Fluxbase ⚡

The Serverless SQL Platform Built for Speed, Scale, and Developers.

The Serverless SQL Platform Built for Speed, Scale, and Developers.

---

Table of Contents

• What is Fluxbase?
• Key Features
• API Reference
• Ingestion Worker
• Real-Time Subscriptions
• Performance
• Client Integration Examples
• Monitoring & Observability
• Project Structure
• Contributing
• License

---

What is Fluxbase?

Fluxbase is a serverless SQL platform that wraps your relational databases in a developer-friendly REST API — complete with a high-performance async ingestion pipeline and real-time subscription capabilities.

It enables multi-dialect database execution (PostgreSQL and MySQL) with strict tenant isolation and a zero-trust security model, so you can focus on building features instead of managing infrastructure.

---

Key Features

Feature	Description
🗄️ Native SQL Execution	Execute raw SQL on bare-metal PostgreSQL or MySQL drivers — no heavy ORM abstractions.
🚀 High-Throughput Ingestion	Dedicated ingestion pipeline capable of writing 80,000+ rows/second asynchronously.
📡 Real-Time Data Streaming	Stream row-level events (INSERT, UPDATE, DELETE) to clients over resilient Server-Sent Events (SSE).
🔒 Security-First Architecture	Built-in AST-based SQL validation, JWT-claim RLS mapping, and scoped API key authorization.
🌐 Multi-Dialect Support	First-class support for both PostgreSQL and MySQL with dialect-aware query generation.
🤖 Flux AI Multimodal Gateway	OpenAI-compatible drop-in gateway for Text Reasoning, Image Generation, Voice (STT & TTS), Video, and 768-dim Embeddings with unlimited quotas for Enterprise, Employee, and PAYG accounts.
📊 Observability	Prometheus metrics endpoint + preconfigured Grafana dashboard out of the box.

---

API Reference

All requests require a project-scoped API key passed via the Authorization header:

Authorization: Bearer <your-api-key>

Base URL: https://fluxbasedb.me

---

POST /api/execute-sql

Executes an arbitrary SQL query under the project's namespace.

Request Body:

{
  "query": "SELECT * FROM users WHERE active = true LIMIT 5"
}

Response:

{
  "success": true,
  "result": {
    "rows": [
      { "id": "018f4a2b-...", "name": "Alice" }
    ],
    "columns": ["id", "name"]
  }
}

Error Response:

{
  "success": false,
  "error": {
    "code": "QUERY_FORBIDDEN",
    "message": "DROP statements are not permitted."
  }
}

---

POST /api/ingest

Queues one or more rows for high-speed asynchronous ingestion. The table is created automatically if it does not exist.

Request Body:

{
  "table": "events",
  "rows": [
    { "event_name": "page_view", "path": "/home" },
    { "event_name": "click", "path": "/pricing" }
  ]
}

Response:

{
  "success": true,
  "queued": 2,
  "batchId": "batch_12345"
}

---

GET /api/realtime

Establishes a Server-Sent Events (SSE) connection to subscribe to live database events for a project.

GET /api/realtime?projectId=<project-id>
Accept: text/event-stream
Authorization: Bearer <your-api-key>

Event Payload Example:

{
  "event": "INSERT",
  "table": "orders",
  "row": { "id": "abc123", "status": "pending" }
}

Note: Clients should implement exponential backoff reconnection logic. The Fluxbase JS client SDK handles this automatically.

---

Flux (OpenAI-Compatible AI Gateway)

Fluxbase provides a built-in, drop-in replacement for OpenAI endpoints named Flux with dynamic model whitelabeling, S3 media persistence, and multi-provider failover.

