Skip to content

All related to Nous, MaLLaM πŸŒ™, Retrieval, Classifier, Speech, Translation, cookbooks πŸ“– , bugs πŸ› and feature requests πŸ™‡πŸ½.

License

Notifications You must be signed in to change notification settings

mesolitica/Nous

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

22 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Nous

All related to Nous, MaLLaM πŸŒ™, Retrieval, Classifier, Speech, Translation, cookbooks πŸ“– , bugs πŸ› and feature requests πŸ™‡πŸ½. This repository only applicable for commercial models from https://mesolitica.com, not open models from https://mesolitica.com/opensource

MaLLaM πŸŒ™

MaLLaM πŸŒ™ is Multi-lingual Malaysian Chat Language Model, 32k context length, Malaysian centric and private, in the future we will support longer context length and code interpreter, get your API key at https://app.nous.mesolitica.com/

API documentation at https://llm-router.nous.mesolitica.com/scalar#tag/default/post/chat/completions

Currently we only support Chat Completion.

OpenAI compatible

MaLLaM πŸŒ™ is compatible with OpenAI library for Python and Node JS.

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://llm-router.nous.mesolitica.com",
)

completion = client.chat.completions.create(
  model="mallam-small",
  messages=[
    { "role": "system", "content": "You are a helpful assistant." },
    { "role": "user", "content": "Hello!" }
  ]
)
print(completion.choices[0].message)
ChatCompletionMessage(content='hello! Bagaimanakah saya boleh membantu anda hari ini?', role='assistant', function_call=None, tool_calls=None)

Node JS

const OpenAI = require('openai');

const openai = new OpenAI({
    baseURL: 'https://llm-router.nous.mesolitica.com',
});

async function main() {
    const completion = await openai.chat.completions.create({
        model: "mallam-small",
        messages: [
            { "role": "system", "content": "Awak pembantu AI yang berguna." },
            { "role": "user", "content": "Hello!" }
        ],
    });

    console.log(completion.choices[0]);
}

main();
{
  index: 0,
  message: {
    role: 'assistant',
    content: 'hello! Saya di sini untuk memberikan maklumat dan menjawab sebarang soalan yang anda ada.'
  },
  logprobs: null,
  finish_reason: 'stop'
}

cURL

curl -X 'POST' \
'https://llm-router.nous.mesolitica.com/chat/completions' \
-H 'accept: application/json' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer api_key' \
-d '{
"model": "mallam-small",
"temperature": 0.9,
"top_p": 0.95,
"top_k": 50,
"max_tokens": 1024,
"messages": [
    {
        "role": "system",
        "content": "Awak pembantu AI yang berguna."
    },
    {
        "role": "user",
        "content": "Hello!"
    }
],
"tools": [],
"stream": false
}'
{"id":"cmpl-2def863689fa4502bc7d554a687d1f8c","object":"chat.completion","created":1570278,"model":"mallam-small","choices":[{"index":0,"message":{"role":"assistant","content":"helo! Bagaimana saya boleh membantu anda hari ini?"},"logprobs":null,"finish_reason":"stop"}],"usage":{"prompt_tokens":25,"total_tokens":48,"completion_tokens":23}}

We covered RAG, JSON prompts and so much more.

Evaluation

  1. json-mode-eval,

Originally from https://huggingface.co/datasets/NousResearch/json-mode-eval, this test is to test how good the models able to convert human natural text into JSON output given the OpenAPI schema.

Speech

Speech is End-to-End streamable Malaysian Speech-to-Text and Speech Translation with Speaker Diarization, get your API key at https://app.nous.mesolitica.com/

API documentation at https://llm-router.nous.mesolitica.com/scalar#tag/default/post/audio/transcriptions

Self-hosted Enterprise

If you are interested to self-host Nous in your virtual private network, Contact us at [email protected] or [email protected] to know more.

What do you get

  1. ♾️ usage of all quota, MaLLaM πŸŒ™, Speech, Translation and chatbots.
  2. Up-to-date enterprise versioning including evolution of MaLLaM πŸŒ™, Speech, Translation and chatbots.
  3. Can be either on private cloud or on premise, Dashboard and all the APIs will be hosted using virtual private IP.

Hardware

  1. If you do not have hardware accelerator, we partner with NVIDIA APAC to include hardware solution with support.
  2. Bring your own hardware accelerator, we do not provide support.

Software

We provide quarterly and yearly licensing, and yearly licencing is much more cheaper in the long term, include support.

About

All related to Nous, MaLLaM πŸŒ™, Retrieval, Classifier, Speech, Translation, cookbooks πŸ“– , bugs πŸ› and feature requests πŸ™‡πŸ½.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages