Browser & Mobile Runtimes

Use the PraisonAI TypeScript SDK inside Tauri, React Native, browser webviews, Deno, Bun, and edge runtimes

The PraisonAI TypeScript SDK runs in any JavaScript runtime — not just Node. Pass the API key through the Agent config, and the same code that runs on your server runs inside a Tauri app, a React Native mobile app, a browser webview, Deno, Bun, or a Cloudflare Worker.

Quick Start

1. Create an agent in a browser bundle

   Pass the API key explicitly — a browser bundle has no process.env.

   import { Agent } from 'praisonai';
   
   const agent = new Agent({
     instructions: 'You are a helpful assistant',
     llm: 'gpt-4o-mini',
   });
   
   const answer = await agent.start('Say hello');
   console.log(answer);

Choosing Where to Run

Pick the runtime that matches how your agent ships.

What Changed

The Agent constructor now works on runtimes without a process global.

Previously the constructor read process.env.OPENAI_MODEL_NAME, PRAISONAI_MODEL, PRAISON_VERBOSE, and PRAISON_PRETTY directly. On any runtime without a process global (Tauri webview, React Native, plain browser), the very first new Agent(...) threw ReferenceError: process is not defined — before your code ever ran.

Now those reads go through an internal getEnv() helper that returns undefined when there is no process. Constructors succeed, defaults apply (gpt-4o-mini, verbose on, pretty off), and any explicit config value still wins.

Streaming also survives: the default token sink (writeTokenToStdout) no-ops when process.stdout is unavailable, so agent.start() and agent.stream() no longer die on the first token in a webview.

Runtime Interaction Flow

A single agent call streams tokens straight into your UI.

Passing API Keys Per Runtime

Read the key where each runtime keeps it, then pass it to the Agent.

Minimal Tauri Example

Read the key in Rust, hand it to the renderer, and construct the agent.

import { invoke } from '@tauri-apps/api/core';
import { Agent } from 'praisonai';

const apiKey = await invoke<string>('get_openai_key');

const agent = new Agent({
  instructions: 'You are a coding assistant',
  llm: 'openai/gpt-4o-mini',
  apiKey,
});

await agent.start('Explain closures in one paragraph');

Minimal React Native Example

Fetch an ephemeral token from your backend, then run the agent on-device.

import { Agent } from 'praisonai';

async function ask(prompt: string) {
  const { token } = await fetch('https://your-backend.example/agent-token')
    .then(r => r.json());

  const agent = new Agent({
    instructions: 'You are a mobile helper',
    llm: 'openai/gpt-4o-mini',
    apiKey: token,
  });

  return await agent.start(prompt);
}

Streaming Without stdout

Use stream() and render each chunk into the UI instead of the terminal.

const agent = new Agent({
  instructions: 'You are helpful',
  llm: 'openai/gpt-4o-mini',
  apiKey,
});

const tokens: string[] = [];
for await (const chunk of agent.stream('Write a haiku')) {
  tokens.push(chunk);
  updateUI(tokens.join(''));  // your renderer
}

Best Practices

Troubleshooting

Related