# File: enhanced_gradio_interface.py import asyncio from collections import defaultdict import json import os import re import time import uuid from typing import List, Dict, Any, Optional from dataclasses import dataclass from threading import Lock import threading import json import os import queue import traceback import uuid from typing import Coroutine, Dict, List, Any, Optional, Callable from dataclasses import dataclass from queue import Queue, Empty from threading import Lock, Event, Thread import threading from concurrent.futures import ThreadPoolExecutor import time import gradio as gr from openai import AsyncOpenAI, OpenAI import pyttsx3 from rich.console import Console BASE_URL="http://localhost:1234/v1" BASE_API_KEY="not-needed" BASE_CLIENT = AsyncOpenAI( base_url=BASE_URL, api_key=BASE_API_KEY ) # Global state for client BASEMODEL_ID = "leroydyer/qwen/qwen3-0.6b-q4_k_m.gguf" # Global state for selected model ID CLIENT =OpenAI( base_url=BASE_URL, api_key=BASE_API_KEY ) # Global state for client # --- Global Variables (if needed) --- console = Console() # --- Configuration --- LOCAL_BASE_URL = "http://localhost:1234/v1" LOCAL_API_KEY = "not-needed" # HuggingFace Spaces configuration HF_INFERENCE_URL = "https://api-inference.huggingface.co/models/" HF_API_KEY = os.getenv("HF_API_KEY", "") # Available model options MODEL_OPTIONS = { "Local LM Studio": LOCAL_BASE_URL, "Codellama 7B": "codellama/CodeLlama-7b-hf", "Mistral 7B": "mistralai/Mistral-7B-v0.1", "Llama 2 7B": "meta-llama/Llama-2-7b-chat-hf", "Falcon 7B": "tiiuae/falcon-7b-instruct" } DEFAULT_TEMPERATURE = 0.7 DEFAULT_MAX_TOKENS = 5000 console = Console() # --- Canvas Artifact Support --- @dataclass class CanvasArtifact: id: str type: str # 'code', 'diagram', 'text', 'image' content: str title: str timestamp: float metadata: Dict[str, Any] @dataclass class LLMMessage: role: str content: str message_id: str = None conversation_id: str = None timestamp: float = None metadata: Dict[str, Any] = None def __post_init__(self): if self.message_id is None: self.message_id = str(uuid.uuid4()) if self.timestamp is None: self.timestamp = time.time() if self.metadata is None: self.metadata = {} @dataclass class LLMRequest: message: LLMMessage response_event: str = None callback: Callable = None def __post_init__(self): if self.response_event is None: self.response_event = f"llm_response_{self.message.message_id}" @dataclass class LLMResponse: message: LLMMessage request_id: str success: bool = True error: str = None # --- Event Manager (copied from your original code or imported) --- class EventManager: def __init__(self): self._handlers = defaultdict(list) self._lock = threading.Lock() def register(self, event: str, handler: Callable): with self._lock: self._handlers[event].append(handler) def unregister(self, event: str, handler: Callable): with self._lock: if event in self._handlers and handler in self._handlers[event]: self._handlers[event].remove(handler) def raise_event(self, event: str, data: Any): with self._lock: handlers = self._handlers[event][:] for handler in handlers: try: handler(data) except Exception as e: console.log(f"Error in event handler for {event}: {e}", style="bold red") EVENT_MANAGER = EventManager() def RegisterEvent(event: str, handler: Callable): EVENT_MANAGER.register(event, handler) def RaiseEvent(event: str, data: Any): EVENT_MANAGER.raise_event(event, data) def UnregisterEvent(event: str, handler: Callable): EVENT_MANAGER.unregister(event, handler) class LLMAgent: """Main Agent Driver ! Agent For Multiple messages at once , has a message queing service as well as agenerator method for easy intergration with console applications as well as ui !""" def __init__( self, model_id: str = BASEMODEL_ID, system_prompt: str = None, max_queue_size: int = 1000, max_retries: int = 3, timeout: int = 30000, max_tokens: int = 5000, temperature: float = 0.3, base_url: str = "http://localhost:1234/v1", api_key: str = "not-needed", generate_fn: Callable[[List[Dict[str, str]]], Coroutine[Any, Any, str]] = None ): self.model_id = model_id self.system_prompt = system_prompt or "You are a helpful AI assistant." self.request_queue = Queue(maxsize=max_queue_size) self.max_retries = max_retries self.timeout = timeout self.is_running = False self._stop_event = Event() self.processing_thread = None # Conversation tracking self.conversations: Dict[str, List[LLMMessage]] = {} self.max_history_length = 20 self._generate = generate_fn or self._default_generate self.api_key = api_key self.base_url = base_url self.max_tokens = max_tokens self.temperature = temperature self.async_client = self.CreateClient(base_url, api_key) # Canvas Artifacts - NEW self.canvas_artifacts: Dict[str, List[CanvasArtifact]] = {} self.canvas_lock = Lock() # Active requests waiting for responses self.pending_requests: Dict[str, LLMRequest] = {} self.pending_requests_lock = Lock() # Speech synthesis try: self.tts_engine = pyttsx3.init() self.setup_tts() self.speech_enabled = True except Exception as e: console.log(f"[yellow]TTS not available: {e}[/yellow]") self.speech_enabled = False # Register internal event handlers self._register_event_handlers() # Start the processing thread immediately self.start() def setup_tts(self): """Configure text-to-speech engine""" if hasattr(self, 'tts_engine'): voices = self.tts_engine.getProperty('voices') if voices: self.tts_engine.setProperty('voice', voices[0].id) self.tts_engine.setProperty('rate', 150) self.tts_engine.setProperty('volume', 0.8) def speak(self, text: str): """Convert text to speech in a non-blocking way""" if not hasattr(self, 'speech_enabled') or not self.speech_enabled: return def _speak(): try: # Clean text for speech (remove markdown, code blocks) clean_text = re.sub(r'```.*?```', '', text, flags=re.DOTALL) clean_text = re.sub(r'`.*?`', '', clean_text) clean_text = clean_text.strip() if clean_text: self.tts_engine.say(clean_text) self.tts_engine.runAndWait() else: self.tts_engine.say(text) self.tts_engine.runAndWait() except Exception as e: console.log(f"[red]TTS Error: {e}[/red]") thread = threading.Thread(target=_speak, daemon=True) thread.start() async def _default_generate(self, messages: List[Dict[str, str]]) -> str: """Default generate function if none provided""" return await self.openai_generate(messages) def _register_event_handlers(self): """Register internal event handlers for response routing""" RegisterEvent("llm_internal_response", self._handle_internal_response) def _handle_internal_response(self, response: LLMResponse): """Route responses to the appropriate request handlers""" console.log(f"[bold cyan]Handling internal response for: {response.request_id}[/bold cyan]") request = None with self.pending_requests_lock: if response.request_id in self.pending_requests: request = self.pending_requests[response.request_id] del self.pending_requests[response.request_id] console.log(f"Found pending request for: {response.request_id}") else: console.log(f"No pending request found for: {response.request_id}", style="yellow") return # Raise the specific response event if request.response_event: console.log(f"[bold green]Raising event: {request.response_event}[/bold green]") RaiseEvent(request.response_event, response) # Call callback if provided if request.callback: try: console.log(f"[bold yellow]Calling callback for: {response.request_id}[/bold yellow]") request.callback(response) except Exception as e: console.log(f"Error in callback: {e}", style="bold red") def _add_to_conversation_history(self, conversation_id: str, message: LLMMessage): """Add message to conversation history""" if conversation_id not in self.conversations: self.conversations[conversation_id] = [] self.conversations[conversation_id].append(message) # Trim history if too long if len(self.conversations[conversation_id]) > self.max_history_length * 2: self.conversations[conversation_id] = self.conversations[conversation_id][-(self.max_history_length * 2):] def _build_messages_from_conversation(self, conversation_id: str, new_message: LLMMessage) -> List[Dict[str, str]]: """Build message list from conversation history""" messages = [] # Add system prompt if self.system_prompt: messages.append({"role": "system", "content": self.system_prompt}) # Add conversation history if conversation_id in self.conversations: for msg in self.conversations[conversation_id][-self.max_history_length:]: messages.append({"role": msg.role, "content": msg.content}) # Add the new message messages.append({"role": new_message.role, "content": new_message.content}) return messages def _process_llm_request(self, request: LLMRequest): """Process a single LLM request""" console.log(f"[bold green]Processing LLM request: {request.message.message_id}[/bold green]") try: # Build messages for LLM messages = self._build_messages_from_conversation( request.message.conversation_id or "default", request.message ) console.log(f"Calling LLM with {len(messages)} messages") # Call LLM - Use sync call for thread compatibility response_content = self._call_llm_sync(messages) console.log(f"[bold green]LLM response received: {response_content}...[/bold green]") # Create response message response_message = LLMMessage( role="assistant", content=response_content, conversation_id=request.message.conversation_id, metadata={"request_id": request.message.message_id} ) # Update conversation history self._add_to_conversation_history( request.message.conversation_id or "default", request.message ) self._add_to_conversation_history( request.message.conversation_id or "default", response_message ) # Create and send response response = LLMResponse( message=response_message, request_id=request.message.message_id, success=True ) console.log(f"[bold blue]Sending internal response for: {request.message.message_id}[/bold blue]") RaiseEvent("llm_internal_response", response) except Exception as e: console.log(f"[bold red]Error processing LLM request: {e}[/bold red]") traceback.print_exc() # Create error response error_response = LLMResponse( message=LLMMessage( role="system", content=f"Error: {str(e)}", conversation_id=request.message.conversation_id ), request_id=request.message.message_id, success=False, error=str(e) ) RaiseEvent("llm_internal_response", error_response) def _call_llm_sync(self, messages: List[Dict[str, str]]) -> str: """Sync call to the LLM with retry logic""" console.log(f"Making LLM call to {self.model_id}") for attempt in range(self.max_retries): try: response = CLIENT.chat.completions.create( model=self.model_id, messages=messages, temperature=self.temperature, max_tokens=self.max_tokens ) content = response.choices[0].message.content console.log(f"LLM call successful, response