Base URL: https://fluxbasedb.me/api/v1

Supported Modalities & Models

Modality	Canonical Model	OpenAI Alias	Upstream / Capability
Text Reasoning	flux (Default)	gpt-3.5-turbo, flux-fast	Ultra-fast reasoning & SQL synthesis (128k context)
Deep Reasoning	flux-ultra, flux-pro, flux-5.2	gpt-4o, claude-3-5-sonnet	Complex multi-table reasoning & migration logic
Image Synthesis	flux-image, flux-image-fast	dall-e-3, dall-e-2	Photorealistic image generation with durable S3 URLs
Speech-to-Text	flux-listen, flux-listen-pro	whisper-1, whisper	Multilingual transcription with timestamps
Text-to-Speech	flux-speak, flux-speak-hd	tts-1, tts-1-hd	Natural voices: alloy, echo, fable, onyx, nova, shimmer
Video Generation	flux-video, flux-video-pro	cogvideox	Text-to-video with async task polling (HTTP 202)
Embeddings	flux-embed	text-embedding-3-small	768-dimensional vectors for semantic search & RAG

Unlimited Quota Policy

Accounts on employee, org_owner, and pay_as_you_go (PAYG) receive unlimited quotas across all modalities (no RPM/TPM restrictions, no daily limits, and access to all models).

Python OpenAI SDK Example

from openai import OpenAI

client = OpenAI(
    base_url="https://fluxbasedb.me/api/v1",
    api_key="flx_live_your_fluxbase_key"
)

# Text Reasoning
chat = client.chat.completions.create(
    model="flux-fast",
    messages=[{"role": "user", "content": "Write an index optimization plan."}]
)

# Image Generation
img = client.images.generate(
    model="flux-image",
    prompt="A futuristic neon server farm, 8k octane render",
    n=1
)
print(f"Generated Image: {img.data[0].url}")

---

Ingestion Worker

The ingestion worker is a standalone Python service (ingestion-worker/) that dequeues rows from a Redis buffer and streams them into the database.

• Fast COPY Protocol — Batches are streamed using asyncpg's COPY protocol instead of parameterized INSERT statements, maximizing throughput and eliminating per-row overhead.
• Dynamic Union Schema Merging — Before importing a batch, the worker calculates the union of all keys across rows to automatically add missing columns, keeping schemas flexible.
• Strict Identifier Sanitization — Any table or column name not matching ^[a-zA-Z_][a-zA-Z0-9_]*$ is rejected and quarantined to a Dead Letter Queue (DLQ), blocking SQL injection at the ingestion boundary.
• Auto-Scaler — The scaler.py module monitors queue depth and adjusts worker concurrency automatically.

Running the worker locally

cd ingestion-worker
pip install -r requirements.txt
python main.py

---

Real-Time Subscriptions

Fluxbase implements low-latency SSE subscriptions with built-in connection resilience:

• Throttled Cache Invalidation — Invalidation events are throttled to prevent UI blocking under high-frequency database writes.
• Connection Resilience — Exponential backoff with jitter and automatic heartbeats ensure clients reconnect gracefully after network drops.
• Shared Event Source — A single SSE connection is reused per project across all UI components to prevent connection exhaustion on the server.

---

Performance

Technique	Details
COPY vs INSERT	Bulk inserts are translated into PostgreSQL binary stream copies, eliminating SQL parsing and planning overhead.
Monotonic UUID v7	Recommended for primary keys to prevent B-tree page fragmentation and index splits under heavy insert loads.
Async WAL Writing	Transaction-local SET synchronous_commit = off allows fast ingestion replies without waiting for WAL disk flushes.
Redis Queue Buffer	Incoming rows are buffered in Redis, decoupling the API from the database and absorbing traffic spikes.

