# Multilingual E5 Large - Core.Today AI API

> Multilingual text embeddings (74M+ runs) — 1024-dimension vectors across 100 languages including Korean. Pairs perfectly with Core.Today customer databases' vector search (knn_vector).

- **Provider**: E5
- **Model ID**: beautyyuyanli/multilingual-e5-large
- **Category**: Image Generation
- **Credits**: 2 per run (batch multiple texts in one call)
- **Speed**: Fast
- **Quality**: High

## Features
- 1024-dim embeddings, 100 languages incl. Korean
- Batch embedding of multiple texts per call
- Normalized embeddings option (cosine-ready)
- Pairs with customer DB vector search

## Use Cases
- Semantic search over documents
- RAG pipelines with Core.Today databases
- Multilingual deduplication
- Recommendation similarity
- Clustering and topic discovery

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Optional
- **texts**: string (default: ["In the water, fish are swimming.", "Fish swim in the water.", "A book lies open on the table."]) - Texts to embed as a JSON list of strings, e.g. ["hello", "world"]
- **batch_size**: integer (default: 32) - Batch size for processing
- **normalize_embeddings**: boolean (default: true) - Normalize embeddings (recommended for cosine similarity)

## Examples

### Embed Korean Documents
Embed multiple Korean texts in one call
```json
{
  "model": "beautyyuyanli/multilingual-e5-large",
  "input": {
    "texts": "[\"코어닷투데이는 AI API 게이트웨이입니다.\", \"크레딧 기반 과금을 지원합니다.\"]",
    "normalize_embeddings": true
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- Batch many texts per call — one run embeds the whole list for 3 credits
- Store vectors in a Core.Today database with a knn_vector field (dimension: 1024) for search
- Keep normalize_embeddings: true so cosine similarity works out of the box
- E5 convention: prefix queries with 'query: ' and passages with 'passage: ' for best retrieval

## Documentation
https://replicate.com/beautyyuyanli/multilingual-e5-large