WAL SQLiteEdge SQLite Archive
BP-06 // OPEN-SOURCE LIBSQL

Self-Hosted LibSQL (sqld): Turso HTTP/Hrana & Embedded Replicas Without Row Fees

SQLiteEdge LibSQL blueprint: running open-source Rust sqld behind serverless edge functions with embedded read replicas and bottomless S3 backups.

MODELED DESIGN TARGET
19,200 TPS
READ PATH / RPO TARGET
0.15 ms (Embedded)
REFERENCE NVME PROFILE
$24–$48 HF Primary

LibSQL (`sqld`) extends SQLite with an HTTP/WebSocket Hrana wire protocol, native `F32_BLOB` DiskANN vector indexes, virtual WAL hooks, and local embedded replicas—powering serverless functions that cannot mount a persistent disk. Self-hosting `sqld` on a Vultr High-Frequency NVMe primary gives you the exact `@libsql/client` SDK ergonomics with zero per-billion row-read or per-million row-write overage fees.

In our SQLiteEdge BP-06 reference guide, every configuration rule below follows official SQLite 3.46 documentation and public vendor specifications retrieved 2026-11-12.

SQLiteEdge Architecture Rules for Self-Hosted LibSQL sqld & Embedded Replicas

  1. Deploy sqld primary on a 3.5GHz+ Vultr High Frequency NVMe node with JWT Ed25519 token auth (`--auth-jwt-key-file`) exposed over TLS/HTTP2.
  2. Use @libsql/client Embedded Replicas inside regional app servers (`syncUrl` + local file path) for 0.1ms local NVMe reads while routing writes automatically to the primary.
  3. Store native AI embeddings right inside SQLite using `F32_BLOB(1536)` + `libsql_vector_idx`—eliminating separate $70/mo Pinecone or Qdrant clusters.
  4. Stream bottomless WAL backups from sqld directly to your $6/mo InterServer S3/MinIO bucket via built-in `--enable-bottomless-replication`.

SQLiteEdge Reference Snippet: Self-Hosted sqld Primary + Bottomless S3 Compose Stack

Self-Hosted sqld Primary + Bottomless S3 Compose Stack 
services:
  sqld-primary:
    image: ghcr.io/tursodatabase/libsql-server:latest
    restart: always
    ports:
      - "127.0.0.1:8080:8080"
      - "127.0.0.1:5001:5001"
    environment:
      SQLD_NODE: primary
      SQLD_HTTP_LISTEN_ADDR: 0.0.0.0:8080
      SQLD_GRPC_LISTEN_ADDR: 0.0.0.0:5001
      SQLD_AUTH_JWT_KEY_FILE: /etc/sqld/jwt_public.pem
      LIBSQL_BOTTOMLESS_ENDPOINT: https://minio-vault.yourdomain.com
      LIBSQL_BOTTOMLESS_BUCKET: libsql-bottomless-prod
      AWS_ACCESS_KEY_ID: ${S3_ACCESS_KEY}
      AWS_SECRET_ACCESS_KEY: ${S3_SECRET_KEY}
    volumes:
      - /var/lib/sqld-nvme:/var/lib/sqld

SQLiteEdge Modeled Profile & Tariff Table (BP-06)

Public vendor prices retrieved 2026-11-12 (promotional and renewal rates may change at checkout); throughput figures reflect theoretical single-queue sizing models rather than unarchived lab claims.

Workload ScaleTurso Cloud ManagedSelf-Hosted sqld on Vultr HF + S3
5B Row Reads + 15M Writes / 15 GB$29/mo (Scaler)$24/mo HF NVMe + $6/mo S3 Vault
40B Row Reads + 80M Writes / 60 GB$156/mo$48/mo HF NVMe + $6/mo S3 Vault
200B Row Reads + 350M Writes / 180 GB$640+/mo$96/mo HF NVMe + $6/mo S3 Vault

SQLiteEdge Spec-Matched NVMe & S3 Targets for BP-06

Open this BP-06 workload preset inside the SQLiteEdge PRAGMA Studio →

SQLiteEdge Engineering FAQ for BP-06

  • Is my existing @libsql/client and Drizzle/Prisma code 100% compatible with self-hosted sqld?

    Yes—you literally swap url: "libsql://your-db.turso.io" to "https://db.yourdomain.com" and supply your self-signed Ed25519 JWT authToken. Zero application query changes required.

  • How do sqld read replicas work across Singapore, Frankfurt, and New York?

    Secondary sqld nodes connect to the primary over gRPC port 5001 (`SQLD_NODE=replica`), serving local regional reads in <1ms and transparently forwarding write statements back to the primary.