feat: Add Deep Infra to known endpoints

[erkserkserks]

Summary

Add Deep Infra to known endpoints

Motivation: qwen2.5-72b-instruct is a strong model that is competitive with proprietary models in coding: https://livebench.ai/

Deep Infra is the currently the cheapest way to run this model and many open weight models:
https://openrouter.ai/models/qwen/qwen-2.5-72b-instruct/providers
https://openrouter.ai/models/meta-llama/llama-3.1-405b-instruct/providers
https://openrouter.ai/models/meta-llama/llama-3.2-90b-vision-instruct

Sample custom endpoint in librechat.yaml:

endpoints:
  custom:
    # Deep Infra
    - name: 'DeepInfra'
      apiKey: '${DEEPINFRA_API_KEY}'
      baseURL: 'https://api.deepinfra.com/v1/openai/'
      models:
        default:
          [
            'Qwen/Qwen2.5-72B-Instruct'
          ]
        fetch: false
      titleConvo: true
      titleModel: 'meta-llama/Llama-3.2-3B-Instruct'
      summarize: false
      summaryModel: 'Qwen/Qwen2.5-72B-Instruct'
      forcePrompt: false
      modelDisplayLabel: "Qwen"

Change Type

• New feature (non-breaking change which adds functionality)

Testing

1. Add Deep Infra custom endpoint to librechat.yaml
2. npm run frontend; npm run backend
3. Visit LibreChat, add Deep Infra API key, test chat feature

Checklist

• My code adheres to this project's style guidelines
• I have performed a self-review of my own code
• My changes do not introduce new warnings

[erkserkserks]

Hi @danny-avila could you please review this?

I have been using OpenRouter for open weight LLMs, and I have noticed that the majority of my requests are routed to DeepInfra. Since OpenRouter load balances based by price, DeepInfra is probably one of the most popular providers on OpenRouter.

I believe the DeepInfra meets the threshold for being a notable provider:

1. A large amount (the majority?) of open weight model requests on OpenRouter get routed to DeepInfra
2. TypingMind supports a limited number of providers, and DeepInfra is supported natively: https://docs.typingmind.com/chat-models-settings/use-with-deepinfra
3. Peer providers like Groq consider DeepInfra to be a competitor. Here's an image from Groq's homepage:
   

It would be great to add the ability use DeepInfra directly because of improved pricing, lower latency, and data privacy.