---

Client Integration Examples

JavaScript / TypeScript

async function executeQuery<T>(sql: string): Promise<T[]> {
  const response = await fetch("https://fluxbasedb.me/api/v1/sql", {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "Authorization": "Bearer <API_KEY>"
    },
    body: JSON.stringify({ projectId: "<PROJECT_ID>", query: sql })
  });
  const data = await response.json();
  if (!data.success) throw new Error(data.error.message);
  return data.rows;
}

Python

import requests

BASE_URL = "https://fluxbasedb.me"
HEADERS  = {"Authorization": "Bearer <API_KEY>", "Content-Type": "application/json"}

def execute_query(project_id: str, sql: str) -> list[dict]:
    response = requests.post(
        f"{BASE_URL}/api/v1/sql",
        headers=HEADERS,
        json={"projectId": project_id, "query": sql}
    )
    data = response.json()
    if not data["success"]:
        raise RuntimeError(data["error"]["message"])
    return data["rows"]

def ingest_rows(table: str, rows: list[dict]) -> dict:
    response = requests.post(
        f"{BASE_URL}/api/ingest",
        headers=HEADERS,
        json={"table": table, "rows": rows}
    )
    return response.json()

Real-Time (JavaScript EventSource)

const source = new EventSource(
  `https://fluxbasedb.me/api/realtime/subscribe?projectId=${PROJECT_ID}&table=orders`
);

source.onmessage = (event) => {
  const { table, eventType, record } = JSON.parse(event.data);
  console.log(`[${eventType}] on ${table}:`, record);
};

source.onerror = () => {
  // The Fluxbase SDK handles reconnection automatically.
};

---

Monitoring & Observability

The ingestion worker exposes a /metrics Prometheus endpoint tracking:

Metric	Type	Description
rows_ingested_total	Counter	Total rows successfully written to the database
rows_failed_total	Counter	Total rows that failed processing
rows_dlq_total	Counter	Total rows quarantined to the Dead Letter Queue
insert_latency_ms	Histogram	End-to-end latency from queue dequeue to DB write

A preconfigured Grafana dashboard is available at ingestion-worker/grafana_dashboard.json for monitoring ingest throughput, error rates, and queue latency in real time.

Alert rules are defined in ingestion-worker/alert_rules.yml.

---

Project Structure

Fluxbase/
├── src/
│   ├── app/                        # Next.js App Router
│   │   ├── (app)/                  # Authenticated dashboard views
│   │   ├── api/                    # API routes (execute-sql, ingest, realtime)
│   │   ├── pricing/                # Public pricing page
│   │   ├── docs/                   # Public documentation page
│   │   ├── layout.tsx              # Root layout
│   │   └── manifest.ts             # PWA manifest
│   ├── components/                 # Reusable UI components
│   ├── lib/                        # DB pools, auth helpers, utilities
│   ├── hooks/                      # Custom React hooks
│   ├── contexts/                   # React context providers
│   ├── actions/                    # Next.js server actions
│   └── server/                     # WebSocket & server-side modules
├── ingestion-worker/               # Async Python ingestion service
│   ├── main.py                     # Worker entrypoint
│   ├── worker.py                   # Schema merger & COPY implementation
│   ├── scaler.py                   # Auto-scaling logic
│   ├── metrics.py                  # Prometheus metrics definitions
│   ├── health.py                   # Health check endpoint
│   ├── throttle.py                 # Rate-limiting / throttle logic
│   ├── grafana_dashboard.json      # Preconfigured Grafana dashboard
│   ├── alert_rules.yml             # Prometheus alert rules
│   ├── Dockerfile                  # Container image for the worker
│   └── requirements.txt
├── fluxbase-client/                # Official JavaScript/TypeScript SDK
│   └── src/
├── src-tauri/                      # Tauri desktop app wrapper
│   └── tauri.conf.json
├── public/                         # Static assets & PWA icons
├── next.config.ts
├── tailwind.config.ts
├── package.json
└── README.md

---

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository.
2. Create a feature branch:

   git checkout -b feat/my-feature

3. Commit your changes using Conventional Commits:

   git commit -m "feat: add my feature"

4. Push to your fork:

   git push origin feat/my-feature

5. Open a Pull Request against the main branch and describe your changes.

Please ensure your code passes linting (npm run lint) and type-checking (npm run typecheck) before submitting.

---

License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Made with ❤️ by the Fluxbase Team