g1-api ⚡

Unified API gateway — REST endpoints, web search, and GPU-powered speech synthesis.

---

✨ Overview

g1-api is the public API surface of Generate One. It bundles three compose stacks: Fusio as the API gateway for building and managing OAuth2-secured REST endpoints, a search stack with SearXNG + Crawl4AI for web intelligence APIs, and Chatterbox TTS for GPU-accelerated text-to-speech synthesis on an RTX 5090. All routes are documented via OpenAPI and accessible through a developer portal.

🏗️ Architecture

graph TD
    subgraph g1-api ["g1-api · api.generate.one"]
        subgraph gateway ["🔐 API Gateway"]
            F[Fusio 5.2<br/>OAuth2 + Scopes<br/>:80]
            MDB[MariaDB 10.11<br/>API Metadata]
        end
        subgraph search ["🔍 Search APIs"]
            SX[SearXNG<br/>Meta Search<br/>:8080]
            C4[Crawl4AI 0.8.0<br/>Web Scraping<br/>:11235]
        end
        subgraph tts ["🗣️ Speech Synthesis"]
            CB[Chatterbox TTS<br/>GPU Accelerated<br/>:8004]
        end
    end

    CLIENT[API Consumers] -->|Bearer Token| T[Traefik]
    T -->|api.generate.one| F
    T -->|search.generate.one| SX
    F --> MDB
    F -->|/v1/chat/completions| LLM[LiteLLM<br/>g1-llm]
    F -->|/v1/search| SX
    F -->|/v1/memory/search| MEM[Graphiti<br/>g1-brain]
    F -->|/v1/knowledge/search| KM[knowledge-mcp<br/>g1-brain]
    CB -.->|🎮 NVIDIA RTX 5090| GPU[GPU]

    style g1-api fill:#1a1a2e,stroke:#e94560,stroke-width:2px,color:#fff
    style F fill:#1976d2,color:#fff
    style CB fill:#76b900,color:#fff
    style SX fill:#3daee9,color:#fff
    style C4 fill:#ff6f00,color:#fff

🚀 Quick Start

# Clone the repository
git clone git@git.generate.one:generate-one/g1-api.git
cd g1-api

# Deploy API gateway
docker compose -f compose/api-gateway.yml up -d

# Deploy search APIs
docker compose -f compose/api-search.yml up -d

# Deploy TTS (requires NVIDIA GPU)
docker compose -f compose/api-tts.yml up -d

🔗 Key Endpoints

Endpoint	Purpose
https://api.generate.one	API gateway root
https://api.generate.one/system/generator/spec-openapi	OpenAPI specification
https://api.generate.one/apps/redoc	Interactive API documentation

🛣️ API Routes

Route	Method	Scope	Backend
/v1/chat/completions	POST	api.chat	LiteLLM
/v1/search	POST	api.search	SearXNG
/v1/memory/search	POST	api.memory	Graphiti API
/v1/knowledge/search	POST	api.knowledge	knowledge-mcp
/v1/health	GET	(public)	Composite health check

🔧 Configuration

API Gateway

Variable	Description	Default
FUSIO_PROJECT_KEY	Fusio project encryption key	—
FUSIO_URL	Public API URL	https://api.generate.one
FUSIO_DB_PASSWORD	MariaDB password	—
FUSIO_CONNECTION	Database connection string	pdo-mysql://fusio:…@mariadb/fusio

Search Stack

Variable	Description	Default
SEARXNG_SECRET	SearXNG instance secret	—
SEARXNG_BASE_URL	Public search URL	https://search.generate.one
CRAWL4AI_API_TOKEN	Crawl4AI auth token	—
OPENAI_BASE_URL	LLM proxy for AI extraction	LiteLLM internal

Text-to-Speech

Variable	Description	Default
HF_TOKEN	HuggingFace access token	—
NVIDIA_VISIBLE_DEVICES	GPU device visibility	all

📦 Services

Service	Image	Port	Domain	Purpose
Fusio	fusio/fusio:5.2	80	api.generate.one	API gateway with OAuth2, scopes, rate limiting
MariaDB	mariadb:10.11	3306	Internal	Fusio metadata database
SearXNG	searxng/searxng:latest	8080	search.generate.one	Privacy-first meta search
Crawl4AI	unclecode/crawl4ai:0.8.0	11235	Internal	AI-powered web scraping
Chatterbox	chatterbox-tts:cu128 (local)	8004	Internal	GPU text-to-speech (RTX 5090)

📂 Compose Files

File	Stack	Description
compose/api-gateway.yml	🔐 Gateway	Fusio + MariaDB
compose/api-search.yml	🔍 Search	SearXNG + Crawl4AI
compose/api-tts.yml	🗣️ TTS	Chatterbox (GPU, locally built)

🔐 Authentication

Fusio uses OAuth2 client credentials flow. Token scopes control access:

┌─────────────────┐     ┌──────────────┐     ┌─────────────┐
│  Client App     │────▶│  Fusio       │────▶│  Backend    │
│  Bearer Token   │     │  Scope Check │     │  Service    │
└─────────────────┘     └──────────────┘     └─────────────┘

Scope	Access
api.chat	Chat completions (LiteLLM)
api.search	Web search (SearXNG)
api.memory	Memory/graph search (Graphiti)
api.knowledge	RAG knowledge search

🔗 Dependencies

Dependency	Project	Purpose
LiteLLM	g1-llm	Chat completion backend
SearXNG	g1-web	Web search backend
Graphiti	g1-brain	Memory/knowledge backends
PostgreSQL	svc-infra	Shared database infrastructure
Valkey	svc-infra	Crawl4AI task queue
NVIDIA GPU	Host	Chatterbox TTS inference

🔗 Related Repos

Repo	Relationship
g1-llm	LiteLLM model routing
g1-web	SearXNG + Crawl4AI backends
g1-brain	Graphiti API + knowledge-mcp
g1-auth	Authentik OIDC provider

---

🛡️ Part of Generate One

Generate One — AI infrastructure that answers to you.

Self-hosted, sovereign AI platform. generate.one