CLAIM GRAPH: Autonomous Evidence & Contradiction Intelligence Engine

Bhargav Ram Ambati — Fri, 16 Oct 2026 10:36:17 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.

🚀 What I Built: CLAIM GRAPH

CLAIM GRAPH is an autonomous evidence & contradiction intelligence engine that deconstructs unstructured architectural specifications, research papers, and policy briefs into testable atomic claims. It dynamically maps dependencies, reconciles cross-document temporal revisions, and pinpoints invalidated conclusions whose supporting evidence has fallen.

🌐 Live Access & Code Repository

GitHub Repository: bhargavambati09-crypto/CLAIM-GRAPH
Live Public URL: claim-graph.netlify.app (Tunnel password / server public IP if prompted: 47.29.69.147)
Local Host URL: http://localhost:8000/

bhargavambati09-crypto / CLAIM-GRAPH

CLAIM GRAPH: Evidence & Contradiction Intelligence

An autonomous knowledge graph engine that deconstructs unstructured technical specifications and research documents into testable atomic propositions, tracks temporal contradictions across document revisions, and flags orphaned conclusions whose supporting evidence has fallen.

🌟 Key Features

Atomic Claim Extraction: Parses unstructured document chunks into discrete, testable propositions classified as FACT, SPECIFICATION, CONCLUSION, or ASSUMPTION.
Temporal Contradiction Tracking: Detects direct conflicts between specifications across document revisions and chronologically marks older claims as SUPERSEDED and same-date conflicts as CONTRADICTED.
Orphan & Invalidation Analysis: Constructs directed dependency edges (Premise) -[:DERIVES]-> (Conclusion). If a premise is superseded or contradicted, the dependent conclusion is automatically invalidated as ORPHANED.
Interactive Knowledge Graph Canvas: Physics-simulated interactive visualization with zoom, pan, node inspection, and colored relationship edges.
Dual Inference Engine
- Local AI Mode: Powered by gemma3:1b running locally via Ollama…

View on GitHub

📸 Live Web Application Screenshots

1. Main Dashboard & Scenario Analysis

Figure 1: CLAIM GRAPH running locally with dark glassmorphic UI, real-time metric cards, scenario switcher, and execution controls.

2. Interactive Knowledge Graph Canvas

Figure 2: Real-time physics-simulated node-link property graph. Blue nodes represent documents, green nodes are active propositions, amber/red nodes represent superseded or contradicted claims, and pink nodes represent orphaned conclusions.

3. Contradiction Matrix & Temporal Reconciliation

Figure 3: Side-by-side proposition comparison across document versions (ARCH-V1 vs ARCH-V2), showing conflict reason and temporal resolution.

4. Orphaned Conclusions Watchlist

Figure 4: Automated deduction tracking identifying conclusions whose prerequisite premises were superseded or refuted.

5. Extracted Atomic Claims Directory

Figure 5: Proposition directory with filter chips by status (Active, Superseded, Contradicted, Orphaned) and search by topic.

🛠️ Problem & Market Fit

In complex engineering projects, research literature, and legal compliance briefs, teams constantly revise documents. However, downstream conclusions silently rely on outdated, refuted, or superseded premises. Traditional search and standard LLM chat interfaces fail to maintain strict logical dependency graphs across document versions, leading to costly design flaws and contradictory assumptions.

CLAIM GRAPH solves this by:

Extracting Atomic Propositions: Breaking documents into precise verifiable claims.
NetworkX Directed Property Graph Mapping: Modeling (Premise) -[:DERIVES]-> (Conclusion) relationships.
Temporal Supersession Tracking: Automatically flagging when a newer document replaces an older premise.
Orphaned Conclusion Detection: Catching downstream conclusions whose supporting pillars have been invalidated.

⚙️ Technical Architecture & Open Innovation

Dual-Engine Inference:
- Local Ollama (gemma3:1b): Deep semantic extraction when local AI is online.
- Deterministic NLP Fallback Engine: Sub-0.1s pattern-matching grammar resolver ensuring robust public access even without active GPU/LLM containers.
Robust Rectifications:
- Windows IPv6 localhost Resolution Fix: Replaced localhost with explicit 127.0.0.1 to prevent multi-minute DNS/IPv6 timeouts on Windows.
- CORS & Public Binding: Bound server to 0.0.0.0:8000 with full CORS preflight support for tunneling tools (localtunnel).
- Case-Insensitive Normalization: Fixed strict keyword lookup bugs (e.g. "NVMe" vs "nvme").
Frontend: Glassmorphic dashboard built with vanilla CSS/JS and HTML5 Canvas physics simulation.

💡 How I Built It & Agent Session

Built with Python, FastAPI, NetworkX, Pydantic, and Ollama (gemma3:1b), orchestrated with Antigravity AI coding assistant and verified via automated E2E test runs.

DEV Community: Bhargav Ram Ambati