Clinical Guidelines Assistant
A retrieval-augmented assistant that answers clinicians' questions from approved treatment guidelines, citing the exact source passage.
- Industry
- Healthcare
- Category
- AI Solutions, Healthcare
- Type
- Concept project
This is a concept project created to demonstrate YARWELL's approach. It does not describe a client engagement, and the interface shows illustrative data only.
Project details
Clinical guidelines are long, frequently revised and spread across many documents. Health workers need quick answers at the point of care, but a general-purpose chatbot cannot be trusted to give accurate, source-backed guidance.
The concept pairs a curated document library with retrieval-augmented generation. Guidelines are divided into sections, indexed with semantic embeddings and retrieved at question time. The model answers only from the retrieved passages and shows its citations; questions the library cannot answer are declined instead of guessed.
- Semantic search across guideline documents
- Answers with passage-level citations
- Document versioning that retires outdated guidance
- Feedback capture for continuous evaluation
- Role-based access and query audit logs
- 1InterfaceNext.js web app and embeddable chat widget
- 2AI servicesFastAPI orchestration, embedding and LLM providers
- 3KnowledgePostgreSQL with vector search and document versions
- 4GovernanceSSO, audit logging and an evaluation test set
Have a similar challenge?
Tell us about the problem you are solving. We will share how we would approach it and what a first phase could look like.