app.py

"""
CyberSec-API: REST API Gateway for Cybersecurity AI Models
===========================================================
Provides unified API access to three specialized cybersecurity models:
- ISO27001-Expert (1.5B) - ISO 27001 compliance guidance
- RGPD-Expert (1.5B) - GDPR/RGPD data protection
- CyberSec-Assistant (3B) - General cybersecurity operations
"""

import os
import json
import time
import gradio as gr
from huggingface_hub import InferenceClient

# ---------------------------------------------------------------------------
# Configuration
# ---------------------------------------------------------------------------

MODELS = {
    "ISO27001-Expert": {
        "id": "AYI-NEDJIMI/ISO27001-Expert-1.5B",
        "description": "Specialized in ISO 27001 standards, ISMS implementation, risk assessment, and compliance auditing.",
        "parameters": "1.5B",
        "specialty": "ISO 27001 Compliance",
    },
    "RGPD-Expert": {
        "id": "AYI-NEDJIMI/RGPD-Expert-1.5B",
        "description": "Specialized in GDPR/RGPD regulations, data protection, privacy impact assessments, and DPO guidance.",
        "parameters": "1.5B",
        "specialty": "GDPR/RGPD Data Protection",
    },
    "CyberSec-Assistant": {
        "id": "AYI-NEDJIMI/CyberSec-Assistant-3B",
        "description": "General-purpose cybersecurity assistant for incident response, threat analysis, vulnerability management, and security operations.",
        "parameters": "3B",
        "specialty": "General Cybersecurity",
    },
}

MODEL_NAMES = list(MODELS.keys())

# System prompts per model
SYSTEM_PROMPTS = {
    "ISO27001-Expert": (
        "You are ISO27001-Expert, an AI assistant specialized in ISO 27001 information security management systems. "
        "Provide accurate, professional guidance on ISMS implementation, risk assessment, control selection, "
        "audit preparation, and compliance requirements. Reference specific ISO 27001 clauses and Annex A controls when relevant."
    ),
    "RGPD-Expert": (
        "You are RGPD-Expert, an AI assistant specialized in GDPR (General Data Protection Regulation) / RGPD. "
        "Provide accurate guidance on data protection principles, lawful bases for processing, data subject rights, "
        "DPIA procedures, breach notification requirements, and DPO responsibilities. Reference specific GDPR articles when relevant."
    ),
    "CyberSec-Assistant": (
        "You are CyberSec-Assistant, a general-purpose cybersecurity AI assistant. "
        "Provide expert guidance on incident response, threat intelligence, vulnerability management, "
        "penetration testing, SOC operations, network security, and security architecture. "
        "Be practical and actionable in your recommendations."
    ),
}

# Inference client
HF_TOKEN = os.getenv("HF_TOKEN", "")
client = InferenceClient(token=HF_TOKEN) if HF_TOKEN else None

# Rate limiting state
_request_log: list[float] = []
RATE_LIMIT_WINDOW = 60  # seconds
RATE_LIMIT_MAX = 30     # requests per window


# ---------------------------------------------------------------------------
# Core functions
# ---------------------------------------------------------------------------

def _check_rate_limit() -> bool:
    """Return True if within rate limit."""
    now = time.time()
    _request_log[:] = [t for t in _request_log if now - t < RATE_LIMIT_WINDOW]
    if len(_request_log) >= RATE_LIMIT_MAX:
        return False
    _request_log.append(now)
    return True


def _query_model(message: str, model_name: str, max_tokens: int = 512) -> str:
    """Send a prompt to the specified model via the HF Inference API."""
    if not client:
        return "[Error] HF_TOKEN is not configured. The API is unavailable."

    if model_name not in MODELS:
        return f"[Error] Unknown model '{model_name}'. Available: {', '.join(MODEL_NAMES)}"

    if not _check_rate_limit():
        return "[Error] Rate limit exceeded. Please wait before sending more requests."

