logos/connectors.py

"""
connectors.py - External API/Service Adapters
Protocol 4: Autonomous Resource Integration

This module isolates all external dependencies so the core engine remains pure.
Each connector wraps an external API/library with a standardized interface.
"""

import os
from typing import Optional, Dict, Any, List
from dataclasses import dataclass


# ==========================================
# CONFIGURATION
# ==========================================

@dataclass
class ConnectorConfig:
    """Configuration for external connectors."""
    hf_token: Optional[str] = None
    hf_space_id: Optional[str] = None
    
    @classmethod
    def from_env(cls) -> 'ConnectorConfig':
        """Load configuration from environment variables."""
        return cls(
            hf_token=os.environ.get('HF_TOKEN'),
            hf_space_id=os.environ.get('HF_SPACE_ID')
        )


# ==========================================
# HUGGING FACE CONNECTOR
# ==========================================

class HuggingFaceConnector:
    """
    Adapter for Hugging Face Hub and Inference API.
    Wraps huggingface_hub for model loading and inference.
    """
    
    def __init__(self, config: ConnectorConfig = None):
        self.config = config or ConnectorConfig.from_env()
        self._client = None
    
    def _ensure_client(self):
        """Lazy initialization of HF client."""
        if self._client is None:
            try:
                from huggingface_hub import InferenceClient
                self._client = InferenceClient(token=self.config.hf_token, base_url="https://router.huggingface.co")
            except ImportError:
                raise ImportError("huggingface_hub not installed. Run: pip install huggingface_hub")
        return self._client
    
    def image_to_text(self, image_path: str, model: str = "Salesforce/blip-image-captioning-base") -> str:
        """
        Generate text description from image using HF Inference API.
        
        Args:
            image_path: Path to image file
            model: HF model ID for image captioning
            
        Returns:
            Generated text description
        """
        client = self._ensure_client()
        with open(image_path, 'rb') as f:
            result = client.image_to_text(f.read(), model=model)
        return result
    
    def text_generation(self, prompt: str, model: str = "gpt2", max_length: int = 100) -> str:
        """
        Generate text from prompt using HF Inference API.
        
        Args:
            prompt: Input text prompt
            model: HF model ID for text generation
            max_length: Maximum output length
            
        Returns:
            Generated text
        """
        client = self._ensure_client()
        result = client.text_generation(prompt, model=model, max_new_tokens=max_length)
        return result


# ==========================================
# OCR CONNECTOR
# ==========================================

class OCRConnector:
    """
    Adapter for Optical Character Recognition via Local Vision Model.
    Uses 'google/gemma-3-4b' (or configured local model) to transcribe text from images.
    """
    
    def __init__(self, languages: List[str] = None, gpu: bool = False):
        # We rely on the local LLM connector, 'gpu' arg is ignored as it's handled by LM Studio
        # Hardcoded to Gemma as requested by user ("gemma is your vision model")
        self.client = get_connector('local', model="google/gemma-3-4b")
    
    def extract_text(self, image_path: str) -> Dict[str, Any]:
        """
        Extract text from image using Vision Model.
        """
        try:
            prompt = "Extract and transcribe all visible text from this image exactly as it appears. Return only the text."
            full_text, _ = self.client.chat(message=prompt, image_path=image_path)
            
            # Simple heuristic for word count
            word_count = len(full_text.split())
            
            return {
                "text_blocks": [], # VLM doesn't give bounding boxes easily
                "full_text": full_text,
                "word_count": word_count
            }
        except Exception as e:
            return {
                "text_blocks": [],
                "full_text": f"[OCR ERROR] Vision Model Failed: {e}",
                "word_count": 0
            }


# ==========================================
# VISION CONNECTOR (Future: Multi-modal)
# ==========================================

class VisionConnector:
    """
    Adapter for computer vision operations.
    Wraps OpenCV and scikit-image.
    """
    
    @staticmethod
    def calculate_ssim(image1_path: str, image2_path: str) -> float:
        """
        Calculate Structural Similarity Index between two images.
        Uses scikit-image for accurate SSIM calculation.
        
