使用 ColPali、Cohere 重排器和多模态嵌入进行重排¶

ColPali：ColPali 是一个基于新颖模型架构和训练策略的模型，该架构和策略基于视觉语言模型 (VLM)，可以利用其视觉特征高效地索引文档。

在此笔记本中，我们将演示如何使用 ColPali 作为重排器，对使用 Cohere 多模态嵌入检索到的图像进行重排。

为了演示，以下是步骤

1. 从相关的维基百科文章下载文本、图像和原始 PDF 文件。
2. 使用 Cohere 多模态嵌入为文本和图像构建多模态索引。
3. 使用多模态检索器同时检索查询相关的文本和图像。
4. 使用 Cohere 重排器重排文本节点，使用 ColPali 重排图像节点。
5. 使用 Multi-Modal Query Engine 和 gpt-4o 多模态 LLM 生成查询响应。

安装¶

我们将使用 Cohere 多模态嵌入进行检索，使用 ColPali 作为图像节点的重排器，使用 Cohere 重排器作为文本节点的重排器，使用 Qdrant 向量存储和 OpenAI 多模态 LLM 进行响应生成。

In [ ]

%pip install llama-index-postprocessor-colpali-rerank
%pip install llama-index-postprocessor-cohere-rerank
%pip install llama-index-embeddings-cohere
%pip install llama-index-vector-stores-qdrant
%pip install llama-index-multi-modal-llms-openai

%pip install llama-index-postprocessor-colpali-rerank %pip install llama-index-postprocessor-cohere-rerank %pip install llama-index-embeddings-cohere %pip install llama-index-vector-stores-qdrant %pip install llama-index-multi-modal-llms-openai

设置 API 密钥¶

Cohere - 多模态检索

OpenAI - 多模态 LLM。

In [ ]

import os

os.environ["COHERE_API_KEY"] = "<YOUR COHERE API KEY>"

os.environ["OPENAI_API_KEY"] = "<YOUR OPENAI API KEY>"

import os os.environ["COHERE_API_KEY"] = "" os.environ["OPENAI_API_KEY"] = ""

实用函数¶

1. get_wikipedia_images：从指定标题的维基百科页面获取图片 URL。
2. plot_images：绘制指定图片路径列表中的图片。
3. delete_large_images：删除指定目录中大于 5 MB 的图片。

注意：Cohere API 接受小于 5MB 的图片。

In [ ]

import requests
import matplotlib.pyplot as plt
from PIL import Image
from pathlib import Path
import urllib.request
import os


def get_wikipedia_images(title):
    """
    Get the image URLs from the Wikipedia page with the specified title.
    """
    response = requests.get(
        "https://en.wikipedia.org/w/api.php",
        params={
            "action": "query",
            "format": "json",
            "titles": title,
            "prop": "imageinfo",
            "iiprop": "url|dimensions|mime",
            "generator": "images",
            "gimlimit": "50",
        },
    ).json()
    image_urls = []
    for page in response["query"]["pages"].values():
        if page["imageinfo"][0]["url"].endswith(".jpg") or page["imageinfo"][
            0
        ]["url"].endswith(".png"):
            image_urls.append(page["imageinfo"][0]["url"])
    return image_urls


def plot_images(image_paths):
    """
    Plot the images in the specified list of image paths.
    """
    images_shown = 0
    plt.figure(figsize=(16, 9))
    for img_path in image_paths:
        if os.path.isfile(img_path):
            image = Image.open(img_path)

            plt.subplot(2, 3, images_shown + 1)
            plt.imshow(image)
            plt.xticks([])
            plt.yticks([])

            images_shown += 1
            if images_shown >= 9:
                break


def delete_large_images(folder_path):
    """
    Delete images larger than 5 MB in the specified directory.
    """
    # List to hold the names of deleted image files
    deleted_images = []

