A.I.StateMachine

Running

App Files Files Community

Chris4K commited on Jan 31

Commit

a9243a6

verified ·

1 Parent(s): 38dc5a0

Update app.py

Browse files

Files changed (1) hide show

app.py +174 -281

app.py CHANGED Viewed

@@ -18,293 +18,186 @@ except Exception as e:
-###################
-#################
-from fastapi import FastAPI, HTTPException, BackgroundTasks, WebSocket, Depends
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import StreamingResponse
-from pydantic import BaseModel, Field, ConfigDict
-from typing import List, Optional, Dict, Any, AsyncGenerator
-import asyncio
-import uuid
-from datetime import datetime
-import json
-from huggingface_hub import hf_hub_download
-from contextlib import asynccontextmanager
-class ChatMessage(BaseModel):
-    """A single message in the chat history."""
-    role: str = Field(
-        ...,
-        description="Role of the message sender",
-        examples=["user", "assistant"]
-    )
-    content: str = Field(..., description="Content of the message")
-    model_config = ConfigDict(
-        json_schema_extra={
-            "example": {
-                "role": "user",
-                "content": "What is the capital of France?"
             }
         }
-    )
-class GenerationConfig(BaseModel):
-    """Configuration for text generation."""
-    temperature: float = Field(
-        0.7,
-        ge=0.0,
-        le=2.0,
-        description="Controls randomness in the output. Higher values (e.g., 0.8) make the output more random, lower values (e.g., 0.2) make it more focused and deterministic."
-    )
-    max_new_tokens: int = Field(
-        100,
-        ge=1,
-        le=2048,
-        description="Maximum number of tokens to generate"
-    )
-    top_p: float = Field(
-        0.9,
-        ge=0.0,
-        le=1.0,
-        description="Nucleus sampling parameter. Only tokens with cumulative probability < top_p are considered."
-    )
-    top_k: int = Field(
-        50,
-        ge=0,
-        description="Only consider the top k tokens for text generation"
-    )
-    strategy: str = Field(
-        "default",
-        description="Generation strategy to use",
-        examples=["default", "majority_voting", "best_of_n", "beam_search", "dvts"]
-    )
-    num_samples: int = Field(
-        5,
-        ge=1,
-        le=10,
-        description="Number of samples to generate (used in majority_voting and best_of_n strategies)"
-    )
-class GenerationRequest(BaseModel):
-    """Request model for text generation."""
-    context: Optional[str] = Field(
-        None,
-        description="Additional context to guide the generation",
-        examples=["You are a helpful assistant skilled in Python programming"]
-    )
-    messages: List[ChatMessage] = Field(
-        ...,
-        description="Chat history including the current message",
-        min_items=1
-    )
-    config: Optional[GenerationConfig] = Field(
-        None,
-        description="Generation configuration parameters"
-    )
-    stream: bool = Field(
-        False,
-        description="Whether to stream the response token by token"
-    )
-    model_config = ConfigDict(
-        json_schema_extra={
-            "example": {
-                "context": "You are a helpful assistant",
-                "messages": [
-                    {"role": "user", "content": "What is the capital of France?"}
-                ],
-                "config": {
-                    "temperature": 0.7,
-                    "max_new_tokens": 100
-                },
-                "stream": False
-            }
-        }
-    )
-class GenerationResponse(BaseModel):
-    """Response model for text generation."""
-    id: str = Field(..., description="Unique generation ID")
-    content: str = Field(..., description="Generated text content")
-    created_at: datetime = Field(
-        default_factory=datetime.now,
-        description="Timestamp of generation"
-    )
-# Model and cache management
-async def get_prm_model_path():
-    """Download and cache the PRM model."""
-    return await asyncio.to_thread(
-        hf_hub_download,
-        repo_id="tensorblock/Llama3.1-8B-PRM-Mistral-Data-GGUF",
-        filename="Llama3.1-8B-PRM-Mistral-Data-Q4_K_M.gguf"
-    )
-# Initialize generator globally
-generator = None
-@asynccontextmanager
-async def lifespan(app: FastAPI):
-    """Lifecycle management for the FastAPI application."""
-    # Startup: Initialize generator
-    global generator
-    try:
-        prm_model_path = await get_prm_model_path()
-        generator = LlamaGenerator(
-            llama_model_name="meta-llama/Llama-3.2-1B-Instruct",
-            prm_model_path=prm_model_path,
-            default_generation_config=GenerationConfig(
