A real-time, end-to-end system for Search and Rescue operations that suggests context-aware follow-up questions and extracts actionable clues during interviews with missing persons' contacts.
In Search and Rescue (SAR) operations, time pressure and inexperience can lead to missed opportunities during interviews with a missing person's friends and family. This system leverages large language models (LLMs), agentic design patterns, and integration with the IntelliSAR platform to assist interviewers in surfacing more complete and relevant information. It compiles key insights into a structured clue log, ready for human review, refinement, and dissemination to the rest of the team.
Ultimate Goal: Accelerate clue discovery and reduce the likelihood of critical information being overlooked in time-sensitive SAR missions.
For docker compose, copy .env.example to .env and fill in all the required environment variables. Then expose expose nginx's port 80 and bring up the system:
docker compose upCopy the .env.example to .env, add the required environment variables.
Run these commands to install dependencies and setup the database:
cd backend
uv sync
uv run alembic upgrade headcd frontend
pnpm installIn the root of the repo run:
pre-commit installcd backend
uv run ./src/main.pycd frontend
pnpm devThe frontend will be available at https://localhost:5173 and the backend WebRTC server runs on the configured port.
See an interactive diagram of the architecture:
docker run -it --rm -p 8080:8080 -v docs:/usr/local/structurizr structurizr/lite
To run the LLM eval tests, in the backend/ dir run:
uv run deepeval test run -m "llm" .\src\interview_helper\ai_analysis\eval
To autogenerate a migration with alembic, in the backend/ dir run:
uv run alembic revision --autogenerate -m "<message>"