    # Iterate through each file in the directory
    for file_name in os.listdir(folder_path):
        if file_name.lower().endswith(
            (".png", ".jpg", ".jpeg", ".gif", ".bmp")
        ):
            # Construct the full file path
            file_path = os.path.join(folder_path, file_name)
            # Get the size of the file in bytes
            file_size = os.path.getsize(file_path)
            # Check if the file size is greater than 5 MB (5242880 bytes) and remove it
            if file_size > 5242880:
                os.remove(file_path)
                deleted_images.append(file_name)
                print(
                    f"Image: {file_name} was larger than 5 MB and has been deleted."
                )

import requests import matplotlib.pyplot as plt from PIL import Image from pathlib import Path import urllib.request import os def get_wikipedia_images(title): """ 从指定标题的维基百科页面获取图片 URL。 """ response = requests.get( "https://en.wikipedia.org/w/api.php", params={ "action": "query", "format": "json", "titles": title, "prop": "imageinfo", "iiprop": "url|dimensions|mime", "generator": "images", "gimlimit": "50", }, ).json() image_urls = [] for page in response["query"]["pages"].values(): if page["imageinfo"][0]["url"].endswith(".jpg") or page["imageinfo"][ 0 ]["url"].endswith(".png"): image_urls.append(page["imageinfo"][0]["url"]) return image_urls def plot_images(image_paths): """ 绘制指定图片路径列表中的图片。 """ images_shown = 0 plt.figure(figsize=(16, 9)) for img_path in image_paths: if os.path.isfile(img_path): image = Image.open(img_path) plt.subplot(2, 3, images_shown + 1) plt.imshow(image) plt.xticks([]) plt.yticks([]) images_shown += 1 if images_shown >= 9: break def delete_large_images(folder_path): """ 删除指定目录中大于 5 MB 的图片。 """ # List to hold the names of deleted image files deleted_images = [] # Iterate through each file in the directory for file_name in os.listdir(folder_path): if file_name.lower().endswith( (".png", ".jpg", ".jpeg", ".gif", ".bmp") ): # Construct the full file path file_path = os.path.join(folder_path, file_name) # Get the size of the file in bytes file_size = os.path.getsize(file_path) # Check if the file size is greater than 5 MB (5242880 bytes) and remove it if file_size > 5242880: os.remove(file_path) deleted_images.append(file_name) print( f"图片：{file_name} 大于 5 MB，已删除。" )

/Users/ravithejad/Desktop/llamaindex/lib/python3.9/site-packages/urllib3/__init__.py:35: NotOpenSSLWarning: urllib3 v2 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020
  warnings.warn(

从维基百科下载文本和图片¶

我们将从以下维基百科页面下载相关的文本和图片。

1. Audi e-tron
2. Ford Mustang
3. Porsche Taycan

In [ ]

image_uuid = 0
# image_metadata_dict stores images metadata including image uuid, filename and path
image_metadata_dict = {}
MAX_IMAGES_PER_WIKI = 10

wiki_titles = {
    "Audi e-tron",
    "Ford Mustang",
    "Porsche Taycan",
}


data_path = Path("mixed_wiki")
if not data_path.exists():
    Path.mkdir(data_path)

for title in wiki_titles:
    response = requests.get(
        "https://en.wikipedia.org/w/api.php",
        params={
            "action": "query",
            "format": "json",
            "titles": title,
            "prop": "extracts",
            "explaintext": True,
        },
    ).json()
    page = next(iter(response["query"]["pages"].values()))
    wiki_text = page["extract"]

    with open(data_path / f"{title}.txt", "w") as fp:
        fp.write(wiki_text)

    images_per_wiki = 0
    try:
        list_img_urls = get_wikipedia_images(title)

        for url in list_img_urls:
            if (
                url.endswith(".jpg")
                or url.endswith(".png")
                or url.endswith(".svg")
            ):
                image_uuid += 1
                urllib.request.urlretrieve(
                    url, data_path / f"{image_uuid}.jpg"
                )
                images_per_wiki += 1
                if images_per_wiki > MAX_IMAGES_PER_WIKI:
                    break
    except:
        print(str(Exception("No images found for Wikipedia page: ")) + title)
        continue

image_uuid = 0 # image_metadata_dict 存储图片元数据，包括图片 uuid、文件名和路径 image_metadata_dict = {} MAX_IMAGES_PER_WIKI = 10 wiki_titles = { "Audi e-tron", "Ford Mustang", "Porsche Taycan", } data_path = Path("mixed_wiki") if not data_path.exists(): Path.mkdir(data_path) for title in wiki_titles: response = requests.get( "https://en.wikipedia.org/w/api.php", params={ "action": "query", "format": "json", "titles": title, "prop": "extracts", "explaintext": True, }, ).json() page = next(iter(response["query"]["pages"].values())) wiki_text = page["extract"] with open(data_path / f"{title}.txt", "w") as fp: fp.write(wiki_text) images_per_wiki = 0 try: list_img_urls = get_wikipedia_images(title) for url in list_img_urls: if ( url.endswith(".jpg") or url.endswith(".png") or url.endswith(".svg") ): image_uuid += 1 urllib.request.urlretrieve( url, data_path / f"{image_uuid}.jpg" ) images_per_wiki += 1 if images_per_wiki > MAX_IMAGES_PER_WIKI: break except: print(str(Exception("找不到维基百科页面图片：")) + title) continue

