliosdb.com/cloud-terms.html).
- [Agent integration guide](https://www.heliosdb.com/baas.md): markdown-formatted, agent-friendly description of the HeliosDB Cloud API — drop into an LLM context window for end-to-end app generation.
- OpenAPI spec: https://cloud.heliosdb.com/api/v1/openapi.json
- MCP endpoint: https://cloud.heliosdb.com/api/v1/mcp — remote MCP server (Streamable HTTP; protocol versions 2026-07-28 and 2025-03-26/2025-06-18/2025-11-25; stateless, no sessions) that standard MCP clients connect to directly (Claude Code: `claude mcp add --transport http heliosdb https://cloud.heliosdb.com/api/v1/mcp --header "Authorization: Bearer $HDB_API_KEY"`). Requires authentication as a static header: a per-database API key as `Authorization: Bearer hdb_...` or `X-API-Key: hdb_...` (a session JWT also works); no OAuth; anonymous calls get 401. Get a key from `POST https://cloud.heliosdb.com/api/v1/quick-start` (returns `api_key`) or in the console (Databases → API Keys → Generate key). All 27 tools in `tools/list` work over MCP (including list/create/delete_database); each runs the REST endpoint's checks, a key works only for its own database, and a refused call is an `isError` result with the reason. Account creation is REST-only (`quick-start`), not an MCP tool.
- Self-hosted alternative: HeliosDB Nano serves its own PostgREST-style REST API, auth and realtime WebSocket from one binary ([REST API reference](https://www.heliosdb.com/docs/nano/api/rest-api/), [Nano vs Supabase](https://www.heliosdb.com/vs-supabase.html)); its /rest/v1 table routes and realtime event delivery are known issues in the current release. Row-level security does not filter those requests, and file storage and edge functions are on the HeliosDB roadmap.

## Documentation

- [Nano docs](https://www.heliosdb.com/docs/nano/): getting started, SQL reference, vector search, branching, embedded API
- [Lite docs](https://www.heliosdb.com/docs/lite/): production deployment, HeliosProxy + HeliosCore configuration, monitoring, backups
- [Full docs](https://www.heliosdb.com/docs/full/): protocol listeners, the HTTP gateway and deployment, plus the roadmap for clustering, replication and in-database ML
- [HeliosProxy docs](https://www.heliosdb.com/docs/proxy/): Transaction Replay in-session failover (https://www.heliosdb.com/docs/proxy/transaction-replay/): replay verified against a digest of the response frames the client already saw (divergence → 40001), an interrupted read re-executed only when every function it calls is provably side-effect-free (`tr_read_functions`), no replay of a SERIALIZABLE or REPEATABLE READ transaction, one `write_timeout_secs` deadline for the whole recovery, and the recovery journal with committed-history replay; plugin authoring (WASM), OCI distribution, version-upgrade orchestration
- [Cross-edition document storage comparison](https://www.heliosdb.com/cross-edition-docs.html)
- [REST API reference](https://www.heliosdb.com/docs/nano/api/rest-api/)

## Documentation for agents

Each docs site publishes an llms.txt index of its public pages and an llms-full.txt with the Markdown text of every page in that index.

- [HeliosDB Nano docs index](https://www.heliosdb.com/docs/nano/llms.txt): full text at https://www.heliosdb.com/docs/nano/llms-full.txt
- [HeliosDB Lite docs index](https://www.heliosdb.com/docs/lite/llms.txt): full text at https://www.heliosdb.com/docs/lite/llms-full.txt
- [HeliosDB Full docs index](https://www.heliosdb.com/docs/full/llms.txt): full text at https://www.heliosdb.com/docs/full/llms-full.txt
- [HeliosProxy docs index](https://www.heliosdb.com/docs/proxy/llms.txt): full text at https://www.heliosdb.com/docs/proxy/llms-full.txt

## AI services

- [AI-DBA Agent](https://www.heliosdb.com/ai-dba.html): autonomous database administration for HeliosDB Lite and Full (early access): monitoring, anomaly detection, preventive maintenance and performance tuning, in Monitor, Advise or Proactive mode; sold through sales
- [AI Support](https://www.heliosdb.com/ai-support.html): docs Q&A assistant over the HeliosDB documentation for Nano, Lite, Full and HeliosProxy, answering with cited docs URLs and ready-to-run SQL; free on HeliosDB Cloud, on-premises use with a licensed API key

