How rasbon99/factcheckerai works
FOX AI is a news-claim fact-checking application: users submit a claim through the dashboard, the controller forwards it to the backend, and the backend orchestrates claim preprocessing, evidence discovery and filtering, GraphRAG analysis, and reporting. Its knowledge graph supports evidence retrieval and visual explanations; a separate evaluation branch benchmarks GraphRAG against documented baselines and datasets. The excerpts confirm controller-to-backend calls and internal preprocessing/RAG responsibilities; backend wiring and several dashboard/reporting details rely on the README and module layout, so uncertain links are omitted.
Main entry points: backend.py Pipeline backend · controller.py API controller · query_engine.py Evidence query · rag_pipeline.py GraphRAG orchestrator
Components 19
User Interface
- Dashboard
Dashboard/dashboard.py - API controller
controller.py - Conversation database
Database/sqldb.py - Interactive results
Dashboard/dashboard.py
Fact-check Pipeline
- Pipeline backend
backend.py - Claim and evidence prep
Preprocessor/preprocessing_pipeline.py - Search and source filter
WebScraper/scraper.py - Domain credibility data
Datasets/iffy_index.csv
Knowledge and Models
- GraphRAG orchestrator
GraphRAG/rag_pipeline.py - Graph loading and views
GraphRAG/graph_manager.py - Evidence query
GraphRAG/query_engine.py - Knowledge graph store
Neo4j/neo4j_api.py - Local model service
Ollama/ollama_api.py
Evaluation
- Benchmark runners
Evaluation/Runners - Evaluation datasets
Evaluation/Utils/dataset_manager.py - Experiment metrics
Evaluation/Utils/experiment_tracker.py - Effectiveness and efficiency
Evaluation/Analysis
Other components
- Claim submitter
- Groq inference
Connections 25
- Claim submitter → Dashboard: submits claim
- Dashboard → API controller: sends request
- API controller → Pipeline backend: forwards request
- API controller → Pipeline backend: requests history
- API controller → Pipeline backend: requests cleanup
- Pipeline backend → Claim and evidence prep: preprocesses claim
- Pipeline backend → Search and source filter: retrieves sources
- Search and source filter → Domain credibility data: checks domains
- Search and source filter → Groq inference: checks relevance
- Pipeline backend → GraphRAG orchestrator: runs evidence analysis
- GraphRAG orchestrator → Graph loading and views: loads evidence
- GraphRAG orchestrator → Evidence query: queries evidence
- Graph loading and views → Knowledge graph store: writes graph
- Evidence query → Knowledge graph store: retrieves evidence
- Evidence query → Local model service: embeds query
- Evidence query → Groq inference: generates verdict
- Claim and evidence prep → Groq inference: summarizes and extracts
- Pipeline backend → Conversation database: persists conversations
- Pipeline backend → Interactive results: returns results
- Interactive results → Claim submitter: shows cited report
- Graph loading and views → Interactive results: provides graph views
- Benchmark runners → Evaluation datasets: loads claims
- Benchmark runners → Experiment metrics: records runs
- Effectiveness and efficiency → Experiment metrics: analyzes metrics
- Benchmark runners → GraphRAG orchestrator: benchmarks GraphRAG