scrapy-vectors 0.2.1

Vector embeddings generation and storage for Scrapy

scrapy-vectors

Vector embeddings generation and storage for Scrapy spiders.

Features

• Embeddings Pipeline: Generate vector embeddings using LiteLLM (supports OpenAI, Cohere, and other providers)
• S3 Vectors Storage: Store embeddings in AWS S3 Vectors service

Installation

pip install scrapy-vectors

Quick Start

In your scrapy_settings.py:

ITEM_PIPELINES = {
    # Outputs as jsonlines in Pinecone format, which s3-vectors can use
    "scrapy_vectors.EmbeddingsLiteLLMPipeline": 300,
}
EXTENSIONS = {
    "scrapy.extensions.feedexport.FeedExporter": None,  # Disable standard
    "scrapy_vectors.S3VectorsFeedExporter": 300,        # Use custom
}
FEED_STORAGES = {
    "s3-vectors": "scrapy_vectors.S3VectorsFeedStorage",
}
FEEDS = {
    "s3-vectors://vectors-bucket/vectors-index": {
        "format": "jsonlines",
        "batch_item_count": 100,
    }
}

# LiteLLM will route for you
LITELLM_API_KEY = "your_provider_api_key"          # (e.g. OpenAI API Key)
LITELLM_EMBEDDING_MODEL = "text-embedding-3-small" # This is default when unspecified

AWS_REGION_NAME = "us-east-1"
AWS_ACCESS_KEY_ID = "access_key_id"
AWS_SECRET_ACCESS_KEY = "access_key"

In your scraper:

import scrapy


class MySpider(scrapy.Spider):
    name = "example"
    start_urls = ["https://example.com"]

    # Must yield with: id, page_content, and metadata
    def parse(self, response):
        yield {
            "id": response.url,
            "page_content": response.css("article::text").get(),
            "metadata": {
                "title": response.css("h1::text").get(),
                "url": response.url,
            }
        }

Configuration

Embeddings Pipeline Settings

• LITELLM_API_KEY: API key for your embedding provider (required)
• LITELLM_EMBEDDING_MODEL: Model to use (default: OpenAI's text-embedding-3-small)

S3 Vectors Storage Settings

• AWS_REGION_NAME: AWS region (required)
• AWS_ACCESS_KEY_ID: AWS access key
• AWS_SECRET_ACCESS_KEY: AWS secret key