## Comparisons

- [vs PostgreSQL](https://www.heliosdb.com/vs-postgresql.html)
- [vs MySQL](https://www.heliosdb.com/migrate-mysql.html)
- [vs SQLite](https://www.heliosdb.com/vs-sqlite.html)
- [vs MongoDB](https://www.heliosdb.com/vs-mongodb.html)
- [vs Supabase](https://www.heliosdb.com/vs-supabase.html)
- [vs Firebase](https://www.heliosdb.com/vs-firebase.html)
- [vs Pinecone / Weaviate / Qdrant / Milvus](https://www.heliosdb.com/vs-vector-dbs.html)
- [vs AIStor](https://www.heliosdb.com/vs-aistor.html)
- [vs MinIO](https://www.heliosdb.com/vs-minio.html)
- [Comparison hub](https://www.heliosdb.com/comparisons.html)

## Migrations

- [From MySQL](https://www.heliosdb.com/migrate-mysql.html)
- [From MongoDB](https://www.heliosdb.com/migrate-mongodb.html)
- [From DynamoDB](https://www.heliosdb.com/migrate-dynamodb.html)
- [Migration hub](https://www.heliosdb.com/migrations.html)

## Build with AI agents

- [Build with agents hub](https://www.heliosdb.com/build-with-agents.html): one-line install for Claude Code & Codex CLI skill bundles across all four products.
- HeliosProxy installs its agent skills with `heliosdb-proxy install skills`; the Nano, Lite and Full skill bundles ship with their sources (Lite also copies its bundle with `heliosdb-lite skills install <dir>`).
- [HeliosDB CodeKB MCP](https://www.heliosdb.com/codekb-mcp.html): single-binary MCP server that indexes a repo and exposes 11 LSP-shaped + Graph-RAG tools to Claude Code, Codex CLI, Cursor, Continue, and Aider over stdio or HTTP. Apache 2.0. Embeds HeliosDB-Nano. Repo: https://github.com/HeliosDatabase/HeliosDB-CodeKB-MCP

## Developer resources

- [Use cases](https://www.heliosdb.com/use-cases.html): catalog filterable by tier (Nano / Lite / Full / HeliosProxy) and area (AI/ML, enterprise, analytics, security, edge/IoT, etc.); entries copied from the Nano and HeliosProxy documentation link to their docs page
- [Code examples](https://www.heliosdb.com/examples.html): worked examples for a RAG backend, AI agent memory, multi-tenant SaaS, semantic product search, event analytics and feature flags
- [Integrations](https://www.heliosdb.com/integrations.html): client SDKs, AI frameworks (LangChain, LlamaIndex, AutoGen, CrewAI), wire protocols and platform integrations
- [SQL Playground](https://www.heliosdb.com/playground.html): a real embedded HeliosDB Nano instance per browser session, no signup or install, with vector search, branching and time-travel
- [HTAP guide](https://www.heliosdb.com/htap-guide.html)
- [Pagination at scale](https://www.heliosdb.com/pagination-performance.html)
- [Benchmarks](https://www.heliosdb.com/benchmarks.html)

## Illustrative scenarios

Scenarios built on the documented product surface, not customer accounts.

- [Illustrative scenario: fintech](https://www.heliosdb.com/case-study-fintech.html): payments compliance platform consolidated onto HeliosDB Lite
- [Illustrative scenario: IoT](https://www.heliosdb.com/case-study-iot.html): edge fleet on ARM gateways with offline-first sync
- [Illustrative scenario: AI legal](https://www.heliosdb.com/case-study-ai-legal.html): RAG-powered contract review with time-travel queries for audit
- [Illustrative scenario: SaaS multi-tenancy](https://www.heliosdb.com/case-study-saas.html): RLS + per-tenant branching + connection multiplexing
- [Illustrative scenarios hub](https://www.heliosdb.com/case-studies.html)

<!-- BEGIN: generated blog -->
## Blog

Every published post, newest first. Each post also has a markdown twin at the same URL with `.md` in place of `.html`.

