Begin your agentic commerce protocol implementation with a killer advantage. Enrich product data with BookWyrm to stand out from your competitors by giving the agent way more information about your products from your existing materials. Use PDF extraction for product collateral with the structured output summary to get the data in a format of your choosing.
BookWyrm supports structured output using custom Pydantic models, allowing you to extract specific information in a consistent format. This is particularly powerful for literary analysis, research, data extraction tasks, and in this example, data enrichment.
The example below takes a product specification PDF and uses it to enrich data according to the OpenAI agentic commerce protocol.
In order to get to this stage, we have used the Extract from PDF and phrasal analysis endpoints to get the data AI-ready. For the sake of brevity, we’ve omitted this code, but you can see examples of both endpoints in our BookWyrm Client documentation.
import json
from typing import Optional, List
from pydantic import BaseModel, Field, HttpUrl
# Pydantic model for OpenAI's Agentic Commerce specification
class EnrichedProductData(BaseModel):
id: str = Field(
None,
description="Merchant product ID (unique, stable identifier)"
)
gtin: Optional[str] = Field(
None,
description="Universal product identifier (GTIN, UPC, ISBN)"
)
mpn: Optional[str] = Field(
None,
description="Manufacturer part number"
)
title: str = Field(
None,
description="Descriptive and clear product title"
)
description: str = Field(
None,
description="Full product description with key features and benefits"
)We created an API server using BookWyrm's API which is available to clone. The PDF Structured Output Server takes a PDF along with a Pydantic model and returns a structured output.
GitHub RepositoryThe agentic commerce protocol (ACP) is an open standard, co-developed by OpenAI and Stripe, designed to enable secure, programmatic commerce flows between buyers, AI agents (like ChatGPT), and businesses. It provides the necessary framework for e-commerce to seamlessly integrate with the new era of conversational and agent-led shopping.
This protocol ensures that an AI agent can act on behalf of a user, finding products, sharing information, and initiating a secure, fast checkout, without the user ever having to leave the AI interface.
Read more here.
Commerce is and always will be competitive. To succeed, your products need to be found. Perhaps the success of tools like ChatGPT is reflective of the failures of Google and the SEO game. Users are tired of the old and are increasingly turning to AI to search and buy products, hence the introduction of ACP.
The problem most online retailers have is the lack of product data held within their e-commerce platforms. LLMs require content to be able to accurately pair a search and a recommendation. Having a three-line product description, cost, and a specification leaves a lot of gray area for an LLM to fill.
Businesses are not short of product data, however. Brochures, sales presentations, and a range of other collateral are all readily available now. The challenge is combining your merchant data with this goldmine of information
BookWyrm allows you to programmatically enrich your merchant data with product information, user problems, customer reviews, product benefits, and whatever collateral you possess. This data is imperative when it comes to helping OpenAI provide a service that meets the needs of shoppers.
AI is the curator, advisor and transaction facilitator. Ensuring it has the data to accurately do its job will provide your business with commerce superpowers. No SEO games, just the product information that you already have.
BookWyrm is a Python SDK and API that seamlessly slots into your pipeline and can be set up in hours without any pre-processing pain. Learn more about BookWyrm.
File System/Source
PDFs, Docs, Brochures
The classify endpoint intelligently routes the document. Output includes classification metadata.
Documents are split into semantically balanced chunks using the phrasal endpoint.
Pydantic Model
BookWyrm supports structured output using custom Pydantic models, allowing you to extract specific information in a consistent format.
Using a Pydantic model, the structured summary endpoint provides a structured output to your specification.
Deploy to OpenAI Agent
Verifiable RAG Answer
In the world of agentic commerce, the AI agent acts as a curator, advisor, and transaction facilitator. To secure a sale, your product must be highly discoverable and compelling to the AI agent itself.
AI agents prioritize products with complete and rich data to ensure the best customer match. Missing or poor data makes your product 'invisible' to the AI's complex search algorithms.
Complete data, including collateral like PDFs and video links, eliminates buyer friction, reduces cart abandonment, and provides the user's agent with all the necessary details to finalize the purchase confidently.
Detailed product information and official documentation (PDF manuals, spec sheets) increase perceived quality and brand trust, directly influencing the user's agent toward an acceptance during the secure checkout phase.
BookWyrm turns unstructured collateral, product sheets, brochures, PDFs, and specs, into structured, AI-ready data that drives discoverability, accuracy, and trust across your sales and ACP stack.
Read more about GenAI for business in our blog.