NexusAI-Client

Configuration

Configure API keys, environment variables, and fallback parameters.

Configuration ⚙️

NexusAI-Client automatically loads credentials from your system environment or a local .env file at the root of your project.


🔑 Environment Variables (.env)

Create a .env file in the working directory where your Python application runs:

# ==============================================================================
# Priority 1: 100% FREE ZERO-COST TIERS (Recommended for zero-cost operation)
# ==============================================================================
GEMINI_FREE_API_KEY=AIzaSy...              # Google AI Studio Free Tier (1M Context)
GROQ_API_KEY=gsk_...                       # Groq Cloud LPU (Ultra-fast gpt-oss-120b & Llama 3.3)
CEREBRAS_API_KEY=csk-...                   # Cerebras CS-3 Wafer-Scale Engine
COHERE_API_KEY=...                         # Cohere Free Trial (Command R+)
NVIDIA_API_KEY=nvapi-...                   # Nvidia NIM (1,000 Free Credits)
OPENROUTER_API_KEY=sk-or-v1-...            # OpenRouter (19 Free Models Live)
ORCAROUTER_API_KEY=sk-orca-...            # OrcaRouter (Free Qwen & DeepSeek models)
MISTRAL_API_KEY=...                        # Mistral AI (Codestral & Small)

# ==============================================================================
# Priority 2: ULTRA-LOW-COST PAID BACKUP TIERS (Safety nets)
# ==============================================================================
DEEPSEEK_API_KEY=sk-...                    # DeepSeek V3 / R1 ($0.27 / 1M tokens)
GEMINI_PRO_API_KEY=AIzaSy...               # Google Cloud Gemini 3.1 Pro (Pay-as-you-go)

[!TIP] You only need to configure the providers you intend to use. When using AIGateway.auto_fallback(), NexusAI-Client automatically detects all present keys and skips unconfigured providers.


⚡ Direct Initialization (Without .env)

You can also pass API keys, custom base URLs, or custom timeouts directly when instantiating AIGateway:

from nexusai_client import AIGateway

# Pass explicit credentials and custom parameters
async with AIGateway(
    "groq",
    api_key="gsk_custom_key_here",
    model="openai/gpt-oss-120b",
    timeout=30.0,
) as client:
    res = await client.generate_text("Hello Groq!")
    print(res.text)

🌐 Custom Base URLs (Proxies & Gateways)

If you use an internal proxy, caching gateway, or custom endpoint:

async with AIGateway(
    "deepseek",
    api_key="sk-...",
    base_url="https://my-internal-proxy.company.internal/v1",
) as client:
    res = await client.generate_text("Hello from behind proxy!")

⏱️ Timeout Management

NexusAI-Client defaults to a generous timeout suitable for heavy reasoning models. You can override it globally or per-request:

# Set a 15-second network timeout for this gateway instance
async with AIGateway("gemini_free", timeout=15.0) as client:
    res = await client.generate_text("Quick summarize...")

On this page