# Gemini Embedding 001 - Core.Today AI API

> Google's text embedding model for generating vector representations. Optimized for semantic search, clustering, and similarity tasks.

- **Provider**: Google
- **Model ID**: gemini-embedding-001
- **Category**: LLM
- **Credits**: 0.1 per 1K tokens
- **Speed**: Fast
- **Quality**: High

## Model Specifications
- **Context Window**: 2K tokens
- **Max Output**: 0K tokens
- **Training Cutoff**: August 2024
- **Supported Formats**: text
- **Compatible SDK**: Google AI

### Token Pricing (per 1M tokens)
- **Input Tokens**: 300 credits ($0.20)
- **Output Tokens**: 0 credits ($0.00)

## Features
- High-quality text embeddings
- 2,048 token max input per request
- Output dimensions: 128–3,072 (default 3,072; recommended 768 / 1,536 / 3,072)
- Matryoshka Representation Learning (MRL) for flexible dimensionality
- 8 task types: SEMANTIC_SIMILARITY, CLASSIFICATION, CLUSTERING, RETRIEVAL_DOCUMENT, RETRIEVAL_QUERY, CODE_RETRIEVAL_QUERY, QUESTION_ANSWERING, FACT_VERIFICATION
- Optimized for semantic search, clustering, classification, and RAG

## Use Cases
- Semantic search
- Document clustering
- Similarity matching
- Recommendation systems
- RAG (Retrieval-Augmented Generation)

## API Endpoint

Base URL: https://api.core.today
Endpoint: POST /llm/gemini/v1beta/openai/embeddings

## Authentication
Header: Authorization: Bearer YOUR_API_KEY

Note: LLM endpoints use OpenAI-compatible format with Authorization Bearer token.

## Input Parameters

### Required
- **input**: string | array - Text or array of texts to embed. Max 2,048 input tokens per item.
- **model**: string (default: gemini-embedding-001) - Model identifier.

### Optional
- **dimensions**: integer (default: 3072) - Output embedding dimensionality. Range: 128–3,072. Recommended: 768, 1,536, 3,072. Manual L2 normalization required for non-3,072 dimensions.
- **task_type**: string - Optimize embedding for a downstream task. One of: SEMANTIC_SIMILARITY, CLASSIFICATION, CLUSTERING, RETRIEVAL_DOCUMENT, RETRIEVAL_QUERY, CODE_RETRIEVAL_QUERY, QUESTION_ANSWERING, FACT_VERIFICATION.
  Options: SEMANTIC_SIMILARITY, CLASSIFICATION, CLUSTERING, RETRIEVAL_DOCUMENT, RETRIEVAL_QUERY, CODE_RETRIEVAL_QUERY, QUESTION_ANSWERING, FACT_VERIFICATION

## Examples

### Text Embedding
Generate embeddings for semantic search
```json
{
  "model": "gemini-embedding-001",
  "input": "What is the meaning of life?"
}
```

## Response Format
```json
{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Response text here"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "total_tokens": 150
  }
}
```

## Tips
- Max input: 2,048 tokens per item — split long documents into chunks first
- Default output: 3,072 dimensions; use 768 or 1,536 to reduce storage/cost via MRL
- Set `task_type` to match your use case (e.g., RETRIEVAL_DOCUMENT for indexing, RETRIEVAL_QUERY for queries)
- Manually L2-normalize embeddings if you request a non-3,072 dimensionality
- Combine with vector databases for RAG pipelines
- Batch multiple texts in a single request for efficiency

## Documentation
https://ai.google.dev/gemini-api/docs/embeddings