        Args:
            image1_path: Path to first image
            image2_path: Path to second image
            
        Returns:
            SSIM score (0-1, higher is better)
        """
        try:
            import cv2
            from skimage.metrics import structural_similarity as ssim
            
            img1 = cv2.imread(image1_path)
            img2 = cv2.imread(image2_path)
            
            # Resize if needed
            if img1.shape != img2.shape:
                img2 = cv2.resize(img2, (img1.shape[1], img1.shape[0]))
            
            # Convert to grayscale for SSIM
            gray1 = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
            gray2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
            
            return ssim(gray1, gray2)
            
        except ImportError as e:
            raise ImportError(f"Required library not installed: {e}")
    
    @staticmethod
    def analyze_entropy(image_path: str) -> Dict[str, float]:
        """
        Analyze image entropy (information density).
        
        Args:
            image_path: Path to image file
            
        Returns:
            Dict with entropy metrics
        """
        try:
            import cv2
            import numpy as np
            from skimage.measure import shannon_entropy
            
            img = cv2.imread(image_path)
            gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
            
            # Calculate entropy
            entropy = shannon_entropy(gray)
            
            # Calculate histogram entropy
            hist = cv2.calcHist([gray], [0], None, [256], [0, 256])
            hist = hist.flatten() / hist.sum()
            hist_entropy = -np.sum(hist[hist > 0] * np.log2(hist[hist > 0]))
            
            return {
                "shannon_entropy": entropy,
                "histogram_entropy": hist_entropy,
                "mean_intensity": float(np.mean(gray)),
                "std_intensity": float(np.std(gray))
            }
            
        except ImportError as e:
            raise ImportError(f"Required library not installed: {e}")

# ==========================================
# DOLPHIN AGENT CONNECTOR (HF Inference)
# ==========================================

class DolphinAgentConnector:
    """
    Adapter for Dolphin AI (via Hugging Face Inference).
    Replaces NeMo/OpenAI dependency with open weights.
    """
    
    def __init__(self, model: str = "cognitivecomputations/dolphin-2.9-llama3-8b"):
        self.model = model
        self.config = ConnectorConfig.from_env()
        self._client = None
    
    def _ensure_client(self):
        """Lazy initialization of HF Inference Client."""
        if self._client is None:
            try:
                from huggingface_hub import InferenceClient
                self._client = InferenceClient(token=self.config.hf_token, base_url="https://router.huggingface.co")
            except ImportError:
                raise ImportError("huggingface_hub not installed.")
        return self._client
    
    def chat(self, message: str, system_prompt: str = None) -> str:
        """
        Chat with Dolphin agent.
        """
        try:
            client = self._ensure_client()
            
            messages = []
            if system_prompt:
                messages.append({"role": "system", "content": system_prompt})
            messages.append({"role": "user", "content": message})
            
            # Using basic text generation if chat template fails, but try chat first
            # Many HF models support chat_completion API via InferenceClient
            try:
                response = client.chat_completion(
                    messages=messages,
                    model=self.model,
                    max_tokens=500
                )
                return response.choices[0].message.content, response.choices[0].get('logprobs')
            except Exception:
                # Fallback to text generation
                prompt = f"<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
                res = client.text_generation(prompt, model=self.model, max_new_tokens=500)
                return res, None
                
        except Exception as e:
            return f"[Dolphin Error] {e}", None
    
    def analyze_diagram(self, image_path: str, prompt: str = "Describe this architectural diagram.") -> str:
        """
        Analyze diagram using visual model (fallback to simple captioning if Dolphin is text-only).
        """
        try:
            # Dolphin is text-only usually. Route to a Vision model.
            from .connectors import get_connector # lazy import
            hf = get_connector('hf')
            return hf.image_to_text(image_path)
        except Exception as e:
            return f"[Vision Error] {e}"


