Add Level-3 academic integrity roadmap and FastAPI scoring scaffold

README.md

@@ -1,10 +1,63 @@
 # TrustCode Platform 🚀
 
-AI-assisted Trust-aware CI/CD platform that improves software reliability
-using automated code audits, testing pipelines, and trust metrics.
-
-## Features
-- AI-based code review (LangChain)
-- Automated CI/CD with GitHub Actions
-- Trust Loop for commit verification
-- Test execution & failure analysis
+TrustCode is an **Academic Integrity & Research Authenticity Platform** that helps faculty, students, and administrators evaluate originality, argument quality, AI overuse risk, and research novelty with explainable evidence.
+
+## What is new (Next-Level Scope)
+
+This repository now includes a practical implementation roadmap for the advanced Level-3 capabilities:
+
+- Concept-level plagiarism detection (idea theft)
+- Argument & logic-flow scoring
+- AI-vs-human probability analysis
+- Student stylometric fingerprinting
+- Group contribution verification
+- Research novelty scoring
+- Viva support question generation
+- Learning outcome mapping (CO/PO/Bloom)
+- Batch/college analytics dashboards
+- Policy-driven integrity rules
+- Student improvement timeline
+- Evidence-backed decision reports
+- Secure dataset builder for research
+- Multi-language support
+- Composite Academic Integrity Index
+
+## Top 10 must-build first
+
+1. Policy-Driven Integrity Rules Engine
+2. Evidence-Backed Decision Reports
+3. Concept-Level Plagiarism Detector
+4. AI vs Human Writing Probability
+5. Stylometric Fingerprint + Drift Detection
+6. Novelty Score Engine
+7. Contribution Verification for Group Projects
+8. Learning Outcome Mapping
+9. Batch/College Analytics
+10. Academic Integrity Index
+
+Why this order:
+- Delivers governance and trust early.
+- Reduces false-positive risk by requiring explainability.
+- Builds reusable scoring infrastructure before advanced UX layers.
+
+## Initial API scaffold
+
+A lightweight FastAPI scaffold is included at `backend/main.py` with:
+
+- health endpoint
+- policy registration endpoint
+- integrity index scoring endpoint
+- explainable evidence payload in the response
+
+Run locally:
+
+```bash
+pip install -r backend/requirements.txt
+uvicorn backend.main:app --reload --port 8000
+```
+
+## Documentation
+
+Detailed architecture, data model, phased rollout, and IEEE-friendly research framing:
+
+- `docs/academic-integrity-next-level.md`

backend/main.py

@@ -0,0 +1,91 @@
+from typing import List, Optional
+
+from fastapi import FastAPI
+from pydantic import BaseModel, Field
+
+app = FastAPI(title="TrustCode Academic Integrity API", version="0.1.0")
+
+
+class IntegrityPolicy(BaseModel):
+    name: str
+    max_similarity: float = Field(ge=0.0, le=100.0)
+    ai_strictness: float = Field(ge=0.0, le=1.0)
+    originality_weight: float = Field(ge=0.0, le=1.0)
+    citation_weight: float = Field(ge=0.0, le=1.0)
+    novelty_weight: float = Field(ge=0.0, le=1.0)
+
+
+class IntegritySignals(BaseModel):
+    originality_score: float = Field(ge=0.0, le=100.0)
+    citation_quality_score: float = Field(ge=0.0, le=100.0)
+    logic_flow_score: float = Field(ge=0.0, le=100.0)
+    novelty_score: float = Field(ge=0.0, le=100.0)
+    writing_quality_score: float = Field(ge=0.0, le=100.0)
+    ai_overuse_risk: float = Field(ge=0.0, le=100.0)
+    idea_similarity_risk: float = Field(ge=0.0, le=100.0)
+
+
+class EvidenceItem(BaseModel):
+    section: str
+    reason: str
+    confidence: float = Field(ge=0.0, le=1.0)
+    source: Optional[str] = None
+
+
+class IntegrityScoreResponse(BaseModel):
+    integrity_index: float
+    verdict: str
+    evidence: List[EvidenceItem]
+
+
+POLICIES: List[IntegrityPolicy] = []
+
+
+@app.get("/health")
+def health() -> dict:
+    return {"status": "ok"}
+
+
+@app.post("/policies", response_model=IntegrityPolicy)
+def create_policy(policy: IntegrityPolicy) -> IntegrityPolicy:
+    POLICIES.append(policy)
+    return policy
+
+
+@app.get("/policies", response_model=List[IntegrityPolicy])
+def list_policies() -> List[IntegrityPolicy]:
+    return POLICIES
+
+
+@app.post("/score", response_model=IntegrityScoreResponse)
+def score_submission(signals: IntegritySignals) -> IntegrityScoreResponse:
+    index = (
+        0.25 * signals.originality_score
+        + 0.15 * signals.citation_quality_score
+        + 0.15 * signals.logic_flow_score
+        + 0.15 * signals.novelty_score
+        + 0.10 * signals.writing_quality_score
+        - 0.10 * signals.ai_overuse_risk
+        - 0.10 * signals.idea_similarity_risk
+    )
+
+    verdict = "pass" if index >= 60 else "review"
+    evidence = [
+        EvidenceItem(
+            section="Overall",
+            reason="Composite index derived from originality, citation quality, logic flow, novelty, and risk penalties.",
+            confidence=0.86,
+        ),
+        EvidenceItem(
+            section="AI Usage",
+            reason="AI-overuse risk reduces the final index under institutional policy.",
+            confidence=0.78,
+            source="policy:default-weight-profile",
+        ),
+    ]
+
+    return IntegrityScoreResponse(
+        integrity_index=round(index, 2),
+        verdict=verdict,
+        evidence=evidence,
+    )