- [Build SQL Visually: A Tour of the HeliosDB Cloud Query Builder](https://www.heliosdb.com/blog/cloud-query-builder-tour.html): From a new account to a filtered, sorted result in five minutes: pick a table, tick columns, add WHERE conditions, set ORDER BY and LIMIT, and the console writes the SQL for you. Then take the same query to REST with bound parameters, or to an AI agent over MCP.
- [HeliosProxy 2.0.0 - Committed History, Patroni Failover, and a Cache That Waits for COMMIT](https://www.heliosdb.com/blog/heliosproxy-2-0-0-committed-history-patroni-failover.html): Two releases since 1.7. 1.9 records what the backend actually committed and replays it in commit order, from disk. 2.0 drives query-cache invalidation from that record on both wire protocols and every tier, makes Patroni a write authority, and fails closed on conflicting primaries. With the cache on, write throughput is back at the cache-off level (+23–45% against 1.9).
- [HeliosProxy 1.7.0 — Transaction Replay That Checks Its Own Work](https://www.heliosdb.com/blog/heliosproxy-1-7-0-verified-transaction-replay.html): Replay now digests every response as the client saw it and re-verifies it on the replacement backend — divergence rolls back with 40001. Snapshot-pinned transactions are refused, read re-execution is a real side-effect policy, and one deadline bounds the whole recovery. In-session failover is core; no cargo feature required.
- [HeliosProxy 1.6.0 — Transaction Replay Drives In-Session Failover](https://www.heliosdb.com/blog/heliosproxy-1-6-0-in-session-failover.html): Sessions survive a primary failure: re-home and restore state, re-run interrupted reads, opt-in replay of uncommitted transactions. A COMMIT with unknown outcome is never retried.
- [HeliosDB Full v8.3.0 — Durable-by-Default WAL, Request-Path Hardening, and a Concurrency Jump](https://www.heliosdb.com/blog/full-8-3-0-durable-wal-concurrency.html): The WAL is now fsync-durable on acknowledgement, with real group commit so durable throughput scales with concurrency. Plus request-path panics eliminated, a startup-handshake timeout and connection cap, PRIMARY KEY uniqueness enforced across restarts, and ~17× read / ~36× write gains at eight clients on representative fixtures.
- [Vector Search Inside a Postgres-Wire-Compatible, Apache-2.0 Database — Why pgvector Users Are Switching](https://www.heliosdb.com/blog/vector-search-postgres-wire-apache-2-pgvector.html): Three structural ceilings of pgvector at corpus scale: RAM cost, deletes that rot recall, licensing if you want to host it. v3.32 closes all three — over the Postgres wire, under Apache 2.0, with existing clients and ORMs unchanged.
- [Benchmarking Nano v3.32: 16× Less Vector RAM, < 0.5% Recall Drop, Zero Regression on the Existing OLTP Path](https://www.heliosdb.com/blog/nano-3-32-vector-benchmark.html): Two clean signals. The new persistent PQ-HNSW path hits 16× compression at sub-0.5% recall drop and is restart-survivable; the default-features binary is byte-identical to v3.31.2 with measured cargo bench deltas (−13 % … +16 %) that scatter bidirectionally — the signature of machine noise, not code.
- [What Changed in HeliosDB Full Perf: Vectorized Comparisons, HNSW for Hybrid, RowDescription Cache](https://www.heliosdb.com/blog/full-perf-vectorized-comparisons-hnsw-hybrid.html): v7.7–v7.8 was a 12-release perf wave: vectorized Bool/String/Timestamp + mixed-numeric comparisons, HNSW for hybrid search (`/api/v1/vectors/stores/{store}/search/hybrid`), RowDescription cache, a faster parse-cache hash, compaction throttle. Numbers and the deltas.
- [Building a Persistent, Quantized, Deletable HNSW in Rust — No hnsw_rs, No faiss, No BSL](https://www.heliosdb.com/blog/persistent-quantized-deletable-hnsw-in-rust.html): Eight phases (P0–P7) of the in-house, from-scratch persistent PQ-HNSW: Malkov & Yashunin graph, RocksDB persistence under __vidx:, PQ + ADC + two-stage rerank, online deletes with neighbour repair, filtered KNN evaluated during traversal, F32/F16/I8 rerank dial. Apache 2.0 end-to-end, no third-party graph crate.
