Chat with your Device
Proof of concept for an AI assistant for querying user manuals of medical devices with source references.
About the Project
Technicians on-site at medical and operational equipment often face the challenge of quickly finding relevant information from extensive user manuals. Manually searching through documentation costs time and is error-prone.
The Concept
The concept of this proof of concept: A chat interface through which technicians simply enter the device type and receive relevant information about the device – without manual document search. In addition to internal documentation, a web search is offered to also include external information sources.
Technical Implementation
The technical implementation is based on LangChain as an agent framework. A dataset of approximately 6,000 documents – primarily user manuals – was embedded and stored in a Milvus Vector Database. The agent orchestrates various tools: Semantic Search on the embeddings, web search for supplementary information and database queries.
Source Referencing
A central feature is the consistent source referencing: Every answer contains references that should be verified – for example, links to the user manual including page number or links to external websites. This ensures that technicians can verify the information before taking action.
Highlights
- Chat interface for intuitive device queries
- Semantic search over 6,000+ embedded documents
- LangChain agent with tool integration
- Consistent source references with links and page numbers
Challenges
Embedding and indexing large document volumes. Precise source references in semantic search.
Results
Functional proof of concept for chat-based document access. Validation of the concept for technician daily work.
Interested?
Are you interested in a similar project or want to learn more? Get in touch – we look forward to hearing from you!
Get in touch