删除较大图片文件¶

Cohere 多模态嵌入模型接受小于 5MB 的文件，因此这里我们删除较大的图片文件。

In [ ]

delete_large_images(data_path)

delete_large_images(data_path)

Image: 8.jpg was larger than 5 MB and has been deleted.
Image: 13.jpg was larger than 5 MB and has been deleted.
Image: 11.jpg was larger than 5 MB and has been deleted.
Image: 21.jpg was larger than 5 MB and has been deleted.
Image: 23.jpg was larger than 5 MB and has been deleted.
Image: 32.jpg was larger than 5 MB and has been deleted.
Image: 19.jpg was larger than 5 MB and has been deleted.
Image: 4.jpg was larger than 5 MB and has been deleted.
Image: 5.jpg was larger than 5 MB and has been deleted.
Image: 7.jpg was larger than 5 MB and has been deleted.
Image: 6.jpg was larger than 5 MB and has been deleted.
Image: 1.jpg was larger than 5 MB and has been deleted.

设置嵌入模型和 LLM。¶

使用 Cohere 多模态嵌入模型进行检索，使用 OpenAI 多模态 LLM 进行响应生成。

In [ ]

from llama_index.embeddings.cohere import CohereEmbedding
from llama_index.multi_modal_llms.openai import OpenAIMultiModal
from llama_index.core import Settings

Settings.embed_model = CohereEmbedding(
    api_key=os.environ["COHERE_API_KEY"],
    model_name="embed-english-v3.0",  # current v3 models support multimodal embeddings
)

gpt_4o = OpenAIMultiModal(model="gpt-4o", max_new_tokens=4096)

from llama_index.embeddings.cohere import CohereEmbedding from llama_index.multi_modal_llms.openai import OpenAIMultiModal from llama_index.core import Settings Settings.embed_model = CohereEmbedding( api_key=os.environ["COHERE_API_KEY"], model_name="embed-english-v3.0", # 当前 v3 模型支持多模态嵌入 ) gpt_4o = OpenAIMultiModal(model="gpt-4o", max_new_tokens=4096)

设置 Cohere 重排器¶

In [ ]

from llama_index.postprocessor.cohere_rerank import CohereRerank

cohere_rerank = CohereRerank(api_key=os.environ["COHERE_API_KEY"], top_n=3)

from llama_index.postprocessor.cohere_rerank import CohereRerank cohere_rerank = CohereRerank(api_key=os.environ["COHERE_API_KEY"], top_n=3)

设置 ColPali 重排器¶

In [ ]

from llama_index.postprocessor.colpali_rerank import ColPaliRerank

colpali_reranker = ColPaliRerank(
    top_n=3,
    model="vidore/colpali-v1.2",
    keep_retrieval_score=True,
    device="cuda",  # or "cpu" or "cuda:0" or "mps" for Apple
)

from llama_index.postprocessor.colpali_rerank import ColPaliRerank colpali_reranker = ColPaliRerank( top_n=3, model="vidore/colpali-v1.2", keep_retrieval_score=True, device="cuda", # 或 "cpu" 或 "cuda:0" 或 Apple 的 "mps" )

`config.hidden_act` is ignored, you should use `config.hidden_activation` instead.
Gemma's activation function will be set to `gelu_pytorch_tanh`. Please, use
`config.hidden_activation` if you want to override this behaviour.
See https://github.com/huggingface/transformers/pull/29402 for more details.

Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]

/Users/ravithejad/Desktop/llamaindex/lib/python3.9/site-packages/bitsandbytes/cextension.py:34: UserWarning: The installed version of bitsandbytes was compiled without GPU support. 8-bit optimizers, 8-bit multiplication, and GPU quantization are unavailable.
  warn("The installed version of bitsandbytes was compiled without GPU support. "

'NoneType' object has no attribute 'cadam32bit_grad_fp32'

加载数据¶

我们将加载下载的文本和图片数据。

In [ ]

from llama_index.core import SimpleDirectoryReader

documents = SimpleDirectoryReader("./mixed_wiki/").load_data()

from llama_index.core import SimpleDirectoryReader documents = SimpleDirectoryReader("./mixed_wiki/").load_data()