length: {len(content)}") return content except Exception as e: console.log(f"LLM call attempt {attempt + 1} failed: {e}") if attempt == self.max_retries - 1: raise e # Wait before retry def _process_queue(self): """Main queue processing loop""" console.log("[bold cyan]LLM Agent queue processor started[/bold cyan]") while not self._stop_event.is_set(): try: request = self.request_queue.get(timeout=1.0) if request: console.log(f"Got request from queue: {request.message.message_id}") self._process_llm_request(request) self.request_queue.task_done() except Empty: continue except Exception as e: console.log(f"Error in queue processing: {e}", style="bold red") traceback.print_exc() console.log("[bold cyan]LLM Agent queue processor stopped[/bold cyan]") def send_message( self, content: str, role: str = "user", conversation_id: str = None, response_event: str = None, callback: Callable = None, metadata: Dict = None ) -> str: """Send a message to the LLM and get response via events""" if not self.is_running: raise RuntimeError("LLM Agent is not running. Call start() first.") # Create message message = LLMMessage( role=role, content=content, conversation_id=conversation_id, metadata=metadata or {} ) # Create request request = LLMRequest( message=message, response_event=response_event, callback=callback ) # Store in pending requests BEFORE adding to queue with self.pending_requests_lock: self.pending_requests[message.message_id] = request console.log(f"Added to pending requests: {message.message_id}") # Add to queue try: self.request_queue.put(request, timeout=5.0) console.log(f"[bold magenta]Message queued: {message.message_id}, Content: {content[:50]}...[/bold magenta]") return message.message_id except queue.Full: console.log(f"[bold red]Queue full, cannot send message[/bold red]") with self.pending_requests_lock: if message.message_id in self.pending_requests: del self.pending_requests[message.message_id] raise RuntimeError("LLM Agent queue is full") async def chat(self, messages: List[Dict[str, str]]) -> str: """ Async chat method that sends message via queue and returns response string. This is the main method you should use. """ # Create future for the response loop = asyncio.get_event_loop() response_future = loop.create_future() def chat_callback(response: LLMResponse): """Callback when LLM responds - thread-safe""" console.log(f"[bold yellow]✓ CHAT CALLBACK TRIGGERED![/bold yellow]") if not response_future.done(): if response.success: content = response.message.content console.log(f"Callback received content: {content}...") # Schedule setting the future result on the main event loop loop.call_soon_threadsafe(response_future.set_result, content) else: console.log(f"Error in response: {response.error}") error_msg = f"❌ Error: {response.error}" loop.call_soon_threadsafe(response_future.set_result, error_msg) else: console.log(f"[bold red]Future already done, ignoring callback[/bold red]") console.log(f"Sending message to LLM agent...") # Extract the actual message content from the messages list user_message = "" for msg in messages: if msg.get("role") == "user": user_message = msg.get("content", "") break if not user_message.strip(): return "" # Send message with callback using the queue system try: message_id = self.send_message( content=user_message, conversation_id="default", callback=chat_callback ) console.log(f"Message sent with ID: {message_id}, waiting for response...") # Wait for the response and return it try: response = await asyncio.wait_for(response_future, timeout=self.timeout) console.log(f"[bold green]✓ Chat complete! Response length: {len(response)}[/bold green]") return response except asyncio.TimeoutError: console.log("[bold red]Response timeout[/bold red]") # Clean up the pending request with self.pending_requests_lock: if message_id in self.pending_requests: del self.pending_requests[message_id] return "❌ Response timeout - check if LLM server is running" except Exception as e: console.log(f"[bold red]Error sending message: {e}[/bold red]") traceback.print_exc() return f"❌ Error sending message: {e}" def start(self): """Start the LLM agent""" if not self.is_running: self.is_running = True self._stop_event.clear() self.processing_thread = Thread(target=self._process_queue, daemon=True) self.processing_thread.start() console.log("[bold green]LLM Agent started[/bold green]") def stop(self): """Stop the LLM agent""" console.log("Stopping LLM Agent...") self._stop_event.set() if self.processing_thread and self.processing_thread.is_alive(): self.processing_thread.join(timeout=10) self.is_running = False console.log("LLM Agent stopped") def get_conversation_history(self, conversation_id: str = "default") -> List[LLMMessage]: """Get conversation history""" return self.conversations.get(conversation_id, [])[:] def clear_conversation(self, conversation_id: str = "default"): """Clear conversation history""" if conversation_id in self.conversations: del self.conversations[conversation_id] async def _chat(self, messages: List[Dict[str, str]]) -> str: return await self._generate(messages) @staticmethod async def openai_generate(messages: List[Dict[str, str]], max_tokens: int = 8096, temperature: float = 0.4, model: str = BASEMODEL_ID,tools=None) -> str: """Static method for generating responses using OpenAI API""" try: resp = await