Feature Request: Add support for Text Embedding API (/v1/embeddings)

[bendusy]

🚀 Feature Request

Summary

Request to add support for the OpenAI-compatible Text Embedding API endpoint (/v1/embeddings).

Motivation

Text embeddings are essential for many AI applications including:

• Semantic search
• RAG (Retrieval-Augmented Generation) systems
• Document similarity analysis
• Vector database integration

Currently, CLIProxyAPI only proxies chat/completion endpoints, which limits its use in applications requiring both LLM inference and embedding capabilities.

Current State

After reviewing the documentation and source code:

Documentation:

• README.md and related docs do not mention Text Embedding API support
• The term "embedding" in docs (e.g., "Reusable Go SDK for embedding the proxy") refers to SDK integration, not AI text embeddings

Code (internal/api/server.go):
Currently registered endpoints:

• /v1/chat/completions ✅
• /v1/completions ✅
• /v1/messages ✅
• /v1beta/models ✅
• /v1/embeddings ❌ Missing

Proposed Solution

Add a new endpoint /v1/embeddings that proxies to upstream embedding models such as:

• text-embedding-3-small
• text-embedding-3-large
• text-embedding-ada-002
• Other provider-specific embedding models

Expected API Format

Request:

{
  "model": "text-embedding-3-small",
  "input": "Your text string goes here"
}

Response:

{
  "object": "list",
  "data": [
    {
      "object": "embedding",
      "embedding": [0.0023064255, -0.009327292, ...],
      "index": 0
    }
  ],
  "model": "text-embedding-3-small",
  "usage": {
    "prompt_tokens": 5,
    "total_tokens": 5
  }
}

Additional Context

This would make CLIProxyAPI a more complete solution for AI application development, allowing users to consolidate both LLM and embedding API calls through a single proxy.

[wangai-studio]

Hi! I've submitted a PR to implement this feature: #1191

The implementation includes:

• ✅ /v1/embeddings endpoint
• ✅ Dynamic endpoint detection for openai-compatibility providers
• ✅ Comprehensive unit tests
• ✅ Configuration examples

Please review when you have time. Thanks!

[Saopanda]

i need this

[razgzy]

we need this

[TAT-Hins]

Any updates here? embedding models calling is quite important for a lot of people.

[SpiGAndromeda]

The feature would be highly appreciated. By now, I have to think about moving away from CPA because of this missing feature. The alternative would be to operate a parallel system just for embeddings.