jina embeddings v4 vllm retrieval EU AI Act Compliance Profile

<br><br>

Jina Embeddings v4: Universal Embeddings for Multimodal Multilingual Retrieval

Original Model | Blog | Technical Report | API

Model Overview

This repository hosts a vLLM-compatible version of jina-embeddings-v4 with the retrieval adapter merged into the base Qwen2.5-VL weights. This architecture modification enables native compatibility with vLLM without requiring custom adapter-handling code.

Usage

import torch
from PIL import Image

from vllm import LLM
from vllm.config import PoolerConfig
from vllm.inputs.data import TextPrompt

# Initialize model
model = LLM(
    model="jinaai/jina-embeddings-v4-vllm-retrieval",
    task="auto",
    override_pooler_config=PoolerConfig(pooling_type="ALL", normalize=False),
    dtype="float16",
)

# Create text prompts
query = "Overview of climate change impacts on coastal cities"
query_prompt = TextPrompt(
    prompt=f"Query: {query}"
)

passage = "The impacts of climate change on coastal cities are significant.."
passage_prompt = TextPrompt(
    prompt=f"Passage: {passage}"
)

# Create image prompt
image = Image.open("<path_to_image>")
image_prompt = TextPrompt(
    prompt="<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>Describe the image.<|im_end|>\n",
    multi_modal_data={"image": image},
)

# Encode all prompts
prompts = [query_prompt, passage_prompt, image_prompt]
outputs = model.encode(prompts)


def get_em

Your obligations as a jina embeddings v4 vllm retrieval deployer

Track your progress

Sign up to track obligations, receive deadline reminders, and generate audit packages.

Tags

Languages

Library