Managing Assistants & Threads

Overview

Threads and Messages represent a conversation session between an Assistant and a user. There is no limit to the number of Messages you can store in a Thread. When the size of the Messages exceeds the model's context window, the Thread will smartly truncate messages before fully dropping the least important ones.

API Reference

Delete an assistant.

Deletes an assistant, removing it from the system.

path=/assistants/{assistantId} method=delete

Request

JavaScript

const response = await fetch('https://api.umamiai.xyz/v1/assistants/{assistantId}', {
method: 'DELETE',
headers: {"Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
});
const data = await response.json();

Python

import requests

response = requests.delete(
  "https://api.umamiai.xyz/v1/assistants/{assistantId}",
  headers={"Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
)
data = response.json()

Curl

curl -L \
  -H 'Authorization: Bearer YOUR_UMAMIAI_API_KEY' \
  -X DELETE \
  'https://api.umamiai.xyz/v1/assistants/{assistantId}'

Response

{
  "id": "text",
  "object": "assistant.deleted",
  "deleted": false
}

Create a new assistant.

Creates a new AI assistant, configuring it to perform tasks based on predefined models and settings.

path=/assistants method=post

Request

JavaScript

const response = await fetch('https://api.umamiai.xyz/v1/assistants', {
  method: 'POST',
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer YOUR_UMAMIAI_API_KEY",
  },
  body: JSON.stringify({
    "name": "text",
    "description": "text",
    "model": "text"
  }),
});
const data = await response.json();

Python

import requests

response = requests.post(
    "https://api.umamiai.xyz/v1/assistants",
    headers={"Content-Type":"application/json","Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
    json={"name":"text","description":"text","model":"text"}
)
data = response.json()

Curl

curl -L \
  -X POST \
  -H 'Authorization: Bearer YOUR_UMAMIAI_API_KEY' \
  -H 'Content-Type: application/json' \
  'https://api.umamiai.xyz/v1/assistants' \
  -d '{"name":"text","description":"text","model":"text"}'

Response

{
  "id": "text",
  "created_at": 0,
  "description": "text",
  "instructions": "text",
  "model": "text",
  "name": "text",
  "object": "assistant",
  "tools": [],
  "response_format": {
    "type": "json_schema",
    "json_schema": {}
  },
  "temperature": 0,
  "top_p": 0
}

List all available assistants.

Returns a list of AI assistants that are available, allowing users to view and manage their virtual assistants.

path=/assistants method=get

Request

JavaScript

const response = await fetch('https://api.umamiai.xyz/v1/assistants', {
  method: 'GET',
  headers: {"Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
});
const data = await response.json();

Python

import requests

response = requests.get(
    "https://api.umamiai.xyz/v1/assistants",
    headers={"Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
)
data = response.json()

Curl

curl -L \
  -H 'Authorization: Bearer YOUR_UMAMIAI_API_KEY' \
  'https://api.umamiai.xyz/v1/assistants'

Response

{
  "object": "list",
  "data": [
    {
      "id": "text",
      "created_at": 0,
      "description": "text",
      "instructions": "text",
      "model": "text",
      "name": "text",
      "object": "assistant",
      "tools": [],
      "response_format": {
        "type": "json_schema",
        "json_schema": {}
      },
      "temperature": 0,
      "top_p": 0
    }
  ]
}

Get details of a specific assistant.

Retrieves information about a specific assistant, including its current configuration and status.

path=/assistants/{assistantId} method=get

Request

JavaScript

const response = await fetch('https://api.umamiai.xyz/v1/assistants/{assistantId}', {
  method: 'GET',
  headers: {"Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
});
const data = await response.json();

Python

import requests

response = requests.get(
    "https://api.umamiai.xyz/v1/assistants/{assistantId}",
    headers={"Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
)
data = response.json()

Curl

curl -L \
  -H 'Authorization: Bearer YOUR_UMAMIAI_API_KEY' \
  'https://api.umamiai.xyz/v1/assistants/{assistantId}'

Response

{
  "object": "list",
  "data": [
    {
      "id": "text",
      "created_at": 0,
      "description": "text",
      "instructions": "text",
      "model": "text",
      "name": "text",
      "object": "assistant",
      "tools": [],
      "response_format": {
        "type": "json_schema",
        "json_schema": {}
      },
      "temperature": 0,
      "top_p": 0
    }
  ]
}

Update an assistant’s settings.