    model_id = MODELS[model_name]["id"]
    system_prompt = SYSTEM_PROMPTS[model_name]

    try:
        messages = [
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": message},
        ]
        response = client.chat_completion(
            model=model_id,
            messages=messages,
            max_tokens=max_tokens,
            temperature=0.7,
        )
        return response.choices[0].message.content

    except Exception as e:
        error_str = str(e)
        # Fallback to text_generation if chat_completion is not supported
        if "not supported" in error_str.lower() or "chat" in error_str.lower():
            try:
                prompt = f"### System:\n{system_prompt}\n\n### User:\n{message}\n\n### Assistant:\n"
                response = client.text_generation(
                    prompt=prompt,
                    model=model_id,
                    max_new_tokens=max_tokens,
                    temperature=0.7,
                    do_sample=True,
                )
                return response
            except Exception as fallback_err:
                return f"[Error] Model query failed: {fallback_err}"
        return f"[Error] Model query failed: {e}"


# ---------------------------------------------------------------------------
# API endpoint functions (exposed via Gradio)
# ---------------------------------------------------------------------------

def chat(message: str, model_name: str) -> str:
    """Send a message to a specific cybersecurity model and get a response.

    Args:
        message: The question or prompt to send to the model.
        model_name: One of 'ISO27001-Expert', 'RGPD-Expert', or 'CyberSec-Assistant'.

    Returns:
        The model's response text.
    """
    if not message or not message.strip():
        return "[Error] Message cannot be empty."
    return _query_model(message.strip(), model_name)


def compare(message: str) -> str:
    """Send a message to all 3 models and compare their responses side by side.

    Args:
        message: The question or prompt to send to all models.

    Returns:
        JSON string with responses from each model.
    """
    if not message or not message.strip():
        return json.dumps({"error": "Message cannot be empty."}, indent=2)

    results = {}
    for name in MODEL_NAMES:
        results[name] = {
            "model_id": MODELS[name]["id"],
            "specialty": MODELS[name]["specialty"],
            "response": _query_model(message.strip(), name),
        }

    return json.dumps(results, indent=2, ensure_ascii=False)


def list_models() -> str:
    """List all available cybersecurity models and their details.

    Returns:
        JSON string with model information.
    """
    model_list = []
    for name, info in MODELS.items():
        model_list.append({
            "name": name,
            "model_id": info["id"],
            "description": info["description"],
            "parameters": info["parameters"],
            "specialty": info["specialty"],
            "endpoint": f"/api/chat with model_name='{name}'",
        })
    return json.dumps({"models": model_list, "count": len(model_list)}, indent=2)


def health_check() -> str:
    """Check the health status of the API and its dependencies.

    Returns:
        JSON string with health status information.
    """
    status = {
        "status": "healthy" if client else "degraded",
        "timestamp": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
        "version": "1.0.0",
        "hf_token_configured": bool(HF_TOKEN),
        "models_available": MODEL_NAMES,
        "rate_limit": {
            "window_seconds": RATE_LIMIT_WINDOW,
            "max_requests": RATE_LIMIT_MAX,
            "current_usage": len([t for t in _request_log if time.time() - t < RATE_LIMIT_WINDOW]),
        },
    }
    return json.dumps(status, indent=2)


# ---------------------------------------------------------------------------
# Tab content builders
# ---------------------------------------------------------------------------

API_DOCS_MD = """
# CyberSec-API Documentation

A REST API gateway providing unified access to three specialized cybersecurity AI models hosted on Hugging Face.

---

## Available Models

| Model | Specialty | Parameters | Model ID |
|-------|-----------|------------|----------|
| **ISO27001-Expert** | ISO 27001 compliance, ISMS, risk assessment | 1.5B | `AYI-NEDJIMI/ISO27001-Expert-1.5B` |
| **RGPD-Expert** | GDPR/RGPD, data protection, privacy | 1.5B | `AYI-NEDJIMI/RGPD-Expert-1.5B` |
| **CyberSec-Assistant** | Incident response, threat analysis, SOC | 3B | `AYI-NEDJIMI/CyberSec-Assistant-3B` |

---

## Endpoints

### POST `/api/chat`
Send a message to a specific cybersecurity model.

**Parameters:**
| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `message` | string | Yes | The question or prompt |
| `model_name` | string | Yes | One of: `ISO27001-Expert`, `RGPD-Expert`, `CyberSec-Assistant` |

**Response:** Plain text response from the model.

---

### POST `/api/compare`
Send the same message to all 3 models and compare their responses.

**Parameters:**
| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `message` | string | Yes | The question or prompt |

