@agentprodready/ai-provider-gemini

First-class Google Gemini API adapter for AgentProdReady (gemini-ai).

Not an OpenAI-compatible gateway — use openaiCompatible() for Chat Completions endpoints.

| | | |---|---| | Status | Published 1.0.x | | Install | with @agentprodready/agent-framework | | Credential | GEMINI_API_KEY only | | Node | >=22 <25 · ESM | | License | MIT |

When to use

Use this package when you need the native Gemini Generative Language API (chat, tools, streaming).

Prefer @agentprodready/agent-framework Simple API (gemini()) for most applications — it wires this adapter for you.

Do not use this package for embeddings (deferred in v1.6) or when a generic OpenAI-compatible endpoint is sufficient.

Install

npm install @agentprodready/agent-framework @agentprodready/ai-provider-gemini
export GEMINI_API_KEY="..."   # PowerShell: $env:GEMINI_API_KEY="..."

Optional: GEMINI_MODEL (default gemini-2.0-flash), GEMINI_BASE_URL.

Simple Agent API

import { createAgent, gemini } from "@agentprodready/agent-framework";

const agent = createAgent({
  model: gemini("gemini-2.0-flash"),
  instructions: "You are a helpful assistant.",
});

const result = await agent.invoke("Hello");
console.log(result.text);
await agent.close();

Supports chat, tools, and streaming via the Gemini API. Embeddings are not included in this release.

Ownership

| Owns | Does not own | |---|---| | Gemini request/response translation + normalized AI results | Capability selection; Security authorization; Runtime | | Capability id gemini-ai | Treating Gemini as openaiCompatible | | Env config (GEMINI_API_KEY, optional model/base URL) | Embeddings (deferred) |

Docs

• Gemini guide
• AI providers
• Monorepo README

License

MIT © 2026 ameenmari