BASE_CLIENT.chat.completions.create( model=model, messages=messages, temperature=temperature, max_tokens=max_tokens, tools=tools ) response_text = resp.choices[0].message.content or "" return response_text except Exception as e: console.log(f"[bold red]Error in openai_generate: {e}[/bold red]") return f"[LLM_Agent Error - openai_generate: {str(e)}]" async def _call_(self, messages: List[Dict[str, str]]) -> str: """Internal call method using instance client""" try: resp = await self.async_client.chat.completions.create( model=self.model_id, messages=messages, temperature=self.temperature, max_tokens=self.max_tokens ) response_text = resp.choices[0].message.content or "" return response_text except Exception as e: console.log(f"[bold red]Error in _call_: {e}[/bold red]") return f"[LLM_Agent Error - _call_: {str(e)}]" @staticmethod def CreateClient(base_url: str, api_key: str) -> AsyncOpenAI: '''Create async OpenAI Client required for multi tasking''' return AsyncOpenAI( base_url=base_url, api_key=api_key ) @staticmethod async def fetch_available_models(base_url: str, api_key: str) -> List[str]: """Fetches available models from the OpenAI API.""" try: async_client = AsyncOpenAI(base_url=base_url, api_key=api_key) models = await async_client.models.list() model_choices = [model.id for model in models.data] return model_choices except Exception as e: console.log(f"[bold red]LLM_Agent Error fetching models: {e}[/bold red]") return ["LLM_Agent Error fetching models"] def get_models(self) -> List[str]: """Get available models using instance credentials""" return asyncio.run(self.fetch_available_models(self.base_url, self.api_key)) def get_queue_size(self) -> int: """Get current queue size""" return self.request_queue.qsize() def get_pending_requests_count(self) -> int: """Get number of pending requests""" with self.pending_requests_lock: return len(self.pending_requests) def get_status(self) -> Dict[str, Any]: """Get agent status information""" return { "is_running": self.is_running, "queue_size": self.get_queue_size(), "pending_requests": self.get_pending_requests_count(), "conversations_count": len(self.conversations), "model": self.model_id } # --- ADDED CANVAS FUNCTIONALITY --- def add_canvas_artifact(self, conversation_id: str, artifact_type: str, content: str, title: str = ""): """Add an artifact to the canvas for a specific conversation.""" conv_id = conversation_id or "default" with self.canvas_lock: if conv_id not in self.canvas_artifacts: self.canvas_artifacts[conv_id] = [] artifact = CanvasArtifact( id=str(uuid.uuid4()), type=artifact_type, content=content, title=title, timestamp=time.time(), metadata={} ) self.canvas_artifacts[conv_id].append(artifact) console.log(f"[green]Added {artifact_type} artifact to canvas '{conv_id}'[/green]") def get_canvas_summary(self, conversation_id: str) -> List[Dict]: """Get a summary of artifacts on the canvas for JSON display.""" conv_id = conversation_id or "default" with self.canvas_lock: artifacts = self.canvas_artifacts.get(conv_id, []) # Convert artifacts to dictionaries for JSON serialization return [ { "id": art.id, "type": art.type, "title": art.title, "timestamp": art.timestamp, "content_preview": art.content[:100] + "..." if len(art.content) > 100 else art.content } for art in artifacts ] def clear_canvas(self, conversation_id: str): """Clear all artifacts from the canvas for a specific conversation.""" conv_id = conversation_id or "default" with self.canvas_lock: if conv_id in self.canvas_artifacts: self.canvas_artifacts[conv_id].clear() console.log(f"[yellow]Cleared canvas artifacts for '{conv_id}'[/yellow]") async def chat_with_canvas(self, message: str, conversation_id: str, include_canvas: bool = False): """Chat method that can optionally include canvas context.""" messages = [{"role": "user", "content": message}] if include_canvas: artifacts = self.get_canvas_summary(conversation_id) if artifacts: canvas_context = "Current Canvas Context:\\n" + "\\n".join([ f"- [{art['type'].upper()}] {art['title'] or 'Untitled'}: {art['content_preview']}" for art in artifacts ]) messages.insert(0, {"role": "system", "content": canvas_context}) return await self.chat(messages) class AI_Agent: def __init__(self, model_id: str, system_prompt: str = "You are a helpful assistant. Respond concisely in 1-2 sentences.", history: List[Dict] = None): self.model_id = model_id self.system_prompt = system_prompt self.history = history or [] self.conversation_id = f"conv_{uuid.uuid4().hex[:8]}" # Create agent instance self.client = LLMAgent( model_id=model_id, system_prompt=self.system_prompt, generate_fn=LLMAgent.openai_generate ) console.log(f"[bold green]✓ MyAgent initialized with model: {model_id}[/bold green]") async def call_llm(self, messages: List[Dict], use_history: bool = True) -> str: """ Send messages to LLM and get response Args: messages: List of message dicts with 'role' and 'content' use_history: Whether to include conversation history Returns: str: LLM response """ try: console.log(f"[bold yellow]Sending {len(messages)} messages to LLM (use_history: {use_history})...