**Response:** JSON object with each model's response.

---

### GET `/api/models`
List all available models and their details.

**Parameters:** None

**Response:** JSON object with model information.

---

### GET `/api/health`
Health check endpoint for monitoring.

**Parameters:** None

**Response:** JSON object with API status, version, and rate limit info.

---

## Rate Limits

| Limit | Value |
|-------|-------|
| Requests per minute | 30 |
| Max tokens per request | 512 |
| Concurrent requests | 5 |

---

## Code Examples

### Python (using `gradio_client`)

```python
from gradio_client import Client

# Connect to the API
client = Client("AYI-NEDJIMI/CyberSec-API")

# Chat with a specific model
result = client.predict(
    message="What are the key requirements of ISO 27001 Clause 6?",
    model_name="ISO27001-Expert",
    api_name="/chat"
)
print(result)

# Compare all models
result = client.predict(
    message="How should we handle a data breach?",
    api_name="/compare"
)
print(result)

# List available models
models = client.predict(api_name="/models")
print(models)

# Health check
status = client.predict(api_name="/health")
print(status)
```

### Python (using `requests`)

```python
import requests

SPACE_URL = "https://ayi-nedjimi-cybersec-api.hf.space"

# Chat endpoint
response = requests.post(
    f"{SPACE_URL}/api/chat",
    json={
        "data": [
            "What controls does ISO 27001 Annex A recommend for access management?",
            "ISO27001-Expert"
        ]
    }
)
print(response.json()["data"][0])

# Compare endpoint
response = requests.post(
    f"{SPACE_URL}/api/compare",
    json={
        "data": ["How do you perform a risk assessment?"]
    }
)
print(response.json()["data"][0])
```

### cURL

```bash
# Chat with a model
curl -X POST "https://ayi-nedjimi-cybersec-api.hf.space/api/chat" \\
  -H "Content-Type: application/json" \\
  -d '{"data": ["What is ISO 27001?", "ISO27001-Expert"]}'

# Compare all models
curl -X POST "https://ayi-nedjimi-cybersec-api.hf.space/api/compare" \\
  -H "Content-Type: application/json" \\
  -d '{"data": ["Explain the principle of least privilege"]}'

# List models
curl -X POST "https://ayi-nedjimi-cybersec-api.hf.space/api/models" \\
  -H "Content-Type: application/json" \\
  -d '{"data": []}'

# Health check
curl -X POST "https://ayi-nedjimi-cybersec-api.hf.space/api/health" \\
  -H "Content-Type: application/json" \\
  -d '{"data": []}'
```

### JavaScript

```javascript
import { Client } from "@gradio/client";

const client = await Client.connect("AYI-NEDJIMI/CyberSec-API");

// Chat with a model
const chatResult = await client.predict("/chat", {
    message: "What are GDPR data subject rights?",
    model_name: "RGPD-Expert",
});
console.log(chatResult.data[0]);

// Compare all models
const compareResult = await client.predict("/compare", {
    message: "How to respond to a ransomware attack?",
});
console.log(JSON.parse(compareResult.data[0]));

// List models
const models = await client.predict("/models", {});
console.log(JSON.parse(models.data[0]));
```

---

## Authentication

This API is publicly accessible. No authentication token is required to call the endpoints.
The API uses an internal HF token (configured as a Space secret) to communicate with the
Hugging Face Inference API on your behalf.

---

## Error Handling

All endpoints return error messages in a consistent format:

| Error | Description |
|-------|-------------|
| `[Error] Message cannot be empty.` | The message parameter was empty or missing |
| `[Error] Unknown model '...'` | Invalid model_name provided |
| `[Error] Rate limit exceeded.` | Too many requests -- wait and retry |
| `[Error] Model query failed: ...` | Upstream inference error |
"""

INTEGRATION_GUIDE_MD = """
# Integration Guide

Integrate CyberSec-API into your security infrastructure, automation pipelines, and communication tools.