# ==========================================
# LOCAL LLM CONNECTOR (Ollama/LM Studio)
# ==========================================

class LocalLLMConnector:
    """
    Adapter for Local Inference (Ollama / LM Studio).
    Uses OpenAI-compatible endpoint structure.
    Optimization: Direct localhost access (no Docker bridge lag).
    """
    
    def __init__(self, base_url: str = None, model: str = "dolphin-x1-8b"):
        # Prioritize Environment -> Argument -> Default
        env_url = os.environ.get("LOGOS_LLM_ENDPOINT")
        self.base_url = base_url or env_url or "http://localhost:1234/v1"
        self.model = model
    
    async def chat_async(self, message: str, system_prompt: str = None, model: str = None, **kwargs):
        """
        Asynchronous chat with local model via aiohttp.
        Supports extra params via kwargs (e.g., max_tokens, temperature).
        """
        import aiohttp
        import json

        target_model = model or self.model
        payload = {
            "model": target_model,
            "messages": [],
            "temperature": 0.7,
            "stream": False
        }
        if system_prompt:
            payload["messages"].append({"role": "system", "content": system_prompt})
        payload["messages"].append({"role": "user", "content": message})
        payload["logprobs"] = True
        payload["top_logprobs"] = 1
        
        # Merge extra args (e.g. max_tokens)
        payload.update(kwargs)

        endpoint = f"{self.base_url}/chat/completions"
        try:
            async with aiohttp.ClientSession() as session:
                async with session.post(endpoint, json=payload, timeout=30) as response:
                    if response.status == 200:
                        data = await response.json()
                        content = data['choices'][0]['message'].get('content', "")
                        logprobs = data['choices'][0].get('logprobs')
                        return content, logprobs
                    else:
                        return f"[Error] Local LLM returned status {response.status}", None
        except Exception as e:
            return f"[Async Local LLM Error] {e}", None

    def chat(self, message: str, system_prompt: str = None, model: str = None, image_path: str = None) -> str:
        """
        Chat with local model via requests. Supports Vision if image_path is provided.
        Auto-detects Docker host.
        """
        import requests
        import json
        import base64
        import os

        # Helper to encode image
        def encode_image(path):
            with open(path, "rb") as image_file:
                return base64.b64encode(image_file.read()).decode('utf-8')

        # Potential endpoints to try
        endpoints = [self.base_url]
        if "localhost" in self.base_url:
             endpoints.append(self.base_url.replace("localhost", "host.docker.internal"))
        
        # Use instance default if no specific model requested
        target_model = model or self.model
        
        payload = {
            "model": target_model,
            "messages": [],
            "temperature": 0.7,
            "stream": False,
            "logprobs": True,
            "top_logprobs": 1
        }
        
        if system_prompt:
            payload["messages"].append({"role": "system", "content": system_prompt})
            
        if image_path and os.path.exists(image_path):
            # Format message for Vision API (OpenAI compatible)
            base64_image = encode_image(image_path)
            user_content = [
                {"type": "text", "text": message},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": f"data:image/jpeg;base64,{base64_image}"
                    }
                }
            ]
            payload["messages"].append({"role": "user", "content": user_content})
        else:
            # Standard Text Chat
            payload["messages"].append({"role": "user", "content": message})
        
        last_error = ""
        
        for base in endpoints:
            endpoint = f"{base}/chat/completions"
            try:
                # Increased timeout for complex/vision tasks
                response = requests.post(endpoint, json=payload, timeout=30)
                response.raise_for_status()
                if response.status_code == 200:
                    data = response.json()
                    content = data['choices'][0]['message'].get('content', "")
                    logprobs = data['choices'][0].get('logprobs')
                    return content, logprobs
            except Exception as e:
                last_error = str(e)
                continue
                
        return f"[Local LLM Error] Could not connect to Local Swarm on {endpoints}. Is LM Studio running? ({last_error})", None


