How aamna-khan-git/amr-sentinel works
AMR Sentinel answers natural-language One Health surveillance questions: its dashboard or REST endpoint sends a question through intent parsing, structured-data and literature retrieval, and grounded narrative generation with severity labels. Separate ingestion paths populate the surveillance database and literature index. The README documents dashboard charts and recommendation capabilities; sampled excerpts do not establish every UI or subsystem connection, so uncertain wiring is omitted.
Main entry points: orchestrator.py Query orchestrator · api_server.py Dashboard and API · pdf_ingest.py PDF ingestion · literature_store.py Index and search
Components 19
Query and answers
- Query orchestrator
orchestrator.py - Surveillance queries
agents/data_agent.py - Literature search
agents/literature_agent.py - Narrative and severity
agents/narrative_agent.py - Database queries
db_setup.py - Index and search
literature_store.py
Data ingestion
- PDF ingestion
pdf_ingest.py - EFSA ingestion
fetch_efsa.py - Abstract retrieval
fetch_abstracts_ncbi.py
Evidence stores
- Surveillance database
amr_sentinel.db - Literature index
chroma_store - Abstract records
data/abstracts.jsonl
User interfaces
- Dashboard and API
api_server.py - Charts and metrics
api_server.py
Other components
- User
- LLM provider
- ECDC reports
- EFSA data
- NCBI PubMed
Connections 21
- User → Dashboard and API: asks
- Dashboard and API → Query orchestrator: forwards question
- Query orchestrator → LLM provider: parses intent
- Query orchestrator → Surveillance queries: queries data
- Query orchestrator → Literature search: searches literature
- Query orchestrator → Narrative and severity: generates answer
- Narrative and severity → LLM provider: generates narrative
- Surveillance queries → Database queries: dispatches query
- Database queries → Surveillance database: reads
- Literature search → Index and search: searches
- Index and search → Literature index: searches index
- ECDC reports → PDF ingestion: provides reports
- PDF ingestion → Surveillance database: writes surveillance data
- PDF ingestion → Literature index: writes text chunks
- EFSA data → EFSA ingestion: provides data
- EFSA ingestion → Surveillance database: writes surveillance data
- NCBI PubMed → Abstract retrieval: provides abstracts
- Abstract retrieval → Abstract records: writes records
- Abstract records → Index and search: supplies documents
- Query orchestrator → Dashboard and API: returns answer
- Dashboard and API → Charts and metrics: renders results