adapt-gemini

name	adapt-gemini
description	How to adapt the existing geminiService.ts by ADDING new methods for MoSPI competency assessment and MCQ generation. All existing methods are preserved untouched.

Adapt Gemini Service — Add MoSPI Competency Methods

When to Use

After running strip-and-fork. This adds new methods to geminiService.ts without modifying any existing ones.

---

Step 1: Read the Existing Service

Read src/services/geminiService.ts completely. Understand these existing methods (ALL are preserved):

• createChat() — Creates a streaming chat with optional PDF grounding
• sendMessageStream() — Sends messages with comprehension modes
• generateQuizQuestion() — Generates structured quiz JSON
• generateSocraticFAQ() — Generates Q&A pairs from PDF
• evaluateAnswer() — Grades student answers
• generateDeepDiveGuide() — Creates study guides from PDF
• generateNotebookLMPodcast() — Creates podcast scripts

IMPORTANT: Do NOT modify or rename any existing method. Only ADD new methods.

Step 2: Add New System Instruction for StatSarthi Chat Sessions

Add a new method createStatSarthiChat() alongside the existing createChat():

public async createStatSarthiChat(history: Message[] = [], pdfData?: string): Promise<void> {
  // Same logic as createChat() but with StatSarthi system instruction:
  const systemInstruction = `You are StatSarthi AI, an expert in India's Official Statistical System,
  competency-based capacity building, and the iGOT Karmayogi framework. You help MoSPI officials
  identify skill gaps, recommend training, and generate assessments.

  Guidelines:
  1. Use formal, professional language appropriate for government officials.
  2. Reference MoSPI terminology: NSSTA, SQAF, FRAC, Official Statistical System.
  3. When generating MCQs, tag each question with Bloom's Taxonomy level.
  4. Support 4 competency domains with 28+ areas:
     - Statistical: Survey Design, Sampling, National Accounts, Price Statistics, Labour Statistics,
       Agricultural Statistics, Industrial Statistics, SDG Indicators, Metadata Standards, Data Quality (SQAF)
     - Technical: Python, R, SQL, Stata, SPSS, SAS, GIS, Data Visualization, AI/ML, Cloud Computing, APIs, Open Data
     - Digital Governance: Cybersecurity, Data Privacy, Digital Signatures, Government Cloud, Digital Public Infrastructure
     - Behavioural & Managerial: Leadership, Communication, Project Management, Ethics, Decision Making, Change Management
  5. When grading assessments, map scores to competency levels (1-5 scale).`;
  // ... rest of chat initialization
}

Step 3: Add New Methods

generateDiagnosticQuestions()

Generates diagnostic assessment questions for a given competency area.

public async generateDiagnosticQuestions(
  competencyArea: string,
  subCompetencies: string[],
  difficulty: 'beginner' | 'intermediate' | 'advanced',
  count: number = 5
): Promise<DiagnosticQuestion[]>

Output schema (use Gemini's responseSchema):

interface DiagnosticQuestion {
  id: string;
  question: string;
  options: string[];           // 4 options
  correctIndex: number;        // 0-3
  competencyArea: string;
  subCompetency: string;
  bloomsLevel: 'Remember' | 'Understand' | 'Apply' | 'Analyze' | 'Evaluate' | 'Create';
  difficulty: string;
  explanation: string;
}

generateMCQsFromDocument()

Takes PDF content and generates MCQs from it. This is the hero feature.

public async generateMCQsFromDocument(
  pdfData: string,
  options: {
    count: number;
    difficulty: 'easy' | 'medium' | 'hard' | 'mixed';
    bloomsLevels: string[];
    competencyArea?: string;
  }
): Promise<GeneratedMCQ[]>

Output schema:

interface GeneratedMCQ {
  id: string;
  question: string;
  options: string[];
  correctIndex: number;
  explanation: string;
  bloomsLevel: string;
  difficulty: string;
  sourceReference: string;     // Which section of the document
}

evaluateCompetency()

Takes answers and produces a competency score.

public async evaluateCompetency(
  questions: DiagnosticQuestion[],
  userAnswers: number[],
  competencyArea: string
): Promise<CompetencyScore>

Output:

interface CompetencyScore {
  competencyArea: string;
  score: number;               // 0-100
  level: 1 | 2 | 3 | 4 | 5;
  strengths: string[];
  weaknesses: string[];
  recommendations: string[];
}

Step 4: Summary of Changes

Method	Action	Notes
createChat()	✅ KEEP	Original CS tutor chat — untouched
sendMessageStream()	✅ KEEP	Streaming with modes — untouched
generateQuizQuestion()	✅ KEEP	Original quiz gen — untouched
generateSocraticFAQ()	✅ KEEP	FAQ generation — untouched
evaluateAnswer()	✅ KEEP	Answer grading — untouched
generateDeepDiveGuide()	✅ KEEP	Study guide gen — untouched
generateNotebookLMPodcast()	✅ KEEP	Podcast gen — untouched
createStatSarthiChat()	➕ ADD	New StatSarthi-specific chat
generateDiagnosticQuestions()	➕ ADD	Competency assessment questions
generateMCQsFromDocument()	➕ ADD	PDF → MCQ generation
evaluateCompetency()	➕ ADD	Competency scoring

Step 5: Add New Types

Create src/types/statsarthi.ts with all new interfaces:

• DiagnosticQuestion
• GeneratedMCQ
• CompetencyScore
• CompetencyArea
• RoleProfile
• GapReport
• IGOTCourse
• LearningPathway
• PathwayStep

Step 6: Verify

Test each new method by calling it from a temporary page or browser console. Ensure Gemini returns valid JSON matching the schemas.