Providers

Configure LLM providers for your agents — built-in services, custom endpoints, and runtime switching.

Overview

Traditional agent architectures couple prompt execution directly to specific API libraries, creating fragile systems and vendor lock-in. Generative Layers resolves this by decoupling the agent's symbolic reasoning from concrete neural endpoints. Through the bind() operator, agents establish a transactional governance boundary around any provider, allowing runtime model switching and out-of-band policy checks directly at the resource boundary.

Built-in Providers

Provider	Free Tier	Tested Models	Status
Groq	30 req/min, 14,400/day	`llama-3.3-70b-versatile`	✓ Tested
Groq — console.groq.com/keys Sign up (free) → Create API Key → Export as `GROQ_API_KEY` `export GROQ_API_KEY="gsk_...your-key-here"`
Cerebras	Free (no credit card)	`gpt-oss-120b`	✓ Tested
Cerebras — cloud.cerebras.ai Sign up (no credit card) → API Keys → Export as `CEREBRAS_API_KEY` `export CEREBRAS_API_KEY="csk-...your-key-here"`
Gemini	15 req/min, 1,500/day	`gemini-2.0-flash`, `gemini-2.5-flash`	✓ Tested
Gemini — aistudio.google.com/apikey Sign in with Google → Create API Key → Export as `GEMINI_API_KEY` `export GEMINI_API_KEY="AIzaSy...your-key-here"`
OpenAI	Paid only	`gpt-4o`, `gpt-4o-mini`	Compatible
OpenAI — platform.openai.com/api-keys Sign up → Create API Key → Export as `OPENAI_API_KEY` `export OPENAI_API_KEY="sk-proj-...your-key-here"`
DeepSeek	Small initial credit	`deepseek-chat`	Compatible
DeepSeek — platform.deepseek.com/api_keys Sign up → Create API Key → Export as `DEEPSEEK_API_KEY` `export DEEPSEEK_API_KEY="sk-...your-key-here"`

Usage Examples

Groq (Recommended Free Option)

Offers high-throughput hardware-accelerated inference with a generous daily rate limit, suitable for executing multi-phase validation loops.

gl.bind("agent1", "groq", "llama-3.3-70b-versatile", "");

gl.bind("agent1", "groq", "llama-3.3-70b-versatile", "", Bid);

bind("agent1", "groq", "llama-3.3-70b-versatile", "", Bid);

Cerebras

Provides low-latency wafer-scale inference. Accessible via a zero-cost API tier with no credit card registration, serving as an ideal environment for rapid prototyping and lifecycle testing.

gl.bind("agent1", "cerebras", "gpt-oss-120b", "");

gl.bind("agent1", "cerebras", "gpt-oss-120b", "", Bid);

bind("agent1", "cerebras", "gpt-oss-120b", "", Bid);

Gemini (Native Google API)

Integrates with Google's foundation models via native API drivers, bypassing generic wrapper translations to support advanced multi-modal and structured reasoning workloads directly.

gl.bind("agent1", "gemini", "gemini-2.5-flash", "");

gl.bind("agent1", "gemini", "gemini-2.5-flash", "", Bid);

bind("agent1", "gemini", "gemini-2.5-flash", "", Bid);

Using Any Custom Provider (Zero Code Changes)

Switching Providers at Runtime

Auto-Detection

Omitting the provider parameter during binding triggers a fallback resolution sequence that scans the environment variables. The framework traverses the environment in a fixed priority hierarchy (GROQ_API_KEY → CEREBRAS_API_KEY → GEMINI_API_KEY → OPENAI_API_KEY → DEEPSEEK_API_KEY) and automatically selects the first active key. If no external provider key is present, the framework falls back to a deterministic mock engine to support testing and local compilation without network dependencies.