---

## 1. SIEM Integration

### Splunk Integration

Create a custom Splunk alert action that queries CyberSec-API for incident analysis:

```python
# splunk_cybersec_action.py
# Place in $SPLUNK_HOME/etc/apps/your_app/bin/

import sys
import json
import requests

CYBERSEC_API = "https://ayi-nedjimi-cybersec-api.hf.space"

def analyze_alert(alert_data):
    \"\"\"Send Splunk alert data to CyberSec-Assistant for analysis.\"\"\"
    prompt = f\"\"\"Analyze this security alert and provide:
1. Severity assessment
2. Recommended immediate actions
3. Investigation steps

Alert Data:
{json.dumps(alert_data, indent=2)}
\"\"\"
    response = requests.post(
        f"{CYBERSEC_API}/api/chat",
        json={"data": [prompt, "CyberSec-Assistant"]},
        timeout=60
    )
    return response.json()["data"][0]

if __name__ == "__main__":
    # Read alert payload from Splunk
    alert_payload = json.loads(sys.stdin.read())
    analysis = analyze_alert(alert_payload)
    print(analysis)
```

**Splunk `alert_actions.conf`:**
```ini
[cybersec_analyze]
label = CyberSec AI Analysis
description = Analyze security alerts using CyberSec-API
command = python3 $SPLUNK_HOME/etc/apps/cybersec/bin/splunk_cybersec_action.py
is_custom = 1
```

### Microsoft Sentinel Integration

Use an Azure Logic App or Function to call CyberSec-API from Sentinel playbooks:

```python
# azure_function/cybersec_sentinel/__init__.py
import json
import logging
import requests
import azure.functions as func

CYBERSEC_API = "https://ayi-nedjimi-cybersec-api.hf.space"

def main(req: func.HttpRequest) -> func.HttpResponse:
    \"\"\"Azure Function triggered by Sentinel incident.\"\"\"
    incident = req.get_json()

    prompt = f\"\"\"Analyze this Microsoft Sentinel security incident:
Title: {incident.get('title', 'N/A')}
Severity: {incident.get('severity', 'N/A')}
Description: {incident.get('description', 'N/A')}
Entities: {json.dumps(incident.get('entities', []))}

Provide: severity validation, recommended response actions, and investigation queries.
\"\"\"
    # Check if it is compliance-related
    model = "CyberSec-Assistant"
    title_lower = incident.get("title", "").lower()
    if "gdpr" in title_lower or "data protection" in title_lower:
        model = "RGPD-Expert"
    elif "compliance" in title_lower or "audit" in title_lower:
        model = "ISO27001-Expert"

    response = requests.post(
        f"{CYBERSEC_API}/api/chat",
        json={"data": [prompt, model]},
        timeout=60
    )

    return func.HttpResponse(
        json.dumps({"analysis": response.json()["data"][0], "model_used": model}),
        mimetype="application/json"
    )
```

---

## 2. Chat Bot Integration

### Slack Bot

```python
# slack_cybersec_bot.py
import os
import json
import requests
from slack_bolt import App
from slack_bolt.adapter.socket_mode import SocketModeHandler

CYBERSEC_API = "https://ayi-nedjimi-cybersec-api.hf.space"

app = App(token=os.environ["SLACK_BOT_TOKEN"])

MODEL_MAP = {
    "iso": "ISO27001-Expert",
    "gdpr": "RGPD-Expert",
    "rgpd": "RGPD-Expert",
    "sec": "CyberSec-Assistant",
    "cyber": "CyberSec-Assistant",
}

def detect_model(text):
    \"\"\"Auto-detect the best model based on keywords.\"\"\"
    text_lower = text.lower()
    for keyword, model in MODEL_MAP.items():
        if keyword in text_lower:
            return model
    return "CyberSec-Assistant"  # default

@app.message("!ask")
def handle_ask(message, say):
    \"\"\"Handle '!ask <question>' messages.\"\"\"
    query = message["text"].replace("!ask", "").strip()
    if not query:
        say("Usage: `!ask <your cybersecurity question>`")
        return

    model = detect_model(query)
    say(f"Asking *{model}*... :hourglass:")

    response = requests.post(
        f"{CYBERSEC_API}/api/chat",
        json={"data": [query, model]},
        timeout=60
    )
    answer = response.json()["data"][0]
    say(f"*{model}:*\\n{answer}")

@app.message("!compare")
def handle_compare(message, say):
    \"\"\"Handle '!compare <question>' to get all 3 model responses.\"\"\"
    query = message["text"].replace("!compare", "").strip()
    if not query:
        say("Usage: `!compare <your cybersecurity question>`")
        return

    say("Comparing all 3 models... :hourglass:")
    response = requests.post(
        f"{CYBERSEC_API}/api/compare",
        json={"data": [query]},
        timeout=120
    )
    results = json.loads(response.json()["data"][0])

    for model_name, data in results.items():
        say(f"*{model_name}* ({data['specialty']}):\\n{data['response']}")

if __name__ == "__main__":
    handler = SocketModeHandler(app, os.environ["SLACK_APP_TOKEN"])
    handler.start()
```