- [AI-Native at the Database Layer: Chat Proxy + Vector Store + Transcribe + SSE](https://www.heliosdb.com/blog/ai-native-database-layer-chat-vector-transcribe-sse.html): HeliosDB Full v7.5/v7.6 ships chat proxy, vector store, embeddings, transcribe, SSE, agents, documents, and per-tenant quotas — all behind one Postgres-wire instance. The AI surface that collapses Pinecone + Qdrant + an inference proxy into the engine.
- [Build a RAG App on Nano v3.32: From fastembed to a Persistent, Compressed Vector Index in 80 Lines of Rust](https://www.heliosdb.com/blog/rag-app-on-nano-3-32-in-80-lines-of-rust.html): End-to-end Rust walkthrough: fastembed BGE-Small embeddings, PersistentVectorIndex::create_with_pq, exact two-stage rerank, online deletes with remove() + compact(), restart proof, and the updated Nano vector skill for Claude Code and Codex.
- [HeliosProxy v0.4.2 install skills: 22 Operator Skills for Postgres Ops via Claude/Codex](https://www.heliosdb.com/blog/heliosproxy-0-4-2-operator-skills.html): `heliosdb-proxy install skills` unpacks an embedded skill bundle into `~/.claude/skills` and `~/.codex/skills` — 22 operator playbooks covering pool tuning, failover rehearsal, time-travel replay, plugin signing, and PG-12→17 upgrade orchestration.
- [HeliosDB Nano v3.32 — A Durable, 16×-Smaller Vector Index for RAG, With Recall-Safe Defaults](https://www.heliosdb.com/blog/nano-3-32-durable-smaller-vector-index.html): v3.32.0 lands the persistent PQ-HNSW vector index (opt-in via vector-persist); v3.32.1 fixes the PQ default that controls recall on every callsite, including SQL CREATE INDEX … USING hnsw. Crash-recoverable, ~16× less resident RAM at < 0.5% recall drop, online deletes with neighbour repair, filtered KNN. Apache 2.0, in-house, no third-party graph crate.
- [Per-Request Memory Hygiene: Why HeliosDB Lite Stays Steady Under Sustained Load](https://www.heliosdb.com/blog/lite-per-request-memory-hygiene.html): Design notes on scoping each statement's scratch memory to the request and watching per-query arena bytes. Editor's note: in Lite 3.6.0 and Full 8.3.0, per-request arenas and the helios_query_stats view are planned.
- [PREPARE COMPILED: Real Prepared Plans with an LRU Plan Cache](https://www.heliosdb.com/blog/prepare-compiled-lru-plan-cache.html): Design notes on PREPARE COMPILED, a fingerprint-keyed LRU plan cache shared across connections. Editor's note: PREPARE COMPILED is planned for HeliosDB Lite and Full; Lite 3.6.0 and Full 8.3.0 serve standard PostgreSQL wire clients.
- [HeliosDB Full v8.0: A Native LSM, RocksDB-Free](https://www.heliosdb.com/blog/full-8-0-native-lsm-14-protocols.html): v7.9.0→v8.0.0 retired RocksDB: native LSM, custom WAL, content-addressed CAS. Editor's note: the current Full listeners and their status are on full.html, and branching and time travel for Full are planned.
- [HeliosDB Nano v3.22 → v3.30 — Eight Releases of Postgres-Compat, Auth, and psql Tooling](https://www.heliosdb.com/blog/nano-3-22-to-3-30-postgres-compat-auth-psql.html): Eight Nano releases closed the gaps most evaluators hit in the first half-hour: SCRAM-SHA-256 with proper GS2 parsing, CREATE DATABASE + StartupMessage validation, information_schema completion, real \d / \l / \dt, keyset row-constructor pagination, and FK enforcement on the index path. The unglamorous wire-protocol work that turns 'interesting' into 'drop-in'.
- [RAG in SQL: Graph + BM25 + Hybrid + RRF/MMR, All in HeliosDB Lite v3.6](https://www.heliosdb.com/blog/rag-in-sql-graph-bm25-hybrid-lite-3-6.html): May 2026 design notes on graph traversal, BM25 ranking, hybrid search and RRF/MMR fusion as SQL functions for Lite. Editor's note: in Lite 3.6.0 these are planned, as are PREPARE COMPILED and an MCP server; Lite 3.6.0 serves PostgreSQL wire and REST.
- [HeliosDB Nano v3.22 — The Daily-RAG Correctness Wave](https://www.heliosdb.com/blog/nano-3-22-daily-rag-correctness.html): FK validator now reads-its-own-writes inside a transaction (PG-compatible). Cross-process `INSERT … ON CONFLICT` no longer duplicates. MCP `result_cache` stats published in `helios/info` and `GET /mcp/info`. Three fixes that close the daily-workflow silent-fail matrix.
