Self-hosted AI

By default, AI features in Forest are processed by Forest servers. To keep your data private, you can route AI requests through your own back-end using the addAi method, your data never leaves your infrastructure.

Self-hosted AI is available for the @forestadmin/agent Node.js agent.

Installation

npm install @forestadmin/ai-proxy

Configuration

Pass a provider built with createAiProvider to addAi:

import { createAiProvider } from '@forestadmin/ai-proxy';

const agent = createAgent(options)
  .addDataSource(/* ... */)
  .addAi(
    createAiProvider({
      name: 'my-assistant',
      provider: 'openai',
      apiKey: process.env.OPENAI_API_KEY,
      model: 'gpt-4o',
    }),
  );

import { createAiProvider } from '@forestadmin/ai-proxy';

const agent = createAgent(options)
  .addDataSource(/* ... */)
  .addAi(
    createAiProvider({
      name: 'my-assistant',
      provider: 'anthropic',
      apiKey: process.env.ANTHROPIC_API_KEY,
      model: 'claude-sonnet-4-5',
    }),
  );

Options

Option	Type	Required	Description
`name`	`string`	Yes	Unique identifier for this AI configuration
`provider`	`'openai'` \| `'anthropic'`	Yes	AI provider
`model`	`string`	Yes	Model to use (see supported models below)
`apiKey`	`string`	No	API key (defaults to `OPENAI_API_KEY` or `ANTHROPIC_API_KEY` env var)

All provider-specific options are also supported: ChatOpenAI options (configuration, temperature, maxTokens, etc.) and ChatAnthropic options (anthropicApiUrl, temperature, maxTokens, topK, etc.).

Supported models

Any model with function/tool calling support works. See the list of unsupported models for details.

Zendesk plugins Testing your back-end

⌘I

Overview

Connect

Process

Build

Execute

Collaborate

Manage

Embed

Control

Project Settings

Installation

Configuration

Options

Supported models

​Installation

​Configuration

​Options

​Supported models

Installation

Configuration

Options

Supported models