### Discord Bot

```python
# discord_cybersec_bot.py
import os
import json
import discord
import requests
from discord.ext import commands

CYBERSEC_API = "https://ayi-nedjimi-cybersec-api.hf.space"

bot = commands.Bot(command_prefix="!", intents=discord.Intents.default())

@bot.command(name="ask")
async def ask(ctx, model: str = "CyberSec-Assistant", *, question: str):
    \"\"\"Ask a cybersecurity question. Usage: !ask [model] <question>\"\"\"
    valid_models = ["ISO27001-Expert", "RGPD-Expert", "CyberSec-Assistant"]
    if model not in valid_models:
        question = f"{model} {question}"
        model = "CyberSec-Assistant"

    await ctx.send(f"Querying **{model}**...")

    response = requests.post(
        f"{CYBERSEC_API}/api/chat",
        json={"data": [question, model]},
        timeout=60
    )
    answer = response.json()["data"][0]

    # Discord has a 2000 char limit
    if len(answer) > 1900:
        for i in range(0, len(answer), 1900):
            await ctx.send(answer[i:i+1900])
    else:
        await ctx.send(f"**{model}:**\\n{answer}")

bot.run(os.environ["DISCORD_TOKEN"])
```

---

## 3. CI/CD Pipeline Integration

### GitHub Actions

```yaml
# .github/workflows/security-review.yml
name: AI Security Review
on:
  pull_request:
    paths:
      - '**.py'
      - '**.js'
      - '**.yml'
      - 'Dockerfile'

jobs:
  security-review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Get changed files
        id: changed
        run: |
          FILES=$(git diff --name-only ${{ github.event.pull_request.base.sha }} HEAD)
          echo "files=$FILES" >> $GITHUB_OUTPUT

      - name: AI Security Review
        run: |
          pip install requests
          python - <<'SCRIPT'
          import requests, os, json

          API = "https://ayi-nedjimi-cybersec-api.hf.space"
          files = "${{ steps.changed.outputs.files }}".split("\\n")

          prompt = f\"\"\"Review these changed files for security vulnerabilities,
          hardcoded secrets, and compliance issues:

          Changed files: {', '.join(files)}

          Provide a security assessment with:
          1. Critical issues found
          2. Recommendations
          3. Compliance notes (ISO 27001 / GDPR if applicable)
          \"\"\"

          resp = requests.post(
              f"{API}/api/compare",
              json={"data": [prompt]},
              timeout=120
          )
          results = json.loads(resp.json()["data"][0])
          for model, data in results.items():
              print(f"\\n{'='*60}")
              print(f"Model: {model} ({data['specialty']})")
              print(f"{'='*60}")
              print(data["response"])
          SCRIPT
```

### GitLab CI

```yaml
# .gitlab-ci.yml
security-ai-scan:
  stage: test
  image: python:3.11-slim
  script:
    - pip install requests
    - |
      python3 -c "
      import requests, json

      API = 'https://ayi-nedjimi-cybersec-api.hf.space'
      resp = requests.post(
          f'{API}/api/chat',
          json={'data': [
              'Review this CI/CD pipeline for security best practices and suggest improvements.',
              'CyberSec-Assistant'
          ]},
          timeout=60
      )
      print(resp.json()['data'][0])
      "
  only:
    changes:
      - .gitlab-ci.yml
      - Dockerfile
      - docker-compose*.yml
```

