curl -X POST https://frnklnwrld-me.hf.space/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "What is the meaning of patience in Islam?",
    "user_id": "Abdullah123",
    "category": "Religious Self"
  }'

const response = await fetch('https://frnklnwrld-me.hf.space/chat', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    message: 'What is sabr?',
    user_id: 'user_abc123',
    category: 'Religious Self'
  })
});

const data = await response.json();
console.log(data.voice_answer);   // Short spoken response
console.log(data.middle_section); // Detailed notes
console.log(data.follow_up);      // Next conversation prompt

import requests

res = requests.post(
    "https://frnklnwrld-me.hf.space/chat",
    json={
        "message": "I feel sad today",
        "user_id": "testuser",
        "category": "Emotional Self"
    }
)
data = res.json()
print(data["voice_answer"])
print(data["model_used"])   # Which Llama model responded

{
  "status": "insight_only",
  "voice_answer": "SubhanAllah, I hear the weight in your words. Remember, after hardship comes ease (Quran 94:5). You are not alone — Allah is closer to you than your jugular vein.",
  "middle_section": "Sadness is a human experience acknowledged in the Quran. The Prophet ﷺ himself experienced grief deeply. Allowing yourself to feel is not weakness — it is honesty before Allah.",
  "middle_label": "Detailed Notes",
  "follow_up": "Would you like to explore what specifically is weighing on your heart today?",
  "references": "Quran 94:5 — Indeed, with hardship comes ease.",
  "model_used": "Llama-3-8B-Instruct",
  "next_action_guidance": {
    "type": "general_chat",
    "message": "Assalamu alaikum. How fares your heart today?",
    "suggested_delay_hours": 6,
    "islamic_reminder": "Quranic Principle: Do not despair of Allah's mercy (39:53)."
  }
}

HF_TOKEN=hf_xxxxxxxxxxxxxxxxxx          # Hugging Face token with Inference API access
SUPABASE_URL=https://xxxx.supabase.co   # Your Supabase project URL
SUPABASE_SERVICE_KEY=sbp_xxxxxxxxxxxx   # Service role key (NOT anon key)

{
  "options": ["Always", "Often", "Sometimes", "Rarely", "Never"],
  "scores":  [   5,       4,        3,          2,        1   ]
}

{
  "message": "آج میں بہت اداس ہوں",
  "user_id": "user123"
}

1. meta-llama/Meta-Llama-3-8B-Instruct    → Best quality
2. meta-llama/Llama-3.1-8B-Instruct       → Latest stable  
3. meta-llama/Llama-3.2-1B-Instruct       → Fastest / lightest

for model in MODELS:           # Try each model in priority order
    for attempt in range(2):   # Up to 2 retries per model
        if status == 429:
            time.sleep(2 ** attempt)   # 1s, then 2s backoff
            continue
        if status == 200:
            return answer, model_name  # Success — stop here
        break                  # Other error — try next model

import 'dart:convert';
import 'package:http/http.dart' as http;

class AbdullahBotService {
  static const String baseUrl = 'https://frnklnwrld-me.hf.space';

  Future sendMessage({
    required String message,
    required String userId,
    String category = 'General',
    Map? previousSummary,
    List? answers,
  }) async {
    final response = await http.post(
      Uri.parse('$baseUrl/chat'),
      headers: {'Content-Type': 'application/json'},
      body: jsonEncode({
        'message': message,
        'user_id': userId,
        'category': category,
        if (previousSummary != null) 'previous_summary': previousSummary,
        if (answers != null) 'answers': answers,
      }),
    );

    if (response.statusCode == 200) {
      return jsonDecode(response.body);
    } else if (response.statusCode == 503) {
      throw Exception('Service waking up — please retry in a moment');
    } else {
      throw Exception('API error: ${response.statusCode}');
    }
  }

  Future getJourney(String userId, {String category = 'General'}) async {
    final response = await http.get(
      Uri.parse('$baseUrl/journey/$userId?category=$category'),
    );
    return jsonDecode(response.body);
  }
}

// In your chat widget:
final bot = AbdullahBotService();
final data = await bot.sendMessage(
  message: userInput,
  userId: currentUser.id,
  category: selectedCategory,
);

// Render the structured response:
Text(data['voice_answer'])          // Primary message bubble
Text(data['middle_section'] ?? '')  // Expandable detail card
Text(data['follow_up'] ?? '')       // Suggested reply chip
Text(data['references'] ?? '')      // Quranic reference footer

async function chatWithRetry(message, userId, retries = 3) {
  for (let i = 0; i < retries; i++) {
    try {
      const res = await fetch('/chat', {
        method: 'POST',
        headers: { 'Content-Type': 'application/json' },
        body: JSON.stringify({ message, user_id: userId })
      });

      const text = await res.text();
      let data;
      try { data = JSON.parse(text); }
      catch { throw new Error('Server error: ' + text.slice(0, 100)); }

      if (res.status === 503) {
        // Supabase cold start — wait and retry
        await new Promise(r => setTimeout(r, 5000));
        continue;
      }
      if (!res.ok) throw new Error(data.detail || 'Request failed');
      return data;

    } catch (err) {
      if (i === retries - 1) throw err;
      await new Promise(r => setTimeout(r, 2000 * (i + 1)));
    }
  }
}

Title: Flutter Developer — Islamic AI Companion App (Abdullah Bot)

We're building a Flutter mobile app powered by a live FastAPI backend
(hosted on Hugging Face Spaces). The backend is complete and production-ready.
We need an experienced Flutter developer to build the mobile UI.

