vitrus
Version:
TypeScript client for interfacing with the Vitrus SDK
155 lines (112 loc) ā¢ 3.87 kB
Markdown
[](https://www.npmjs.com/package/vitrus)
[](https://packagephobia.now.sh/result?p=vitrus)
[](https://www.npmjs.com/package/vitrus)
A TypeScript client for multi Agent-World-Actor orchestration, with easy-to-use spatial perception _Workflows_.
For detailed documentation and more examples access [Vitrus Docs](https://vitrus.gitbook.io/docs/concepts).
š” Tip: If anything takes more than 2 minutes to setup, ask in our [Discord channel](https://discord.gg/Xd5f6WSh).
```bash
npm install vitrus
bun add vitrus
```
[](https://app.vitrus.ai)
```typescript
import Vitrus from "vitrus";
// Initialize the client with all options
const vitrus = new Vitrus({
apiKey: process.env["VITRUS_API_KEY"],
});
```
<br/>
**Workflows** have a similar schema as AI tools, so it connects perfectly with [OpenAI function Calling](https://platform.openai.com/docs/guides/function-calling?api-mode=chat). Making Workflows great to compose complex physical tasks with AI Agents. The following is a simple example of how to run a workflow:
```typescript
// running a basic workflow
const result = await vitrus.workflow("hello-world", {
prompt: "hello world!",
});
console.log(result);
```
Workflows are executed in Cloud GPUs (e.g. `Nvidia A100`), and combine multiple AI models for complex tasks.
## Available Workflows
We are continously updating the available workflows, and keeping them up to date with state-of-the-art (SOTA) AI models. For the latest list of workflows, you can execute:
```ts
const workflows = vitrus.list_workflows();
console.log(workflows);
```
<details><summary>Example JSON Response</summary>
```json
[
{
"type": "function",
"function": {
"name": "perception-encoder",
"description": "Encodes perception data based on specified parameters.",
"parameters": {
"type": "object",
"properties": {
"inputData": {
"type": "object",
"description": "The raw data to encode.",
"additionalProperties": true
},
"encodingType": {
"type": "string",
"description": "The encoding method to use (e.g., 'base64', 'json')."
}
},
"required": ["inputData"]
}
}
}
]
```
</details>
<br/>
Create a world at [app.vitrus.ai](https://app.vitrus.ai).
```typescript
import Vitrus from "vitrus";
// Initialize the client
const vitrus = new Vitrus({
apiKey: "your-api-key",
baseUrl: "ws://localhost:3001", //hosted Vitrus URL
});
```
```ts
import Vitrus from "vitrus";
const vitrus = new Vitrus({
apiKey: "<your-api-key>",
world: "<world-id>",
});
const actor = await vitrus.actor("forest", {
human: "Tom Hanks",
eyes: "green",
});
actor.on("walk", (args: any) => {
console.log("received", args);
return "run forest, run!";
});
```
On the Agent side, once connected to, the actor can be treated as "functions".
```ts
import Vitrus from "vitrus";
const vitrus = new Vitrus({
apiKey: "<your-api-key>",
world: "<world-id>", //must match actor's world
});
const actor = await vitrus.actor("forest");
const resp = await actor.run("walk", {
direction: "front",
});
```
---
Vitrus workflows, worlds, actors and agents runs on top of Distributed Actor Orchestration (DAO). A lightweight cloud system that enables the cross-communication of agents-world-actors.