# ==========================================
# BROWSER AUTOMATION CONNECTOR (smolagents)
# ==========================================

class BrowserAutomationConnector:
    """
    Adapter for HuggingFace smolagents + helium for browser automation.
    Enables autonomous web navigation, form filling, and data extraction.
    """
    
    def __init__(self, headless: bool = True):
        self.headless = headless
        self._driver = None
        self._agent = None
    
    def _ensure_browser(self):
        """Initialize browser with helium."""
        if self._driver is None:
            try:
                from helium import start_chrome, go_to, click, write, get_driver
                from selenium.webdriver.chrome.options import Options
                
                options = Options()
                if self.headless:
                    options.add_argument('--headless')
                    options.add_argument('--no-sandbox')
                    options.add_argument('--disable-dev-shm-usage')
                
                start_chrome(headless=self.headless)
                self._driver = get_driver()
            except ImportError:
                raise ImportError("helium/selenium not installed. Run: pip install helium selenium")
        return self._driver
    
    def navigate(self, url: str) -> str:
        """Navigate to a URL and return page title."""
        try:
            from helium import go_to, get_driver
            self._ensure_browser()
            go_to(url)
            return f"Navigated to: {get_driver().title}"
        except Exception as e:
            return f"[Navigation Error] {e}"
    
    def search_page(self, query: str) -> str:
        """Search for text on current page."""
        try:
            driver = self._ensure_browser()
            page_source = driver.page_source.lower()
            if query.lower() in page_source:
                return f"Found '{query}' on page"
            else:
                return f"'{query}' not found on page"
        except Exception as e:
            return f"[Search Error] {e}"
    
    def get_page_text(self) -> str:
        """Extract visible text from current page."""
        try:
            driver = self._ensure_browser()
            from bs4 import BeautifulSoup
            soup = BeautifulSoup(driver.page_source, 'html.parser')
            text = soup.get_text(separator=' ', strip=True)
            return text[:5000]  # Limit to 5k chars
        except Exception as e:
            return f"[Extraction Error] {e}"
    
    def click_element(self, text: str) -> str:
        """Click an element by its text."""
        try:
            from helium import click
            self._ensure_browser()
            click(text)
            return f"Clicked: {text}"
        except Exception as e:
            return f"[Click Error] {e}"
    
    def type_text(self, field: str, text: str) -> str:
        """Type text into a field."""
        try:
            from helium import write
            self._ensure_browser()
            write(text, into=field)
            return f"Typed into {field}"
        except Exception as e:
            return f"[Type Error] {e}"
    
    def close(self):
        """Close the browser."""
        try:
            from helium import kill_browser
            kill_browser()
            self._driver = None
        except:
            pass
    
    def run_agent_task(self, task: str, model: str = "Qwen/Qwen2.5-Coder-32B-Instruct") -> str:
        """
        Run a browser automation task using smolagents Code Agent.
        
        Args:
            task: Natural language task description
            model: HF model for the code agent
            
        Returns:
            Task result
        """
        try:
            from smolagents import CodeAgent, HfApiModel
            from smolagents.tools import Tool
            
            # Create agent with browser tools
            model_instance = HfApiModel(model)
            agent = CodeAgent(tools=[], model=model_instance)
            
            # Run the task
            result = agent.run(task)
            return str(result)
        except ImportError:
            return "[Error] smolagents not installed. Run: pip install smolagents"
        except Exception as e:
            return f"[Agent Error] {e}"


# ==========================================
# FACTORY
# ==========================================

def get_connector(connector_type: str, **kwargs) -> Any:
    """
    Factory function for connectors.
    