设置 Qdrant 向量存储¶

我们将使用 Qdrant 向量存储来存储图片和文本嵌入及相关的元数据。

In [ ]

from llama_index.core.indices import MultiModalVectorStoreIndex
from llama_index.vector_stores.qdrant import QdrantVectorStore
from llama_index.core import StorageContext

import qdrant_client

# Create a local Qdrant vector store
client = qdrant_client.QdrantClient(path="qdrant_mm_db")

text_store = QdrantVectorStore(
    client=client, collection_name="text_collection"
)
image_store = QdrantVectorStore(
    client=client, collection_name="image_collection"
)
storage_context = StorageContext.from_defaults(
    vector_store=text_store, image_store=image_store
)

from llama_index.core.indices import MultiModalVectorStoreIndex from llama_index.vector_stores.qdrant import QdrantVectorStore from llama_index.core import StorageContext import qdrant_client # 创建一个本地 Qdrant 向量存储 client = qdrant_client.QdrantClient(path="qdrant_mm_db") text_store = QdrantVectorStore( client=client, collection_name="text_collection" ) image_store = QdrantVectorStore( client=client, collection_name="image_collection" ) storage_context = StorageContext.from_defaults( vector_store=text_store, image_store=image_store )

创建 MultiModalVectorStoreIndex。¶

In [ ]

index = MultiModalVectorStoreIndex.from_documents(
    documents,
    storage_context=storage_context,
    image_embed_model=Settings.embed_model,
)

index = MultiModalVectorStoreIndex.from_documents( documents, storage_context=storage_context, image_embed_model=Settings.embed_model, )

WARNING:root:Payload indexes have no effect in the local Qdrant. Please use server Qdrant if you need payload indexes.
WARNING:root:Payload indexes have no effect in the local Qdrant. Please use server Qdrant if you need payload indexes.

测试检索¶

在这里我们创建一个检索器并进行测试。

In [ ]

retriever_engine = index.as_retriever(
    similarity_top_k=6, image_similarity_top_k=6
)

retriever_engine = index.as_retriever( similarity_top_k=6, image_similarity_top_k=6 )

In [ ]

query = "Which models of Porsche are discussed here?"
retrieval_results = retriever_engine.retrieve(query)

query = "Which models of Porsche are discussed here?" retrieval_results = retriever_engine.retrieve(query)

检查检索结果¶

In [ ]

from llama_index.core.response.notebook_utils import display_source_node
from llama_index.core.schema import ImageNode

retrieved_image = []
for res_node in retrieval_results:
    if isinstance(res_node.node, ImageNode):
        retrieved_image.append(res_node.node.metadata["file_path"])
    else:
        display_source_node(res_node, source_length=200)

plot_images(retrieved_image)

from llama_index.core.response.notebook_utils import display_source_node from llama_index.core.schema import ImageNode retrieved_image = [] for res_node in retrieval_results: if isinstance(res_node.node, ImageNode): retrieved_image.append(res_node.node.metadata["file_path"]) else: display_source_node(res_node, source_length=200) plot_images(retrieved_image)

节点 ID: e37183fe-8b2b-4bc5-b856-0a2b0221b011
相似度 0.4906621738377948
文本: === Aerodynamics === The Taycan Turbo has a drag coefficient of Cd=0.22, which the manufacturer claims is the lowest of any current Porsche model. The Turbo S model has a slightly higher drag coeff...

节点 ID: d40676ea-00fa-4d45-8149-c22129115d44
相似度 0.4863016369562377
文本: The Porsche Taycan is a battery electric luxury sports sedan and shooting brake car produced by German automobile manufacturer Porsche. The concept version of the Taycan named the Porsche Mission E...

节点 ID: 6e989ddd-581d-4391-8f51-66f627b5baae
相似度 0.4680214751437455
文本: === Porsche Mission E Cross Turismo === The Porsche Mission E Cross Turismo previewed the Taycan Cross Turismo, and was presented at the 2018 Geneva Motor Show. The design language of the Mission E...

节点 ID: 17e0044d-c6a2-4387-b481-08003706333b
相似度 0.4333677171313131
文本: == Specifications ==

=== Chassis === The Taycan's body is mainly steel and aluminium joined by different bonding techniques. The body's B pillars, side roof frame and seat cross member are made f...

