How joaosegurilho/scientific-claim-verifier works
SciVerifier turns scientific claims into evidence-based verdicts by finding papers, extracting propositions, retrieving relevant evidence, and interpreting support or contradiction. Its other user-facing paths ingest local documents, verify claim batches, expose knowledge-base browsing and agent verification through a web app, and run benchmark datasets. The graph shows the documented CLI and sampled pipeline ownership; several component internals and web routes were not sampled, so uncertain wiring is omitted.
Main entry points: verification_pipeline.py Verification pipeline · cli.py CLI · extraction_pipeline.py Extraction pipeline · main.py Web application
Components 23
User entry points
- CLI
scverifier/cli.py - Web application
scverifier/webapp/main.py - Bulk verifier
scverifier/core/bulk_verifier.py
Evidence acquisition
- Literature search
scverifier/core/knowledge/literature_search.py - Paper APIs
scverifier/api/api.py - Extraction pipeline
scverifier/pipelines/extraction_pipeline.py - Document loader
scverifier/data/file_loader.py - Paper processor
scverifier/data/local_paper_processor.py - Proposition extractor
scverifier/core/extraction/proposition_extractor.py
Verification and retrieval
- Verification pipeline
scverifier/pipelines/verification_pipeline.py - Proposition retrieval
scverifier/core/retrieval/retrieval_system.py - Evidence reranker
scverifier/core/retrieval/reranker.py - Claim verifier
scverifier/core/verification/claim_verifier.py - Confidence interpreter
scverifier/core/verification/confidence_interpreter.py - Autonomous agent
scverifier/core/agents/autonomous_agent.py - Verdict and evidence
Knowledge base
- Knowledge base
scverifier/core/knowledge/knowledge_base.py - Paper database
scverifier/data/database.py
Additional capabilities
- Benchmark pipeline
scverifier/pipelines/benchmark_pipeline.py - Benchmark runner
scverifier/core/benchmarking/run_benchmark.py - Timeline visualization
scverifier/core/visualization/timeline_generator.py
Other components
- User
- Literature services
Connections 27
- User → CLI: invokes
- User → Web application: uses
- CLI → Verification pipeline: starts verification
- CLI → Extraction pipeline: starts extraction
- CLI → Bulk verifier: starts batch verification
- CLI → Benchmark pipeline: starts benchmark
- Verification pipeline → Literature search: searches papers
- Literature search → Paper APIs: queries
- Paper APIs → Literature services: requests papers
- Verification pipeline → Proposition extractor: extracts propositions
- Verification pipeline → Knowledge base: loads and updates
- Verification pipeline → Proposition retrieval: retrieves evidence
- Verification pipeline → Evidence reranker: ranks evidence
- Verification pipeline → Claim verifier: evaluates claim
- Claim verifier → Confidence interpreter: interprets confidence
- Verification pipeline → Verdict and evidence: returns verdict
- Extraction pipeline → Document loader: loads documents
- Extraction pipeline → Paper processor: builds paper
- Extraction pipeline → Knowledge base: adds and saves paper
- Document loader → Extraction pipeline: returns document text
- Web application → Knowledge base: browses papers
- Web application → Autonomous agent: streams agent verification
- Autonomous agent → Verdict and evidence: produces verdict
- Knowledge base → Paper database: stores paper records
- Benchmark pipeline → Benchmark runner: runs datasets
- Benchmark runner → Verification pipeline: uses verification methods
- Web application → Timeline visualization: generates timelines