- [Drop a Database into Claude Code in 30 Seconds: HeliosDB Skill Bundles](https://www.heliosdb.com/blog/claude-code-database-skill-bundles.html): Every HeliosDB product (Nano, Lite, Full, HeliosProxy) ships a `.claude/skills/` catalogue plus an `AGENTS.md` aggregate, installed with one command per product (a script for Nano and Full, `heliosdb-lite skills install` with a target directory, `heliosdb-proxy install skills`). Verb-by-verb tool catalogues for Claude Code, Codex CLI, and any MCP-aware agent.
- [HeliosDB Lite v3.6.3 — From v3.6.1, Through the Batched Executor Migration](https://www.heliosdb.com/blog/lite-3-6-3-batched-executor-migration.html): Internal tests of HeliosDB Lite builds labelled 3.6.1 to 3.6.3 (Lite has no release tags; its public version is 3.6.0): window functions −25%, INSERT throughput −17 to −31%, CTE −13%, time-travel index 39 000× over linear scan. Four silently-red regression tests now green.
- [Zero-Downtime PostgreSQL Major Upgrades, Orchestrated by HeliosProxy](https://www.heliosdb.com/blog/zero-downtime-postgresql-major-upgrades.html): Stage-driven upgrades with shadow-execution validation, chaos rehearsal, and time-travel replay rollback. The wedge into the PG-12-EOL migration market.
- [HeliosProxy v0.4 — From Connection Router to Programmable Data-Plane](https://www.heliosdb.com/blog/heliosproxy-0-4-programmable-data-plane.html): Real WASM plugin runtime, signed-plugin distribution via OCI, time-travel replay, zero-downtime PG major-version upgrade orchestrator, built-in admin console. Largest release in the project's history.
- [Native MCP Server: Why Your Database Should Be an MCP Tool Provider](https://www.heliosdb.com/blog/native-mcp-server-database-tool-provider.html): Three HTTP transports, 16 core auto-registered tools in builds with the mcp-endpoint feature, streaming progress events, single-shot discovery. Claude Code, Cursor, Continue, Codex, and Aider connect to HeliosDB Nano with no glue server.
- [Graph-RAG: The WITH CONTEXT Clause and Why RAG Belongs in SQL](https://www.heliosdb.com/blog/graph-rag-with-context-clause.html): WITH CONTEXT expands SQL seed rows over a graph in Nano. Editor's note: the ingest adapters are Rust API methods, and applying RERANK BY in SQL is planned.
- [HeliosDB Lite v3.6.1: April 2026 Monthly Performance Review](https://www.heliosdb.com/blog/lite-3-6-1-april-2026-performance-review.html): Internal tests of a HeliosDB Lite build labelled 3.6.1 vs 3.6.0 (Lite has no release tags; its public version is 3.6.0): 33 of 35 categories match or improve. DELETE point −36% (1000× on a 10K-row table), IN subquery −47%, DISTINCT −45%, JOINs ~−29%.
- [HeliosProxy v0.3.1: Faster Pool Checkout, Lock-Free Cache Reads](https://www.heliosdb.com/blog/heliosproxy-0-3-1-pool-checkout-cache-reads.html): A patch release focused on the request path: 28–37% faster pool checkout, 3.1× faster metrics reads, zero-allocation protocol parsing, and concurrent L1 cache hits.
- [Code-Graph: SQL-Callable LSP Inside Your Database](https://www.heliosdb.com/blog/code-graph-sql-callable-lsp.html): The AST is a table and LSP calls are SQL in Nano. Editor's note: the index is CREATE AST INDEX src_ast ON repo (content) USING tree_sitter(lang).
- [REST as the Universal Database Interface: One Engine, Three Protocols](https://www.heliosdb.com/blog/rest-universal-database-interface.html): Part 3 of the BaaS series. Three callers, three needs: long-lived servers want wire protocol, serverless workers want REST, agents want MCP. Why HeliosDB exposes all three on the same engine and how to pick per caller.
- [HeliosDB Nano v3.19 — From Embedded SQL to AI-Native Retrieval Engine](https://www.heliosdb.com/blog/nano-3-19-ai-native-retrieval-engine.html): Five weeks, eight minor versions, three new tracks. Code-Graph (CREATE EXTENSION hdb_code), Graph-RAG (WITH CONTEXT clause), native MCP server. In-process fastembed, Docling ingestion, Drizzle/Prisma/TypeORM compat.