[/bold yellow]") # Enhance messages based on history setting enhanced_messages = await self._enhance_messages(messages, use_history) response = await self.client.chat(enhanced_messages) console.log(f"[bold green]✓ Response received ({len(response)} chars)[/bold green]") # Update conversation history ONLY if we're using history if use_history: self._update_history(messages, response) return response except Exception as e: console.log(f"[bold red]✗ ERROR: {e}[/bold red]") traceback.print_exc() return f"Error: {str(e)}" async def _enhance_messages(self, messages: List[Dict], use_history: bool) -> List[Dict]: """Enhance messages with system prompt and optional history""" enhanced = [] # Add system prompt if not already in messages has_system = any(msg.get('role') == 'system' for msg in messages) if not has_system and self.system_prompt: enhanced.append({"role": "system", "content": self.system_prompt}) # Add conversation history only if requested if use_history and self.history: enhanced.extend(self.history[-10:]) # Last 10 messages for context # Add current messages enhanced.extend(messages) return enhanced def _update_history(self, messages: List[Dict], response: str): """Update conversation history with new exchange""" # Add user messages to history for msg in messages: if msg.get('role') in ['user', 'assistant']: self.history.append(msg) # Add assistant response to history self.history.append({"role": "assistant", "content": response}) # Keep history manageable (last 20 exchanges) if len(self.history) > 40: # 20 user + 20 assistant messages self.history = self.history[-40:] async def simple_query(self, query: str) -> str: """Simple one-shot query method - NO history/context""" messages = [{"role": "user", "content": query}] return await self.call_llm(messages, use_history=False) async def multi_turn_chat(self, user_input: str) -> str: """Multi-turn chat that maintains context across calls""" messages = [{"role": "user", "content": user_input}] response = await self.call_llm(messages, use_history=True) return response def get_conversation_summary(self) -> Dict: """Get conversation summary""" return { "conversation_id": self.conversation_id, "total_messages": len(self.history), "user_messages": len([msg for msg in self.history if msg.get('role') == 'user']), "assistant_messages": len([msg for msg in self.history if msg.get('role') == 'assistant']), "recent_exchanges": self.history[-4:] if self.history else [] } def clear_history(self): """Clear conversation history""" self.history.clear() console.log("[bold yellow]Conversation history cleared[/bold yellow]") def update_system_prompt(self, new_prompt: str): """Update the system prompt""" self.system_prompt = new_prompt console.log(f"[bold blue]System prompt updated[/bold blue]") def stop(self): """Stop the client gracefully""" if hasattr(self, 'client') and self.client: self.client.stop() console.log("[bold yellow]MyAgent client stopped[/bold yellow]") async def contextual_query(self, query: str, context_messages: List[Dict] = None, context_text: str = None, context_files: List[str] = None) -> str: """ Query with specific context but doesn't update main history Args: query: The user question context_messages: List of message dicts for context context_text: Plain text context (will be converted to system message) context_files: List of file paths to read and include as context """ messages = [] # Add system prompt if self.system_prompt: messages.append({"role": "system", "content": self.system_prompt}) # Handle different context types if context_messages: messages.extend(context_messages) if context_text: messages.append({"role": "system", "content": f"Additional context: {context_text}"}) if context_files: file_context = await self._read_files_context(context_files) if file_context: messages.append({"role": "system", "content": f"File contents:\n{file_context}"}) # Add the actual query messages.append({"role": "user", "content": query}) return await self.call_llm(messages, use_history=False) async def _read_files_context(self, file_paths: List[str]) -> str: """Read multiple files and return as context string""" contexts = [] for file_path in file_paths: try: if os.path.exists(file_path): with open(file_path, 'r', encoding='utf-8') as f: content = f.read() contexts.append(f"--- {os.path.basename(file_path)} ---\n{content}") else: console.log(f"[bold yellow]File not found: {file_path}[/bold yellow]") except Exception as e: console.log(f"[bold red]Error reading file {file_path}: {e}[/bold red]") return "\n\n".join(contexts) if contexts else "" async def query_with_code_context(self, query: str, code_snippets: List[str] = None, code_files: List[str] = None) -> str: """ Specialized contextual query for code-related questions """ code_context = "CODE CONTEXT:\n" if code_snippets: for i, snippet in enumerate(code_snippets, 1): code_context += f"\nSnippet {i}:\n```\n{snippet}\n```\n" if code_files: # Read code files and include them for file_path in code_files: if file_path.endswith(('.py', '.js', '.java', '.cpp', '.c', '.html', '.css')): code_context += f"\nFile: {file_path}\n```\n" try: with open(file_path, 'r') as f: code_context += f.read() except Exception as e: code_context += f"Error reading file: {e}" code_context += "\n```\n" return await self.contextual_query(query, context_text=code_context) async def multi_context_query(self, query: str, contexts: Dict[str, Any]) -> str: """ Advanced contextual query with multiple context types Args: query: The user question contexts: Dict with various context types - 'messages': List of message dicts - 'text': Plain text context - 'files': List of file paths - 'urls': List of URLs - 'code': List of code snippets or files - 'metadata': Any additional metadata """ all_context_messages = [] # Build context from different sources if contexts.get('text'): all_context_messages.append({"role": "system", "content": f"Context: {contexts['text']}"}) if contexts.get('messages'): all_context_messages.extend(contexts['messages']) if contexts.get('files'): file_context = await self._read_files_context(contexts['files']) if file_context: all_context_messages.append({"role": "system", "content": f"File Contents:\n{file_context}"}) if contexts.get('code'): code_context = "\n".join([f"Code snippet {i}:\n```\n{code}\n```" for i, code in enumerate(contexts['code'], 1)]) all_context_messages.append({"role": "system", "content": f"Code Context:\n{code_context}"}) if contexts.get('metadata'): all_context_messages.append({"role": "system", "content": f"Metadata: {contexts['metadata']}"}) return await self.contextual_query(query, context_messages=all_context_messages) console = Console() # --- Canvas Artifact Support --- @dataclass class CanvasArtifact: id: str type: str # 'code', 'diagram', 'text', 'image' content: str title: str timestamp: float metadata: Dict[str, Any] class EnhancedAIAgent: """ Wrapper around your AI_Agent that adds canvas/artifact management without modifying the original agent. """ def __init__(self, ai_agent): self.agent = ai_agent self.canvas_artifacts: Dict[str, List[CanvasArtifact]] = {} self.max_canvas_artifacts = 50 console.log("[bold green]✓ Enhanced AI Agent wrapper initialized[/bold green]") def add_artifact_to_canvas(self, conversation_id: str, content: str, artifact_type: str = "code", title: str = None): """Add artifacts to the collaborative canvas""" if conversation_id not in self.canvas_artifacts: self.canvas_artifacts[conversation_id] = [] artifact = CanvasArtifact( id=str(uuid.uuid4())[:8], type=artifact_type, content=content, title=title or f"{artifact_type}_{len(self.canvas_artifacts[conversation_id]) + 1}", timestamp=time.time(), metadata={"conversation_id": conversation_id} ) self.canvas_artifacts[conversation_id].append(artifact) # Keep only recent artifacts if len(self.canvas_artifacts[conversation_id]) > self.max_canvas_artifacts: self.canvas_artifacts[conversation_id] = self.canvas_artifacts[conversation_id][-self.max_canvas_artifacts:] console.log(f"[green]Added artifact to canvas: {artifact.title}[/green]") return artifact def get_canvas_context(self, conversation_id: str) -> str: """Get formatted canvas context for LLM prompts""" if conversation_id not in self.canvas_artifacts or not self.canvas_artifacts[conversation_id]: return "" context_lines = ["\n=== COLLABORATIVE CANVAS ARTIFACTS ==="] for artifact in self.canvas_artifacts[conversation_id][-10:]: # Last 10 artifacts context_lines.append(f"\n--- {artifact.title} [{artifact.type.upper()}] ---") preview = artifact.content[:500] + "..." if len(artifact.content) > 500 else artifact.content context_lines.append(preview) return "\n".join(context_lines) + "\n=================================\n" async def chat_with_canvas(self, message: str, conversation_id: str = "default", include_canvas: bool = True) -> str: """Enhanced chat that includes canvas context""" # Build context with canvas artifacts if requested full_message = message if include_canvas: canvas_context = self.get_canvas_context(conversation_id) if canvas_context: full_message = f"{canvas_context}\n\nUser Query: {message}" try: # Use your original agent's multi_turn_chat method response = await self.agent.multi_turn_chat(full_message) # Auto-extract and add code artifacts to canvas self._extract_artifacts_to_canvas(response, conversation_id) return response except Exception as e: error_msg = f"Error in chat_with_canvas: {str(e)}" console.log(f"[red]{error_msg}[/red]") return error_msg def _extract_artifacts_to_canvas(self, response: str, conversation_id: str): """Automatically extract code blocks and add to canvas""" # Find all code blocks with optional language specification code_blocks = re.findall(r'```(?:(\w+)\n)?(.*?)```', response, re.DOTALL) for i, (lang, code_block) in enumerate(code_blocks): if len(code_block.strip()) > 10: # Only add substantial code blocks self.add_artifact_to_canvas( conversation_id, code_block.strip(), "code", f"code_snippet_{lang or 'unknown'}_{len(self.canvas_artifacts.get(conversation_id, [])) + 1}" ) def get_canvas_summary(self, conversation_id: str) -> List[Dict]: """Get summary of canvas artifacts for display""" if conversation_id not in self.canvas_artifacts: return [] artifacts = [] for artifact in reversed(self.canvas_artifacts[conversation_id]): # Newest first artifacts.append({ "id": artifact.id, "type": artifact.type.upper(), "title": artifact.title, "preview": artifact.content[:100] + "..." if len(artifact.content) > 100 else artifact.content, "timestamp": time.strftime("%H:%M:%S", time.localtime(artifact.timestamp)) }) return artifacts def get_artifact_by_id(self, conversation_id: str, artifact_id: str) -> Optional[CanvasArtifact]: """Get specific artifact by ID""" if conversation_id not in self.canvas_artifacts: return None for artifact in self.canvas_artifacts[conversation_id]: if artifact.id == artifact_id: return artifact return None def clear_canvas(self, conversation_id: str = "default"): """Clear canvas artifacts""" if conversation_id in self.canvas_artifacts: self.canvas_artifacts[conversation_id] = [] console.log(f"[yellow]Cleared canvas: {conversation_id}[/yellow]") def get_latest_code_artifact(self, conversation_id: str) -> Optional[str]: """Get the most recent code artifact content""" if conversation_id not in self.canvas_artifacts: return None for artifact in reversed(self.canvas_artifacts[conversation_id]): if artifact.type == "code": return artifact.content return None console = Console() # --- LCARS Styled Gradio Interface --- class LcarsInterface: def __init__(self): # Start with HuggingFace by default for Spaces self.use_huggingface = True self.agent = LLMAgent(generate_fn=LLMAgent.openai_generate) self.current_conversation = "default" def create_interface(self): """Create the full LCARS-styled interface""" lcars_css = """ :root { --lcars-orange: #FF9900; --lcars-red: #FF0033; --lcars-blue: #6699FF; --lcars-purple: #CC99FF; --lcars-pale-blue: #99CCFF; --lcars-black: #000000; --lcars-dark-blue: #3366CC; --lcars-gray: #424242; --lcars-yellow: #FFFF66; } body { background: var(--lcars-black); color: var(--lcars-orange); font-family: 'Antonio', 'LCD', 'Courier New', monospace; margin: 0; padding: 0; } .gradio-container { background: var(--lcars-black) !important; min-height: 100vh; } .lcars-container { background: var(--lcars-black); border: 4px solid var(--lcars-orange); border-radius: 0 30px 0 0; min-height: 100vh; padding: 20px; } .lcars-header { background: linear-gradient(90deg, var(--lcars-red), var(--lcars-orange)); padding: 20px 40px; border-radius: 0 60px 0 0; margin: -20px -20px 20px -20px; border-bottom: 6px solid var(--lcars-blue); } .lcars-title { font-size: 2.5em; font-weight: bold; color: var(--lcars-black); margin: 0; } .lcars-subtitle { font-size: 1.2em; color: var(--lcars-black); margin: 10px 0 0 0; } .lcars-panel { background: rgba(66, 66, 66, 0.9); border: 2px solid var(--lcars-orange); border-radius: 0 20px 0 20px; padding: 15px; margin-bottom: 15px; } .lcars-button { background: var(--lcars-orange); color: var(--lcars-black) !important; border: none !important; border-radius: 0 15px 0 15px !important; padding: 10px 20px !important; font-family: inherit !important; font-weight: bold !important; margin: 5px !important; } .lcars-button:hover { background: var(--lcars-red) !important; } .lcars-input { background: var(--lcars-black) !important; color: var(--lcars-orange) !important; border: 2px solid var(--lcars-blue) !important; border-radius: 0 10px 0 10px !important; padding: 10px !important; } .lcars-chatbot { background: var(--lcars-black) !important; border: 2px solid var(--lcars-purple) !important; border-radius: 0 15px 0 15px !important; } .status-indicator { display: inline-block; width: 12px; height: 12px; border-radius: 50%; background: var(--lcars-red); margin-right: 8px; } .status-online { background: var(--lcars-blue); animation: pulse 2s infinite; } @keyframes pulse { 0% { opacity: 1; } 50% { opacity: 0.5; } 100% { opacity: 1; } } """ with gr.Blocks(css=lcars_css, theme=gr.themes.Default(), title="LCARS Terminal") as interface: with gr.Column(elem_classes="lcars-container"): # Header with gr.Sidebar(): gr.LoginButton() with gr.Row(elem_classes="lcars-header"): gr.Markdown("""
🚀 LCARS TERMINAL
STARFLEET AI DEVELOPMENT CONSOLE
SYSTEM ONLINE
""") # Main Content with gr.Row(): # Left Sidebar with gr.Column(scale=1): # Configuration Panel with gr.Column(elem_classes="lcars-panel"): # Connection Type Selector with gr.Row(elem_classes="lcars-panel"): connection_type = gr.Radio(label = "### 🌐 CONNECTION TYPE", choices=["HuggingFace Inference", "Local LM Studio"], value="HuggingFace Inference", elem_classes="lcars-input" ) gr.Markdown("### 🔧 CONFIGURATION") # Connection-specific settings with gr.Row(visible=False) as local_settings: base_url = gr.Textbox( value=LOCAL_BASE_URL, label="LM Studio URL", elem_classes="lcars-input" ) api_key = gr.Textbox( value=LOCAL_API_KEY, label="API Key", type="password", elem_classes="lcars-input" ) with gr.Row(visible=True) as hf_settings: hf_api_key = gr.Textbox( value=HF_API_KEY, label="HuggingFace API Key", type="password", elem_classes="lcars-input", placeholder="Get from https://huggingface.co/settings/tokens" ) with gr.Row(): model_dropdown = gr.Dropdown( choices=list(MODEL_OPTIONS.keys())[1:], value=list(MODEL_OPTIONS.keys())[1], label="AI Model", elem_classes="lcars-input" ) fetch_models_btn = gr.Button("📡 Fetch Models", elem_classes="lcars-button") with gr.Row(): temperature = gr.Slider(0.0, 2.0, value=0.7, label="Temperature") max_tokens = gr.Slider(128, 8192, value=2000, step=128, label="Max Tokens") with gr.Row(): update_config_btn = gr.Button("💾 Apply Config", elem_classes="lcars-button") speech_toggle = gr.Checkbox(value=True, label="🔊 Speech Output") # Canvas Artifacts with gr.Column(elem_classes="lcars-panel"): gr.Markdown("### 🎨 CANVAS ARTIFACTS") artifact_display = gr.JSON(label="") with gr.Row(): refresh_artifacts_btn = gr.Button("🔄 Refresh", elem_classes="lcars-button") clear_canvas_btn = gr.Button("🗑️ Clear Canvas", elem_classes="lcars-button") # Main Content Area with gr.Column(scale=2): # Code Canvas with gr.Accordion("💻 COLLABORATIVE CODE CANVAS", open=False): code_editor = gr.Code( value="# Welcome to LCARS