| layout | default |
|---|---|
| title | 🏥 anamnesis - Accurate clinical insights from patient data |
| description | Ingest FHIR resources into a vector store to power clinical reasoning chat pipelines using Redpanda, Qdrant, and Rust. |
Anamnesis connects clinical data to your workflows. This application processes health records and provides answers based on factual patient history. It uses vector search and language models to prevent false information, commonly known as hallucinations, by citing actual sources within your files. This provides clinicians and researchers with a reliable way to review electronic health records and clinical notes efficiently.
- Windows 10 or Windows 11
- 8 GB RAM (16 GB recommended)
- 500 MB disk space
- Reliable internet connection
- Visit the official releases page to access the software.
- Select the latest version listed under the Assets section.
- Download the file ending in
.exe. - Run the downloaded installer.
- Follow the prompts on the screen to finalize the setup.
- Launch the application from your desktop or Start menu.
The application runs locally on your machine. Upon opening it, you will see a main dashboard. Before the system can provide clear answers, you must link your clinical data sources.
- Locate the Settings menu in the top right corner.
- Select the Data Sources tab.
- Choose the directory where your patient notes or medical records reside.
- The system automatically detects FHIR-formatted data and prepares it for indexing.
- Wait for the status indicator to show Complete.
Once the data indexes, the search bar becomes active. You can ask questions about specific patient histories or clinical conditions.
- Type your question naturally.
- The system searches the connected records.
- Results appear in the center pane.
- Click any highlighted citation to open the source document.
- The system verifies every answer against the underlying records.
This application prioritizes patient confidentiality. It processes most data on your local hardware. When the system requires external intelligence for complex reasoning, it sends only the necessary fragments to the language model. No raw medical files leave your computer without your explicit configuration. The software handles all data through encrypted channels to maintain medical compliance.
You can adjust how the application interprets your clinical data under the Configuration tab:
- Indexing Priority: Balance speed against depth during the data ingestion phase.
- Model Selection: Switch between different reasoning engines based on your preference for factual strictness.
- Cache Management: Clear or store vector indices based on your local disk capacity.
- Update Frequency: Schedule automatic checks for new patient records in your designated folders.
If the application fails to start:
- Check your internet connection.
- Ensure your antivirus does not block the application.
- Verify that you have administrative rights on your machine.
If the search returns no results:
- Confirm that the folder path in Settings points to valid files.
- Check if your files match the required FHIR data format.
- Click the Re-index button in the dashboard to force a refresh of the database.
Anamnesis integrates several components to maintain a high standard of clinical accuracy. It uses a Rust-based backend for speed and memory safety. The Axum framework handles requests efficiently, while Redpanda Connect manages the flow of clinical data streams. The system maps these data points into a Qdrant vector database, which allows for fast, accurate searching. By combining these tools, the software bridges the gap between raw medical records and actionable clinical intelligence.
Keywords: axum, chat, clinical-decision-support, embeddings, kafka, kafka-connect, llm, openai, qdrant, qdrant-vector-database, rag, redpanda, redpanda-connect, rust, rust-api, vector-database, vector-search