- [Zero to a Vector-Search Notes App in One Shell Session](https://www.heliosdb.com/blog/vector-search-notes-app-in-one-shell-session.html): Part 2 of the BaaS series. A hands-on build: provision, schema with VECTOR(1536), seed with embeddings, per-user scoped keys, branch for a feature, time-travel to debug, self-host with the same SQL. No SDK, no backend.
- [Database-Agnostic by Default: Why a RESTful Backend Is the Portable Unit of Infrastructure](https://www.heliosdb.com/blog/database-agnostic-restful-backend.html): Part 1 of the BaaS series. SQL portability isn't database portability — drivers, pools, auth, and provisioning leak through. A RESTful BaaS moves the seam to the HTTP boundary. Bootstrap in one POST, deploy anywhere, scale stateless.
- [Introducing HeliosDB Full v7.1.0](https://www.heliosdb.com/blog/introducing-heliosdb-full-7-1-0.html): The January 2026 v7.1.0 announcement, with an editor's note: the current Full listeners and their status are on full.html, clustering is planned, and GPU acceleration (NVIDIA CUDA) is available in builds on request.
- [How We Built 384x Vector Compression](https://www.heliosdb.com/blog/how-we-built-384x-vector-compression.html): A deep dive into Product Quantization in HeliosDB Nano. Editor's note: the recall-safe default saves about 16x resident RAM (internal tests, Nano 3.32.1 vs an unquantized index, see benchmarks.html); 384x costs most of the recall.
- [Database Branching: Git for Your Data](https://www.heliosdb.com/blog/database-branching-git-for-your-data.html): How zero-cost copy-on-write branching at the storage layer changes the way teams develop, test, and deploy schema changes.
- [Why We Chose Rust for HeliosDB](https://www.heliosdb.com/blog/why-we-chose-rust-for-heliosdb.html): Memory safety without garbage collection, fearless concurrency, and zero-cost abstractions. Here's why Rust was the only choice.
- [Benchmarking HeliosDB vs PostgreSQL](https://www.heliosdb.com/blog/benchmarking-heliosdb-vs-postgresql.html): November 2025 internal tests comparing a HeliosDB build with PostgreSQL 16 across OLTP, OLAP, vector and multi-tenant workloads. Editor's note: not reproduced on current releases; current figures are on benchmarks.html.
- [Building RAG Applications with HeliosDB](https://www.heliosdb.com/blog/building-rag-applications-with-heliosdb.html): A step-by-step guide to building Retrieval-Augmented Generation pipelines using HeliosDB's native vector search and SQL.
- [HeliosProxy: Beyond PgBouncer](https://www.heliosdb.com/blog/heliosproxy-beyond-pgbouncer.html): HeliosProxy: the connection router whose flagship feature is Transaction Replay, plus multi-tier caching, a GraphQL gateway, WASM plugins and query routing.
- [Zero-Knowledge Encryption Explained](https://www.heliosdb.com/blog/zero-knowledge-encryption-explained.html): The 2025 design for client-side encryption. Editor's note: HeliosDB Nano ships zero-knowledge encryption; ZKE for Lite is planned.
- [Time-Travel Queries in Production](https://www.heliosdb.com/blog/time-travel-queries-in-production.html): Real-world patterns for using AS OF queries in production: audit compliance, debugging, data recovery, and A/B testing analysis.
- [Multi-Tenancy Best Practices](https://www.heliosdb.com/blog/multi-tenancy-best-practices.html): Row-Level Security, per-tenant resource quotas, and isolation patterns for building SaaS applications on HeliosDB.
- [Edge Computing with HeliosDB Nano](https://www.heliosdb.com/blog/edge-computing-with-heliosdb-nano.html): Deploy a full-featured database at the edge: IoT sensors, mobile apps, and offline-first architectures with one self-contained binary (about 36-44 MB in the 4.41.0 release, depending on platform).
- [The Future of AI-Native Databases](https://www.heliosdb.com/blog/future-of-ai-native-databases.html): Why vector search is just the beginning. Our vision for databases that understand, predict, and adapt to your application needs.
<!-- END: generated blog -->