---

## 4. Python SDK Example

Create a reusable Python SDK wrapper for clean integration:

```python
# cybersec_sdk.py
\"\"\"CyberSec-API Python SDK\"\"\"

import json
from typing import Optional
from gradio_client import Client


class CyberSecAPI:
    \"\"\"Client for the CyberSec-API gateway.\"\"\"

    MODELS = ["ISO27001-Expert", "RGPD-Expert", "CyberSec-Assistant"]

    def __init__(self, space_id: str = "AYI-NEDJIMI/CyberSec-API"):
        self.client = Client(space_id)

    def chat(self, message: str, model: str = "CyberSec-Assistant") -> str:
        \"\"\"Send a question to a specific model.\"\"\"
        if model not in self.MODELS:
            raise ValueError(f"Unknown model '{model}'. Choose from: {self.MODELS}")
        return self.client.predict(
            message=message,
            model_name=model,
            api_name="/chat"
        )

    def compare(self, message: str) -> dict:
        \"\"\"Get responses from all 3 models for comparison.\"\"\"
        result = self.client.predict(message=message, api_name="/compare")
        return json.loads(result)

    def models(self) -> dict:
        \"\"\"List available models.\"\"\"
        result = self.client.predict(api_name="/models")
        return json.loads(result)

    def health(self) -> dict:
        \"\"\"Check API health status.\"\"\"
        result = self.client.predict(api_name="/health")
        return json.loads(result)

    def ask_iso27001(self, question: str) -> str:
        \"\"\"Shortcut to query the ISO 27001 expert.\"\"\"
        return self.chat(question, model="ISO27001-Expert")

    def ask_rgpd(self, question: str) -> str:
        \"\"\"Shortcut to query the RGPD/GDPR expert.\"\"\"
        return self.chat(question, model="RGPD-Expert")

    def ask_cybersec(self, question: str) -> str:
        \"\"\"Shortcut to query the general cybersecurity assistant.\"\"\"
        return self.chat(question, model="CyberSec-Assistant")


# Usage example
if __name__ == "__main__":
    api = CyberSecAPI()

    # Check health
    print("Health:", api.health())

    # Ask a question
    answer = api.ask_iso27001("What are the mandatory documents for ISO 27001 certification?")
    print("Answer:", answer)

    # Compare models
    comparison = api.compare("What is the best approach to incident response?")
    for model, data in comparison.items():
        print(f"\\n{model}: {data['response'][:200]}...")
```

---

## 5. Webhook Integration

For event-driven architectures, set up a webhook relay:

```python
# webhook_relay.py
from flask import Flask, request, jsonify
import requests

app = Flask(__name__)
CYBERSEC_API = "https://ayi-nedjimi-cybersec-api.hf.space"

@app.route("/webhook/security-alert", methods=["POST"])
def security_alert_webhook():
    \"\"\"Receive security alerts and auto-analyze with CyberSec-API.\"\"\"
    alert = request.json
    prompt = f"Analyze this security alert: {json.dumps(alert)}"

    response = requests.post(
        f"{CYBERSEC_API}/api/chat",
        json={"data": [prompt, "CyberSec-Assistant"]},
        timeout=60
    )

    return jsonify({
        "alert_id": alert.get("id"),
        "ai_analysis": response.json()["data"][0]
    })
```
"""

# ---------------------------------------------------------------------------
# CSS
# ---------------------------------------------------------------------------

CUSTOM_CSS = """
.api-docs {
    max-width: 900px;
    margin: 0 auto;
}
.model-card {
    border: 1px solid #374151;
    border-radius: 8px;
    padding: 16px;
    margin: 8px 0;
    background: #1a1a2e;
}
.header-banner {
    background: linear-gradient(135deg, #0f0f23 0%, #1a1a3e 50%, #2d1b4e 100%);
    padding: 24px;
    border-radius: 12px;
    margin-bottom: 16px;
    border: 1px solid #333;
    text-align: center;
}
.status-badge {
    display: inline-block;
    padding: 4px 12px;
    border-radius: 12px;
    font-size: 0.85em;
    font-weight: 600;
}
.status-healthy { background: #064e3b; color: #6ee7b7; }
.status-degraded { background: #78350f; color: #fcd34d; }
footer { display: none !important; }
"""

# ---------------------------------------------------------------------------
# Gradio UI
# ---------------------------------------------------------------------------

with gr.Blocks(
    title="CyberSec-API",
    css=CUSTOM_CSS,
    theme=gr.themes.Base(
        primary_hue="blue",
        secondary_hue="gray",
        neutral_hue="gray",
    ),
) as demo:

    # Header
    gr.HTML("""
    <div class="header-banner">
        <h1 style="margin:0; font-size:2em; color:#60a5fa;">CyberSec-API</h1>
        <p style="margin:4px 0 0; color:#9ca3af; font-size:1.1em;">
            REST API Gateway for Cybersecurity AI Models
        </p>
        <p style="margin:8px 0 0; color:#6b7280; font-size:0.9em;">
            ISO 27001 &bull; GDPR/RGPD &bull; General Cybersecurity
        </p>
    </div>
    """)

    with gr.Tabs():
        # ===== Tab 1: API Documentation =====
        with gr.Tab("API Documentation", id="docs"):
            gr.Markdown(API_DOCS_MD, elem_classes=["api-docs"])