节点 ID: 237d74ca-ad9e-413b-a897-295fc657318d
相似度 0.4244636106388237
文本: === Range and charging === The EPA lists the Taycan 4S' range at 203 mi (327 km) with a consumption of 69 mpg‑e (49 kWh/100 mi). However, the car's range depends on how it is driven and what drivin...

节点 ID: 0ce4a54a-1660-45f2-b84c-024be65434ed
相似度 0.39880809672817724
文本: === e-tron GT (2021) ===

The e-tron GT is a four-door all-electric grand tourer. The concept model was showcased at the 2018 Los Angeles Auto Show. It is built on the same platform as the Porsche ...

测试多模态查询引擎¶

我们将使用上述 MultiModalVectorStoreIndex 创建一个 QueryEngine。

In [ ]

from llama_index.core.query_engine import (
    CustomQueryEngine,
    SimpleMultiModalQueryEngine,
)
from llama_index.core.retrievers import BaseRetriever
from llama_index.core.schema import (
    ImageNode,
    NodeWithScore,
    MetadataMode,
    TextNode,
)
from llama_index.core.prompts import PromptTemplate
from llama_index.core.base.response.schema import Response
from llama_index.core.indices.query.schema import QueryBundle
from typing import Optional

QA_PROMPT_TMPL = """\
Below we give parsed text and images as context.

Use both the parsed text and images to answer the question. 

---------------------
{context_str}
---------------------
Given the context information and not prior knowledge, answer the query. Explain whether you got the answer
from the text or image, and if there's discrepancies, and your reasoning for the final answer.

Query: {query_str}
Answer: """

QA_PROMPT = PromptTemplate(QA_PROMPT_TMPL)


class MultimodalQueryEngine(CustomQueryEngine):
    """Custom multimodal Query Engine.

    Takes in a retriever to retrieve a set of document nodes.
    Also takes in a prompt template and multimodal model.

    """

    qa_prompt: PromptTemplate
    retriever: BaseRetriever
    multi_modal_llm: OpenAIMultiModal

    def __init__(
        self, qa_prompt: Optional[PromptTemplate] = None, **kwargs
    ) -> None:
        """Initialize."""
        super().__init__(qa_prompt=qa_prompt or QA_PROMPT, **kwargs)

    def custom_query(self, query_str: str):
        # retrieve text nodes
        nodes = self.retriever.retrieve(query_str)
        image_nodes = [n for n in nodes if isinstance(n.node, ImageNode)]
        text_nodes = [n for n in nodes if isinstance(n.node, TextNode)]

        # Make QueryBundle
        query_bundle = QueryBundle(query_str)

        # reranking text nodes
        reranked_text_nodes = cohere_rerank.postprocess_nodes(
            text_nodes, query_bundle
        )

        # reranking image nodes
        reranked_image_nodes = colpali_reranker.postprocess_nodes(
            image_nodes, query_bundle
        )

        # create context string from text nodes, dump into the prompt
        context_str = "\n\n".join(
            [
                r.get_content(metadata_mode=MetadataMode.LLM)
                for r in reranked_text_nodes
            ]
        )
        fmt_prompt = self.qa_prompt.format(
            context_str=context_str, query_str=query_str
        )

        # synthesize an answer from formatted text and images
        llm_response = self.multi_modal_llm.complete(
            prompt=fmt_prompt,
            image_documents=[n.node for n in reranked_image_nodes],
        )
        return Response(
            response=str(llm_response),
            source_nodes=nodes,
            metadata={
                "text_nodes": reranked_text_nodes,
                "image_nodes": reranked_image_nodes,
            },
        )

