本文详细介绍了 7-Eleven 如何利用 Databricks Agent Bricks 开发'技术人员维护助手'（TMA），以简化维护技术人员的知识获取。面对分散且格式多样的设备文档挑战，7-Eleven 旨在创建一个 AI 驱动的助手，实现精确文档检索、基于图像的零件识别以及与 Microsoft Teams 的无缝集成。该解决方案利用 Unity Catalog Volumes 进行安全文档存储，使用带有 BAAI bge-large-en-v1.5 嵌入的 Databricks Vector Search 实现快速信息检索，并通过 Teams Bot API 层协调 Databricks Model Serving。这一举措显著提高了技术人员获取关键维护信息的效率，通过集中化和智能化呈现文档来增强门店运营的一致性。
Empowering Technicians Across Every Store
7‑Eleven’s maintenance technicians keep stores running smoothly by servicing a wide range of equipment — from food service appliances and refrigeration units to fuel dispensers and Slurpee machines. Each repair relies on the technician's knowledge and immediate access to supporting documents, such as service manuals, wiring diagrams, and annotated images.
Creating a Unified and Faster Way for Technicians to Find Equipment Information
Over time, equipment documentation has evolved to include multiple formats, spread across various locations. This makes it harder for Technicians to locate the information they need quickly. Moreover, when encountering unfamiliar equipment, parts, etc., Technicians would often rely on chat or email to get support from their peers.
As such, an opportunity to streamline how information is accessed, shared, etc. was identified; ultimately resulting in more consistent support for store operations.
Building the Technician’s Maintenance Assistant (TMA)
To tackle these challenges, 7‑Eleven envisioned an AI‑powered assistant that could:
- Retrieve precise answers from maintenance documents.
- Identify equipment parts from images and suggest related materials.
- Integrate seamlessly within Microsoft Teams.
Partnering with Databricks, 7-Eleven developed the Technician’s Maintenance Assistant (TMA), an intelligent solution that integrates document retrieval, vision models, and collaboration into a streamlined workflow.
Document Storage and Indexing
All relevant maintenance documents were uploaded to a Unity Catalog Volume, which manages permissions for non-tabular data, such as text and images, across cloud storage.
Using Databricks Vector Search, the development team implemented Delta Sync with Embeddings Compute. They generated vector embeddings using the BAAI bge-large-en-v1.5 model, and served them through a Vector Search endpoint for high-speed, low-latency retrieval.
Microsoft Teams Integration
Technicians access TMA directly through Microsoft Teams. A Teams Bot routes each query through an API layer that orchestrates calls to Databricks Model Serving. The assistant provides contextual answers, matches documentation links, and suggests relevant parts directly in the chat window.