        # ===== Tab 2: Try It =====
        with gr.Tab("Try It", id="try-it"):
            gr.Markdown("## Interactive API Tester")
            gr.Markdown("Select a model, type your cybersecurity question, and get a response.")

            with gr.Row():
                with gr.Column(scale=2):
                    model_selector = gr.Dropdown(
                        choices=MODEL_NAMES,
                        value="CyberSec-Assistant",
                        label="Select Model",
                        info="Choose which cybersecurity expert to query",
                    )
                    user_input = gr.Textbox(
                        label="Your Question",
                        placeholder="e.g., What are the key steps for implementing an ISMS according to ISO 27001?",
                        lines=4,
                    )
                    with gr.Row():
                        submit_btn = gr.Button("Submit", variant="primary", scale=2)
                        clear_btn = gr.Button("Clear", variant="secondary", scale=1)

                with gr.Column(scale=3):
                    response_output = gr.Textbox(
                        label="Model Response",
                        lines=16,
                        interactive=False,
                        show_copy_button=True,
                    )

            gr.Markdown("---")
            gr.Markdown("### Quick Examples")
            gr.Examples(
                examples=[
                    ["What are the mandatory documents required for ISO 27001 certification?", "ISO27001-Expert"],
                    ["Explain the GDPR right to data portability under Article 20.", "RGPD-Expert"],
                    ["How should a SOC team respond to a ransomware incident?", "CyberSec-Assistant"],
                    ["What is the difference between ISO 27001 and ISO 27002?", "ISO27001-Expert"],
                    ["What are the lawful bases for processing personal data under GDPR?", "RGPD-Expert"],
                    ["Explain the MITRE ATT&CK framework and its use in threat hunting.", "CyberSec-Assistant"],
                ],
                inputs=[user_input, model_selector],
                label="Click an example to populate the form",
            )

            # Compare section
            gr.Markdown("---")
            gr.Markdown("### Compare All Models")
            gr.Markdown("Send the same question to all 3 models and see how each expert responds.")
            compare_input = gr.Textbox(
                label="Question for All Models",
                placeholder="e.g., How do you perform a security risk assessment?",
                lines=2,
            )
            compare_btn = gr.Button("Compare All Models", variant="primary")
            compare_output = gr.Textbox(
                label="Comparison Results (JSON)",
                lines=20,
                interactive=False,
                show_copy_button=True,
            )

            # Status section
            gr.Markdown("---")
            gr.Markdown("### API Status")
            with gr.Row():
                models_btn = gr.Button("List Models", variant="secondary")
                health_btn = gr.Button("Health Check", variant="secondary")
            status_output = gr.Textbox(
                label="Status Output",
                lines=10,
                interactive=False,
                show_copy_button=True,
            )

            # Wire up events with api_name for clean API URLs
            submit_btn.click(
                fn=chat,
                inputs=[user_input, model_selector],
                outputs=response_output,
                api_name="chat",
            )

            clear_btn.click(
                fn=lambda: ("", ""),
                inputs=None,
                outputs=[user_input, response_output],
                api_name=False,
            )

            compare_btn.click(
                fn=compare,
                inputs=compare_input,
                outputs=compare_output,
                api_name="compare",
            )

            models_btn.click(
                fn=list_models,
                inputs=None,
                outputs=status_output,
                api_name="models",
            )

            health_btn.click(
                fn=health_check,
                inputs=None,
                outputs=status_output,
                api_name="health",
            )

        # ===== Tab 3: Integration Guide =====
        with gr.Tab("Integration Guide", id="integration"):
            gr.Markdown(INTEGRATION_GUIDE_MD, elem_classes=["api-docs"])

    # Footer
    gr.Markdown(
        "<center style='color:#6b7280; margin-top:16px;'>"
        "CyberSec-API v1.0.0 | "
        "<a href='https://huggingface.co/AYI-NEDJIMI' target='_blank'>AYI-NEDJIMI</a> | "
        "Powered by Hugging Face Inference API"
        "</center>"
    )


# ---------------------------------------------------------------------------
# Launch
# ---------------------------------------------------------------------------

if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        show_api=True,
    )