        return response

from llama_index.core.query_engine import ( CustomQueryEngine, SimpleMultiModalQueryEngine, ) from llama_index.core.retrievers import BaseRetriever from llama_index.core.schema import ( ImageNode, NodeWithScore, MetadataMode, TextNode, ) from llama_index.core.prompts import PromptTemplate from llama_index.core.base.response.schema import Response from llama_index.core.indices.query.schema import QueryBundle from typing import Optional QA_PROMPT_TMPL = """\ 下面我们提供解析后的文本和图片作为上下文。请同时使用解析后的文本和图片来回答问题。 --------------------- {context_str} --------------------- 根据上下文信息，而不是先验知识，回答查询。说明你是从文本还是图片中获得答案的，以及是否存在差异，并给出你的最终答案的推理。 查询：{query_str} 答案： """ QA_PROMPT = PromptTemplate(QA_PROMPT_TMPL) class MultimodalQueryEngine(CustomQueryEngine): """自定义多模态查询引擎。接收一个检索器以检索一组文档节点。还接收一个提示模板和多模态模型。 """ qa_prompt: PromptTemplate retriever: BaseRetriever multi_modal_llm: OpenAIMultiModal def __init__( self, qa_prompt: Optional[PromptTemplate] = None, **kwargs ) -> None: """初始化。""" super().__init__(qa_prompt=qa_prompt or QA_PROMPT, **kwargs) def custom_query(self, query_str: str): # 检索文本节点 nodes = self.retriever.retrieve(query_str) image_nodes = [n for n in nodes if isinstance(n.node, ImageNode)] text_nodes = [n for n in nodes if isinstance(n.node, TextNode)] # 创建 QueryBundle query_bundle = QueryBundle(query_str) # 重排文本节点 reranked_text_nodes = cohere_rerank.postprocess_nodes( text_nodes, query_bundle ) # 重排图片节点 reranked_image_nodes = colpali_reranker.postprocess_nodes( image_nodes, query_bundle ) # 从文本节点创建上下文字符串，放入提示词 context_str = "\n\n".join( [ r.get_content(metadata_mode=MetadataMode.LLM) for r in reranked_text_nodes ] ) fmt_prompt = self.qa_prompt.format( context_str=context_str, query_str=query_str ) # 从格式化文本和图片合成答案 llm_response = self.multi_modal_llm.complete( prompt=fmt_prompt, image_documents=[n.node for n in reranked_image_nodes], ) return Response( response=str(llm_response), source_nodes=nodes, metadata={ "text_nodes": reranked_text_nodes, "image_nodes": reranked_image_nodes, }, ) return response

In [ ]

query_engine = MultimodalQueryEngine(
    retriever=retriever_engine, multi_modal_llm=gpt_4o
)

query_engine = MultimodalQueryEngine( retriever=retriever_engine, multi_modal_llm=gpt_4o )

In [ ]

query = "Which models of Porsche are discussed here?"
response = query_engine.query(query)

query = "Which models of Porsche are discussed here?" response = query_engine.query(query)

You are passing both `text` and `images` to `PaliGemmaProcessor`. The processor expects special image tokens in the text, as many tokens as there are images per each text. It is recommended to add `<image>` tokens in the very beginning of your text and `<bos>` token after that. For this call, we will infer how many images each text has and add special tokens.

In [ ]

print(str(response))

print(str(response))

The models of Porsche discussed in the text are:

1. **Taycan Turbo**
2. **Taycan Turbo S**
3. **Taycan 4S**
4. **Taycan Cross Turismo**
5. **Taycan Sport Turismo**
6. **Base RWD model**
7. **GTS variants**

Additionally, the concept models mentioned are:

1. **Porsche Mission E**
2. **Porsche Mission E Cross Turismo**

The information is derived from the parsed text, which provides details about these models and their features. The images support the text by showing the design and concept of the Porsche Taycan and Mission E, but the specific model names and details are from the text.

检查来源¶

In [ ]

from llama_index.core.response.notebook_utils import display_source_node

for text_node in response.metadata["text_nodes"]:
    display_source_node(text_node, source_length=200)
plot_images(
    [n.metadata["file_path"] for n in response.metadata["image_nodes"]]
)

from llama_index.core.response.notebook_utils import display_source_node for text_node in response.metadata["text_nodes"]: display_source_node(text_node, source_length=200) plot_images( [n.metadata["file_path"] for n in response.metadata["image_nodes"]] )

节点 ID: e37183fe-8b2b-4bc5-b856-0a2b0221b011
相似度 0.9579909
文本: === Aerodynamics === The Taycan Turbo has a drag coefficient of Cd=0.22, which the manufacturer claims is the lowest of any current Porsche model. The Turbo S model has a slightly higher drag coeff...

节点 ID: d40676ea-00fa-4d45-8149-c22129115d44
相似度 0.8467045
文本: The Porsche Taycan is a battery electric luxury sports sedan and shooting brake car produced by German automobile manufacturer Porsche. The concept version of the Taycan named the Porsche Mission E...

节点 ID: 6e989ddd-581d-4391-8f51-66f627b5baae
相似度 0.70212173
文本: === Porsche Mission E Cross Turismo === The Porsche Mission E Cross Turismo previewed the Taycan Cross Turismo, and was presented at the 2018 Geneva Motor Show. The design language of the Mission E...