Collaborative Canvas\\nprint('Hello, Starfleet!')", language="python", lines=15, label="" ) with gr.Row(): load_to_chat_btn = gr.Button("💬 Discuss Code", elem_classes="lcars-button") analyze_btn = gr.Button("🔍 Analyze", elem_classes="lcars-button") optimize_btn = gr.Button("⚡ Optimize", elem_classes="lcars-button") # Chat Interface with gr.Column(elem_classes="lcars-panel"): gr.Markdown("### 💬 MISSION LOG") chatbot = gr.Chatbot(label="", height=300) with gr.Row(): message_input = gr.Textbox( placeholder="Enter your command or query...", show_label=False, lines=2, scale=4 ) send_btn = gr.Button("🚀 SEND", elem_classes="lcars-button", scale=1) # Status with gr.Row(): status_display = gr.Textbox( value="LCARS terminal operational. Awaiting commands.", label="Status", max_lines=2 ) with gr.Column(scale=0): clear_chat_btn = gr.Button("🗑️ Clear Chat", elem_classes="lcars-button") new_session_btn = gr.Button("🆕 New Session", elem_classes="lcars-button") # === EVENT HANDLERS === def switch_connection(connection_type): if connection_type == "Local LM Studio": return [ gr.update(visible=True), gr.update(visible=False), gr.update(choices=list(MODEL_OPTIONS.keys())[1:], value=list(MODEL_OPTIONS.keys())[1]) ] else: return [ gr.update(visible=False), gr.update(visible=True), gr.update(choices=list(MODEL_OPTIONS.keys())[1:], value=list(MODEL_OPTIONS.keys())[1]) ] async def fetch_models_updated(connection_type, base_url_val, api_key_val, hf_api_key_val): # Fixed: Removed the 'use_huggingface' parameter if connection_type == "Local LM Studio": models = await LLMAgent.fetch_available_models( base_url_val, api_key_val ) else: # Using the HF_INFERENCE_URL and the key models = await LLMAgent.fetch_available_models( HF_INFERENCE_URL, hf_api_key_val ) if models: return gr.update(choices=models, value=models[0]) return gr.update(choices=["No models found"]) def update_agent_connection(connection_type, model_id, base_url_val, api_key_val, hf_api_key_val): # Fixed: Removed the 'use_huggingface' parameter from the constructor use_hf = connection_type == "HuggingFace Inference" if use_hf: # Use the model_id directly (it's the model name like 'codellama/CodeLlama-7b-hf') self.agent = LLMAgent( model_id=model_id, base_url=HF_INFERENCE_URL, api_key=hf_api_key_val, generate_fn=LLMAgent.openai_generate ) return f"✅ Switched to HuggingFace: {model_id}" else: self.agent = LLMAgent( model_id=model_id, base_url=base_url_val, api_key=api_key_val, generate_fn=LLMAgent.openai_generate ) return f"✅ Switched to Local: {base_url_val}" async def process_message(message, history, speech_enabled): if not message.strip(): return "", history, "Please enter a message" history = history + [[message, None]] try: # Fixed: Uses the new chat_with_canvas method which includes canvas context response = await self.agent.chat_with_canvas( message, self.current_conversation, include_canvas=True ) history[-1][1] = response if speech_enabled and self.agent.speech_enabled: self.agent.speak(response) artifacts = self.agent.get_canvas_summary(self.current_conversation) status = f"✅ Response received. Canvas artifacts: {len(artifacts)}" return "", history, status, artifacts except Exception as e: error_msg = f"❌ Error: {str(e)}" history[-1][1] = error_msg return "", history, error_msg, self.agent.get_canvas_summary(self.current_conversation) def get_artifacts(): return self.agent.get_canvas_summary(self.current_conversation) def clear_canvas(): self.agent.clear_canvas(self.current_conversation) return [], "✅ Canvas cleared" def clear_chat(): self.agent.clear_conversation(self.current_conversation) return [], "✅ Chat cleared" def new_session(): self.agent.clear_conversation(self.current_conversation) self.agent.clear_canvas(self.current_conversation) return [], "# New session started\\nprint('Ready!')", "🆕 New session started", [] # Connect events connection_type.change(switch_connection, inputs=connection_type, outputs=[local_settings, hf_settings, model_dropdown]) fetch_models_btn.click(fetch_models_updated, inputs=[connection_type, base_url, api_key, hf_api_key], outputs=model_dropdown) update_config_btn.click(update_agent_connection, inputs=[connection_type, model_dropdown, base_url, api_key, hf_api_key], outputs=status_display) send_btn.click(process_message, inputs=[message_input, chatbot, speech_toggle], outputs=[message_input, chatbot, status_display, artifact_display]) message_input.submit(process_message, inputs=[message_input, chatbot, speech_toggle], outputs=[message_input, chatbot, status_display, artifact_display]) refresh_artifacts_btn.click(get_artifacts, outputs=artifact_display) clear_canvas_btn.click(clear_canvas, outputs=[artifact_display, status_display]) clear_chat_btn.click(clear_chat, outputs=[chatbot, status_display]) new_session_btn.click(new_session, outputs=[chatbot, code_editor, status_display, artifact_display]) interface.load(get_artifacts, outputs=artifact_display) return interface # --- Main Application --- def main(): console.log("[bold blue]🚀 Starting LCARS Terminal...[/bold blue]") is_space = os.getenv('SPACE_ID') is not None if is_space: console.log("[green]🌐 Detected HuggingFace Space[/green]") else: console.log("[blue]💻 Running locally[/blue]") interface = LcarsInterface() demo = interface.create_interface() demo.launch( share=is_space ) if __name__ == "__main__": main()