## Company

- [About HeliosDB](https://www.heliosdb.com/about.html)
- [Pricing](https://www.heliosdb.com/pricing.html)
- [Blog](https://www.heliosdb.com/blog.html)
- [Careers](https://www.heliosdb.com/careers.html)
- [Contact](https://www.heliosdb.com/contact.html)
- [Legal documents](https://www.heliosdb.com/terms.html) — [Privacy Notice](https://www.heliosdb.com/privacy.html), [Cloud Terms](https://www.heliosdb.com/cloud-terms.html), [Customer Terms](https://www.heliosdb.com/customer-terms.html), [DPA](https://www.heliosdb.com/dpa.html), [Support Policy](https://www.heliosdb.com/support-policy.html), [Transaction Durability Commitment](https://www.heliosdb.com/transaction-durability.html) and partner terms; HeliosDB is a trading name of Dimen Sigon OÜ (Estonia)
- [Investors](https://www.heliosdb.com/investors.html) — 228 documented database innovations
- [Early Adopters program](https://www.heliosdb.com/early-adopters.html)
- [Partners program](https://www.heliosdb.com/partners.html) — hosting, technology and reseller partners; contact the partner team through the [contact form](https://www.heliosdb.com/contact.html?subject=partnership)
- [Ambassadors & Contributors](https://www.heliosdb.com/ambassadors.html) — community Champions program (recognition and enablement, not a referral scheme)

## Optional

- [HeliosDB Cloud](https://www.heliosdb.com/cloud.html): the managed service for HeliosDB Nano at https://cloud.heliosdb.com; the Free, Starter and Pro plans differ in indicative limits for databases, storage, connections, vectors, API requests and time-travel retention, and all are free of charge for every customer during the current Experimental Period (see [pricing](https://www.heliosdb.com/pricing.html)). There is no trial and no uptime SLA; the [Transaction Durability Commitment](https://www.heliosdb.com/transaction-durability.html) applies, and support is best effort under the [Support Policy](https://www.heliosdb.com/support-policy.html). Every plan has branching with as-of (timestamp or transaction), time-travel queries, managed backups, AES-256-GCM encryption at rest, zero-knowledge encryption, REST and MCP; Starter and Pro add a PostgreSQL connection string. Console accounts are for business email addresses; accounts that agents create through the BaaS API start at the Instant level with any email address. No Enterprise plan is offered during the Experimental Period. A managed service for Lite and Full is planned.
- GitHub repositories:
  - [HeliosDB Nano](https://github.com/HeliosDatabase/HeliosDB-Nano) — Apache 2.0
  - [HeliosProxy](https://github.com/HeliosDatabase/HeliosDB-Proxy) — Apache 2.0
  - [HeliosDB CodeKB MCP](https://github.com/HeliosDatabase/HeliosDB-CodeKB-MCP) — Apache 2.0
  - [HeliosDB SDKs](https://github.com/HeliosDatabase/HeliosDB-SDKs) — Apache 2.0 (Python, TypeScript and Go clients in preview, a Rust HTTP client planned; source only, not published to package registries)
- [AI Support docs Q&A API](https://www.heliosdb.com/ai-support.html) — `POST https://www.heliosdb.com/api/v1/support/ask` `{"question","tier"}`; answers cite docs URLs
- [Discord community](https://discord.gg/yTykuUrFXc)