Updates the configuration or settings of an existing assistant.

path=/assistants/{assistantId} method=post

Request

JavaScript

const response = await fetch('https://api.umamiai.xyz/v1/assistants/{assistantId}', {
  method: 'POST',
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer YOUR_UMAMIAI_API_KEY",
  },
  body: JSON.stringify({}),
});
const data = await response.json();

Python

import requests

response = requests.post(
    "https://api.umamiai.xyz/v1/assistants/{assistantId}",
    headers={"Content-Type":"application/json","Authorization": "Bearer YOUR_UMAMIAI_API_KEY",},
    json={}
)
data = response.json()

Curl

curl -L \
  -X POST \
  -H 'Authorization: Bearer YOUR_UMAMIAI_API_KEY' \
  -H 'Content-Type: application/json' \
  'https://api.umamiai.xyz/v1/assistants/{assistantId}' \
  -d '{}'

Response

{
  "id": "text",
  "created_at": 0,
  "description": "text",
  "instructions": "text",
  "model": "text",
  "name": "text",
  "object": "assistant",
  "tools": [],
  "response_format": {
    "type": "json_schema",
    "json_schema": {}
  },
  "temperature": 0,
  "top_p": 0
}

Using OpenAI threads with UmamiAI API

const { OpenAI } = require('openai');
const dotenv = require('dotenv');
const prompts = require('prompts');
const axios = require('axios').default;

dotenv.config({ path: ['.env.default', '.env'], override: true });

const STEP_NAMES = {
  CREATE_ASSISTANT: 'createAssistant',
  CREATE_THREAD: 'selectAssistant',
  RUN_THREAD: 'runThread',
  INITIAL: 'initial',
};

const steps = {
  [STEP_NAMES.CREATE_ASSISTANT]: async ({ api, vendorByModelId }) => {
    const { model, name, instructions, description } = await prompts([
      {
        type: 'autocomplete',
        name: 'model',
        message: 'Your assistant model',
        choices: Object.entries(vendorByModelId).map(([id, vendor]) => ({
          title: id,
          value: id,
          description: vendor,
        })),
      },
      {
        type: 'text',
        name: 'name',
        message: 'Your assistant name',
      },
      {
        type: 'text',
        name: 'instructions',
        message: 'Your assistant instructions',
      },
      {
        type: 'text',
        name: 'description',
        message: 'Your assistant description',
      },
    ]);

    await api.beta.assistants.create({
      model,
      name,
      description,
      instructions,
    });

    return { step: 'initial' };
  },
  [STEP_NAMES.INITIAL]: async ({ api }) => {
    const assistants = await api.beta.assistants.list({ order: 'desc' });
    const { assistantId } = await prompts({
      type: 'select',
      name: 'assistantId',
      message: 'Select assistant',
      choices: [
        { title: 'Create new', value: 'new' },
        ...assistants.data.map((item) => ({ title: item.name, description: item.model, value: item.id })),
      ],
    });

    if (assistantId === 'new') {
      return { step: STEP_NAMES.CREATE_ASSISTANT };
    }

    return { step: STEP_NAMES.CREATE_THREAD, assistantId };
  },
  [STEP_NAMES.CREATE_THREAD]: async ({ api }) => {
    const thread = await api.beta.threads.create({ messages: [] });
    return { step: STEP_NAMES.RUN_THREAD, thread };
  },

  [STEP_NAMES.RUN_THREAD]: async ({ api, assistantId, thread }) => {
    const { text } = await prompts([
      {
        type: 'text',
        name: 'text',
        message: 'Message',
      },
    ]);

    await api.beta.threads.messages.create(thread.id, {
      role: 'user',
      content: text,
      metadata: {
        userId: 'example-1',
      },
    });

    await api.beta.threads.runs.createAndPoll(thread.id, { assistant_id: assistantId });
    const messages = await api.beta.threads.messages.list(thread.id, { order: 'desc', limit: 1 });
    const msg = messages.data[0].content.find((item) => item.type === 'text').text.value;
    console.log(`Assistant: ${msg}`);

    return { step: STEP_NAMES.RUN_THREAD };
  },
};

const main = async () => {
  const api = new OpenAI({
    baseURL: 'https://api.umamiai.xyz/v1',
    apiKey: process.env.API_TOKEN,
  });

  const vendorByModelId = await axios.get('https://api.umamiai.xyz/v1/models').then((r) => r.data);
  let step = 'initial';
  let payload = { api, vendorByModelId };
  while (step) {
    const result = await steps[step](payload);
    const { step: nextStep, ...nextPayload } = result ?? {};

    payload = { ...payload, ...nextPayload };
    step = nextStep;
  }
};

main();

Overview​

API Reference​

Delete an assistant.​

Create a new assistant.​

List all available assistants.​

Get details of a specific assistant.​

Update an assistant’s settings.​

Using OpenAI threads with UmamiAI API​

Overview

API Reference

Delete an assistant.

Create a new assistant.

List all available assistants.

Get details of a specific assistant.

Update an assistant’s settings.

Using OpenAI threads with UmamiAI API