backend/requirements.txt

@@ -0,0 +1,3 @@
+fastapi==0.115.0
+uvicorn==0.30.6
+pydantic==2.9.2

docs/academic-integrity-next-level.md

@@ -0,0 +1,93 @@
+# Academic Integrity Platform – Level-3 Blueprint
+
+## 1) Product goal
+
+Build an enterprise-ready academic integrity system that moves beyond string matching and supports **idea-level authenticity**, **explainable decisioning**, and **institution-scale analytics**.
+
+## 2) Core modules
+
+| Layer | Responsibilities |
+|---|---|
+| AI Engine | embeddings, stylometry, argument mining, multilingual NLP |
+| Integrity Engine | weighted scoring, policy evaluation, penalties, thresholds |
+| Explainability | source snippets, confidence, rationale trace |
+| Analytics | student, class, department, institution trend intelligence |
+| Governance | policy authoring, audit logs, ethics + consent handling |
+| Research | novelty analysis and anonymized dataset export |
+
+## 3) Prioritized implementation roadmap
+
+### Phase A: Trust foundation (Weeks 1–3)
+- Policy-Driven Rules Engine
+- Evidence-Backed Decision Reports
+- Academic Integrity Index (first version)
+
+### Phase B: Detection intelligence (Weeks 4–8)
+- Concept-Level Plagiarism Detector
+- AI-vs-Human Writing Probability
+- Stylometric fingerprint + anomaly detection
+- Contribution verification for team submissions
+
+### Phase C: Research quality (Weeks 9–12)
+- Argument & Logic Flow Analyzer
+- Novelty Score against historical/internal/public sources
+- Viva-support auto question generator
+- Learning Outcome mapping (Bloom/CO/PO)
+
+### Phase D: Enterprise scale (Weeks 13–16)
+- Batch/college analytics dashboards
+- Improvement timeline per student
+- Multi-language support + cross-language similarity
+- Secure dataset builder with ethics controls
+
+## 4) Academic Integrity Index (proposed v1)
+
+```text
+IntegrityIndex =
+(0.25 * OriginalityScore)
++ (0.15 * CitationQualityScore)
++ (0.15 * LogicFlowScore)
++ (0.15 * NoveltyScore)
++ (0.10 * WritingQualityScore)
+- (0.10 * AIOveruseRisk)
+- (0.10 * IdeaSimilarityRisk)
+```
+
+Policy engine can override weights by institution.
+
+## 5) Data entities (minimal)
+
+- StudentProfile
+- Submission
+- EvidenceItem
+- IntegrityPolicy
+- AnalysisResult
+- ContributionSplit
+- OutcomeMapping
+- AuditEvent
+
+## 6) Explainability contract
+
+Every flagged result must include:
+- evidence snippet
+- matched source
+- score/confidence
+- model signal summary
+- policy/rule that fired
+
+This keeps outcomes auditable and defensible during viva/disciplinary reviews.
+
+## 7) Research/IEEE readiness
+
+- Anonymized dataset pipeline with consent metadata
+- Experiment tracking for model versions
+- Baseline comparison (classical plagiarism vs idea-level model)
+- Fairness checks across language/background segments
+
+## 8) Suggested KPIs
+
+- False positive rate (<5% target)
+- Manual review agreement (>85%)
+- Policy override frequency
+- Semester-over-semester integrity improvement
+- Time saved in faculty review process