=== BACKEND OVERVIEW ===
• Live API: https://frnklnwrld-me.hf.space
• Swagger docs: https://frnklnwrld-me.hf.space/docs
• AI Model: Meta Llama-3 (3-tier fallback)
• Database: Supabase (Postgres)
• Features: Tone detection (13 modes), Journey tracking, MCQ system,
  Multilingual (English/Urdu/Arabic), Quranic references

=== YOUR RESPONSIBILITIES ===
1. Build Flutter chat UI consuming POST /chat endpoint
2. Render structured responses: voice_answer + middle_section + follow_up
3. Implement MCQ journey screen (render current_mcqs array as interactive form)
4. Journey progress screen using GET /journey/{user_id}
5. Category picker using GET /categories
6. Handle cold-start 503 errors with graceful retry UX
7. State management (Provider or Riverpod preferred)
8. Local caching of cumulative_summary

=== API RESPONSE STRUCTURE ===
Every /chat response returns:
- voice_answer (string): Primary message — display as main chat bubble
- middle_section (string?): Detail card — collapsible
- follow_up (string?): Suggested next message chip
- current_mcqs (array?): MCQ questions to render as form
- model_used (string): Which AI model responded
- next_action_guidance (object): UI hints

=== REQUIREMENTS ===
• 3+ years Flutter experience
• REST API integration (http or dio package)
• Clean architecture preferred
• Portfolio of chat/messaging UIs required
• Islamic/Arabic UI experience is a strong plus

=== DELIVERABLES ===
• Complete Flutter project (clean code, well-commented)
• APK for testing
• README with setup instructions

Please share 2-3 examples of chat apps you've built.
Budget: [your budget] | Timeline: 2-3 weeks

Title: FastAPI / Python Developer — Enhance Islamic AI Bot Backend

We have a production FastAPI backend for an Islamic AI companion app.
The core is working. We need help with improvements and scaling.

=== CURRENT STACK ===
• FastAPI + Python 3.10
• Hugging Face Spaces (hosting)
• Supabase (Postgres + REST)
• Meta Llama-3 via HF Router API (3-model fallback)
• Live at: https://frnklnwrld-me.hf.space

=== TASKS ===
1. Add async support — convert sync DB calls to async (supabase-py async client)
2. Add proper rate limiting middleware (slowapi)
3. Implement background tasks for non-critical operations (chat logging)
4. Add Redis caching layer for frequently fetched data (categories, user journeys)
5. Improve error responses — standardise error schema across all endpoints
6. Add /admin endpoints for analytics (total users, avg scores by category)
7. Write pytest suite — target 80% coverage on business logic
8. Add Pydantic v2 migration (currently v1 syntax)

=== CODEBASE HIGHLIGHTS ===
• Tone detection: 13 emotional tones, multilingual (EN/UR/AR)
• Model fallback: 3-tier Llama-3 chain with exponential backoff
• Journey tracking: MCQ system with score smoothing
• Structured responses: voice_answer + middle_section + follow_up format

=== REQUIREMENTS ===
• Strong FastAPI & SQLAlchemy/PostgREST experience
• Supabase or PostgreSQL background
• Experience with async Python (asyncio, httpx)
• Understanding of LLM API integration patterns

Please share your GitHub or sample FastAPI project.
Budget: [your budget] | Timeline: 1-2 weeks

=== ABDULLAH BOT API — TECHNICAL SPEC ===

LIVE ENDPOINTS
  Base: https://frnklnwrld-me.hf.space
  Docs: https://frnklnwrld-me.hf.space/docs
  Health: https://frnklnwrld-me.hf.space/health

KEY ENDPOINTS
  POST /chat          — Main AI conversation (requires: message, user_id)
  GET  /journey/:id   — User progress stats
  GET  /categories    — Available MCQ categories (11 total)
  POST /reset-journey — Reset user journey in a category

RESPONSE FORMAT (all /chat responses)
  status:            "insight_only" | "asking_questions" | "session_complete"
  voice_answer:      string  — short, speakable primary response
  middle_section:    string? — extended detail or notes
  follow_up:         string? — suggested next message
  current_mcqs:      [{question, options[5]}]? — MCQ form data
  model_used:        "Llama-3-8B-Instruct" | "fallback" | ...
  next_action_guidance: {type, message, suggested_delay_hours, islamic_reminder}

MCQ ANSWER SUBMISSION
  Send in: answers: [{question_num: 1, answer: "Often", score: 4}]
  Scores: Always=5, Often=4, Sometimes=3, Rarely=2, Never=1

ERROR CODES
  400 — Missing required fields
  503 — Supabase cold start (retry after 5s)
  500 — Server error (check HF Space logs)

TECH STACK
  Python 3.10, FastAPI, Uvicorn
  supabase-py, httpx, requests
  pydantic v1, python-dotenv

SECRETS REQUIRED (HF Space Settings → Secrets)
  HF_TOKEN, SUPABASE_URL, SUPABASE_SERVICE_KEY