AI • DATA • DOCUMENTATION
Documentation verification & evidence platform
Nexus is a documentation verification platform designed to determine whether software documentation is actually supported by evidence. Modern software repositories change constantly, but their documentation can quickly become outdated or inaccurate. Nexus analyzes documentation, extracts individual claims, retrieves relevant evidence from the underlying repository, and evaluates each claim against that evidence.
Timeline
Built and iterated throughout 2026, from product concept and architecture through backend integration, verification pipeline, testing, and dashboard development.
Background
Nexus was built around a simple question: “Is your documentation actually true?” The platform creates a structured verification pipeline that turns documentation into individual claims and evaluates them using repository evidence rather than relying solely on AI-generated answers.
This category details the step-by-step approach taken during the project, including research, planning, design, development, testing, and optimization phases.
Research & Planning
Defined the documentation-verification problem, researched how documentation becomes unreliable, and designed the claim-to-evidence verification workflow.
System Architecture
Designed the end-to-end pipeline connecting documentation ingestion, claim extraction, evidence retrieval, evidence analysis, verification, and reporting.
Backend & Analysis
Built the verification pipeline using Python and FastAPI, with PostgreSQL and SQLAlchemy supporting the application and analysis workflow.
Testing & Validation
Built automated tests around the backend and verification pipeline to validate the system's behavior across different cases.
Nexus transforms documentation verification from a manual review process into a structured, evidence-driven workflow. The system processes documentation through a multi-stage pipeline: Documentation → Claim Extraction → Evidence Retrieval → Evidence Analysis → Truth Verification → Report
Claim Extraction
Breaks documentation into individual, verifiable claims instead of treating an entire document as one block of information.
Evidence Retrieval
Searches the relevant repository and documentation sources for evidence that can support or challenge each claim.
Truth Verification
Analyzes the relationship between claims and retrieved evidence to classify whether the documentation is supported, uncertain, or contradicted.
Verification Reports
Presents the results in a dashboard so users can quickly understand what is verified, what needs review, and where documentation may be inaccurate.
End-to-End Pipeline
Complete documentation-to-verification workflow integrated and functional.
Automated Claim Extraction
Documentation converted into individual claims ready for verification.
Evidence-Based Verification
Claims evaluated against evidence from the connected repository.
Verification Dashboard
Results presented through a structured dashboard for quick review.
28 / 28 Tests Passing
Backend and analysis pipeline validated with a complete passing test suite.
73 Claims Analyzed
73 documentation claims processed in the GradScope regression test.
36 Verified
Claims supported by available evidence.
37 Uncertain
Claims requiring additional evidence or review.
0 Contradicted
No claims were conclusively contradicted.
75 / 100 Truth Score
Overall verification score for the regression analysis.