-                max_new_tokens=100,
-                temperature=0.7
-            )
-        )
-        yield
-    finally:
-        # Shutdown: Clean up resources
-        if generator:
-            await asyncio.to_thread(generator.cleanup)
-# FastAPI application
-app = FastAPI(
-    title="Inference Deluxe Service",
-    description="""
-    A service for generating text using LLaMA models with various generation strategies.
-    Generation Strategies:
-    - default: Standard autoregressive generation
-    - majority_voting: Generates multiple responses and selects the most common one
-    - best_of_n: Generates multiple responses and selects the best based on a scoring metric
-    - beam_search: Uses beam search for more coherent text generation
-    - dvts: Dynamic vocabulary tree search for efficient generation
-    """,
-    version="1.0.0",
-    lifespan=lifespan
-)
-# CORS middleware
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-async def get_generator():
-    """Dependency to get the generator instance."""
-    if not generator:
-        raise HTTPException(
-            status_code=503,
-            detail="Generator not initialized"
-        )
-    return generator
-@app.post(
-    "/generate",
-    response_model=GenerationResponse,
-    tags=["generation"],
-    summary="Generate text response",
-    response_description="Generated text with unique identifier"
-)
-async def generate(
-    request: GenerationRequest,
-    generator: Any = Depends(get_generator)
-):
-    """
-    Generate a text response based on the provided context and chat history.
-    """
-    try:
-        chat_history = [(msg.role, msg.content) for msg in request.messages[:-1]]
-        user_input = request.messages[-1].content
-        # Extract or set defaults for additional arguments
-        config = request.config or GenerationConfig()
-        model_kwargs = {
-            "temperature": config.temperature if hasattr(config, "temperature") else 0.7,
-            "max_new_tokens": config.max_new_tokens if hasattr(config, "max_new_tokens") else 100,
-            # Add other model kwargs as needed
-        }
-        # Explicitly pass additional required arguments
-        response = await asyncio.to_thread(
-            generator.generate_with_context,
-            context=request.context or "",
-            user_input=user_input,
-            chat_history=chat_history,
-            model_kwargs=model_kwargs,
-            max_history_turns=config.max_history_turns if hasattr(config, "max_history_turns") else 3,
-            strategy=config.strategy if hasattr(config, "strategy") else "default",
-            num_samples=config.num_samples if hasattr(config, "num_samples") else 5,
-            depth=config.depth if hasattr(config, "depth") else 3,
-            breadth=config.breadth if hasattr(config, "breadth") else 2,
-        )
-        return GenerationResponse(
-            id=str(uuid.uuid4()),
-            content=response
-        )
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
-@app.websocket("/generate/stream")
-async def generate_stream(
-    websocket: WebSocket,
-    generator: Any = Depends(get_generator)
-):
-    """
-    Stream generated text tokens over a WebSocket connection.
-    The stream sends JSON messages with the following format:
-    - During generation: {"token": "generated_token", "finished": false}
-    - End of generation: {"token": "", "finished": true}
-    - Error: {"error": "error_message"}
-    """
     await websocket.accept()
     try:
         while True:
-            request_data = await websocket.receive_text()
-            request = GenerationRequest.parse_raw(request_data)
-            chat_history = [(msg.role, msg.content) for msg in request.messages[:-1]]
-            user_input = request.messages[-1].content
-            config = request.config or GenerationConfig()
-            async for token in generator.generate_stream(
-                prompt=generator.prompt_builder.format(
-                    context=request.context or "",
-                    user_input=user_input,
-                    chat_history=chat_history
-                ),
-                config=config
-            ):
-                await websocket.send_text(json.dumps({
-                    "token": token,
-                    "finished": False
-                }))
-            await websocket.send_text(json.dumps({
-                "token": "",
-                "finished": True
-            }))
-    except Exception as e:
-        await websocket.send_text(json.dumps({
-            "error": str(e)
-        }))
-    finally:
-        await websocket.close()
-if __name__ == "__main__":
-    import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=8000)