    Args:
        connector_type: One of 'hf', 'ocr', 'vision', 'nemo', 'browser'
        **kwargs: Connector-specific arguments
        
    Returns:
        Initialized connector instance
    """
    connectors = {
        'hf': HuggingFaceConnector,
        'ocr': OCRConnector,
        'vision': VisionConnector,
        'dolphin': DolphinAgentConnector,
        'browser': BrowserAutomationConnector,
        'local': LocalLLMConnector
    }
    
    if connector_type not in connectors:
        raise ValueError(f"Unknown connector type: {connector_type}. Available: {list(connectors.keys())}")
    
    return connectors[connector_type](**kwargs)


# ==========================================
# REGISTRY (For Protocol 4 Discovery)
# ==========================================

AVAILABLE_CONNECTORS = {
    'hf': {
        'name': 'Hugging Face',
        'capabilities': ['image_to_text', 'text_generation'],
        'requires': ['huggingface_hub'],
        'env_vars': ['HF_TOKEN']
    },
    'ocr': {
        'name': 'EasyOCR',
        'capabilities': ['extract_text'],
        'requires': ['easyocr'],
        'env_vars': []
    },
    'vision': {
        'name': 'Vision (OpenCV/scikit-image)',
        'capabilities': ['calculate_ssim', 'analyze_entropy'],
        'requires': ['opencv-python-headless', 'scikit-image'],
        'env_vars': []
    },
    'dolphin': {
        'name': 'Dolphin AI (HF Inference)',
        'capabilities': ['chat', 'analyze_diagram'],
        'requires': ['huggingface_hub'],
        'env_vars': ['HF_TOKEN']
    },
    'browser': {
        'name': 'Browser Automation (smolagents)',
        'capabilities': ['navigate', 'click_element', 'type_text', 'get_page_text', 'run_agent_task'],
        'requires': ['smolagents', 'helium', 'selenium', 'beautifulsoup4'],
        'env_vars': []
    },
    'local': {
        'name': 'Local LLM (Ollama/LM Studio)',
        'capabilities': ['chat'],
        'requires': ['requests'],
        'env_vars': []
    }
}

# ==========================================
# CLI / TESTING
# ==========================================

def main():
    """
    CLI for testing connectors directly.
    Usage: python -m logos.connectors --test local --model google/gemma-3-4b
    """
    import argparse
    import sys
    
    parser = argparse.ArgumentParser(description="LOGOS Connectors Utilities")
    parser.add_argument("--test", choices=list(AVAILABLE_CONNECTORS.keys()), help="Connector to test")
    parser.add_argument("--model", help="Model name for HF or Local connector")
    parser.add_argument("--prompt", default="Hello, are you online?", help="Prompt to send")
    parser.add_argument("--image", help="Path to image for OCR or Vision test")
    
    args = parser.parse_args()
    
    if not args.test:
        print("Available Connectors:")
        for k, v in AVAILABLE_CONNECTORS.items():
            print(f"  - {k:<10} : {v['name']}")
        print("\nRun with --test <name> to verify a connection.")
        return

    print(f"--- Testing Connector: {args.test.upper()} ---")
    
    try:
        if args.test == 'local':
            # Local LLM / Vision Test
            model = args.model or "local-model"
            print(f"Targeting Model: {model}")
            client = get_connector('local', model=model)
            
            if args.image:
                print(f"Sending Vision Request with {args.image}...")
                resp = client.chat(args.prompt, image_path=args.image)
            else:
                print(f"Sending Chat Request: '{args.prompt}'...")
                resp = client.chat(args.prompt)
            
            print(f"\n[RESPONSE]\n{resp}")

        elif args.test == 'ocr':
            # OCR Test (via Vision)
            if not args.image:
                print("Error: --image argument required for OCR test.")
                return
            
            client = get_connector('ocr')
            print(f"Extracting text from {args.image}...")
            res = client.extract_text(args.image)
            print(f"\n[RESULT]\n{res['full_text']}")

        elif args.test == 'hf':
            # Hugging Face Test
            client = get_connector('hf')
            if args.image:
                # Image Captioning
                resp = client.image_to_text(args.image)
            else:
                # Text Gen
                resp = client.text_generation(args.prompt)
            print(f"\n[RESPONSE]\n{resp}")
            
        else:
            print(f"Test CLI not yet implemented for {args.test}. Import and use in Python.")

    except Exception as e:
        print(f"\n[FAIL] {e}")
        import traceback
        traceback.print_exc()

if __name__ == "__main__":
    main()