+########
+html = """
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>AI State Machine</title>
+    <style>
+        body { font-family: Arial, sans-serif; text-align: center; }
+        #chatbox { width: 80%; height: 300px; border: 1px solid #ccc; overflow-y: auto; margin: 20px auto; padding: 10px; }
+        #inputbox { width: 70%; padding: 5px; }
+        button { padding: 5px 10px; }
+    </style>
+</head>
+<body>
+    <h2>AI State Machine</h2>
+    <div id="chatbox"></div>
+    <input type="text" id="inputbox" placeholder="Type your message...">
+    <button onclick="sendMessage()">Send</button>
+    <script>
+        let ws = new WebSocket("ws://localhost:8000/ws");
+        ws.onmessage = (event) => {
+            let chatbox = document.getElementById("chatbox");
+            chatbox.innerHTML += `<p>${event.data}</p>`;
+            chatbox.scrollTop = chatbox.scrollHeight;
+        };
+        function sendMessage() {
+            let input = document.getElementById("inputbox");
+            let message = input.value.trim();
+            if (message) {
+                ws.send(message);
+                input.value = "";
             }
         }
+    </script>
+</body>
+</html>
+"""
+######
+import asyncio
+import queue
+import threading
+import random
+import time
+from fastapi import FastAPI, WebSocket
+from fastapi.responses import HTMLResponse
+import uvicorn
+# FastAPI App
+app = FastAPI()
+class AIStateManager:
+    def __init__(self):
+        self.state = "awake"
+        self.msg_queue = queue.Queue()
+        self.heartbeat_count = 0
+        self.lock = threading.Lock()
+        self.clients = set()
+        # Research Task List
+        self.research_tasks = ["Explore AI Ethics", "Find latest AI models", "Investigate quantum computing"]
+        self.current_task = None
+    def set_state(self, new_state):
+        with self.lock:
+            print(f"[STATE CHANGE] {self.state} → {new_state}")
+            self.state = new_state
+            self.add_message("system", f"State changed to {new_state}")
+    def receive_message(self, sender, message):
+        """Adds messages to queue and resets heartbeat if human sends input."""
+        with self.lock:
+            self.msg_queue.put((sender, message))
+            if sender == "human":
+                self.heartbeat_count = 0
+                if self.state != "awake":
+                    self.set_state("awake")
+    def add_message(self, sender, message):
+        """AI or System can add messages to queue."""
+        with self.lock:
+            self.msg_queue.put((sender, message))
+    def process_messages(self):
+        """Processes messages in queue."""
+        while not self.msg_queue.empty():
+            sender, message = self.msg_queue.get()
+            print(f"[{sender.upper()}] {message}")
+            asyncio.create_task(self.broadcast(f"[{sender.upper()}] {message}"))
+    async def broadcast(self, message):
+        """Sends message to all connected WebSocket clients."""
+        for ws in self.clients:
+            await ws.send_text(message)
+    async def heartbeat(self):
+        """Basic heartbeat loop."""
+        while True:
+            await asyncio.sleep(1)  # One 'beat'
+            with self.lock:
+                self.heartbeat_count += 1
+                self.process_messages()
+    async def consciousness(self):
+        """Controls research/sleep cycle."""
+        while True:
+            await asyncio.sleep(2)  # Consciousness checks every 2 sec
+            with self.lock:
+                if self.state == "awake":
+                    if self.heartbeat_count >= 5:
+                        self.set_state("research")
+                        asyncio.create_task(self.run_research())
+                elif self.state == "research":
+                    if not self.research_tasks:  # If no tasks left, move to sleep
+                        self.set_state("sleeping")
+                        asyncio.create_task(self.run_sleeping())
+                elif self.state == "sleeping":
+                    if self.heartbeat_count >= 20:
+                        self.set_state("research")  # Restart research after sleep
+                        asyncio.create_task(self.run_research())
+    async def run_research(self):
+        """Runs research tasks in order."""
+        while self.state == "research" and self.research_tasks:
+            await asyncio.sleep(3)
+            self.current_task = self.research_tasks.pop(0)
+            self.add_message("ai", f"Researching: {self.current_task}")
+            if random.random() < 0.3:  # AI might ask a follow-up question
+                self.add_message("ai", f"Question: What do you think about {self.current_task}?")
+            self.process_messages()
+    async def run_sleeping(self):
+        """Runs self-training until state changes."""
+        while self.state == "sleeping":
+            await asyncio.sleep(5)
+            self.add_message("system", "Self-training in progress...")
+            self.process_messages()
+    def modify_research_tasks(self):
+        """Background process to edit research tasks dynamically (subconscious)."""
+        while True:
+            time.sleep(10)  # Runs every 10 seconds
+            with self.lock:
+                if self.state == "research" and random.random() < 0.5:
+                    new_task = f"Investigate {random.choice(['AI Bias', 'Neural Networks', 'Data Privacy'])}"
+                    self.research_tasks.append(new_task)
+                    self.add_message("system", f"New research task added: {new_task}")
+# Initialize AI Manager
+ai_manager = AIStateManager()
+# Start Heartbeat and Consciousness in separate threads
+threading.Thread(target=lambda: asyncio.run(ai_manager.heartbeat()), daemon=True).start()
+threading.Thread(target=lambda: asyncio.run(ai_manager.consciousness()), daemon=True).start()
+threading.Thread(target=ai_manager.modify_research_tasks, daemon=True).start()
+@app.websocket("/ws")
+async def websocket_endpoint(websocket: WebSocket):
+    """WebSocket connection handler."""
     await websocket.accept()
+    ai_manager.clients.add(websocket)
     try:
         while True:
+            data = await websocket.receive_text()
+            ai_manager.receive_message("human", data)
+    except Exception:
+        ai_manager.clients.remove(websocket)
+@app.get("/")
+async def get():
+    """Serve frontend HTML."""
+    return HTMLResponse(html)
+if __name__ == "__main__":
+    uvicorn.run(app, host="localhost", port=8000)