v1

README.md

@@ -1,11 +1,35 @@
----
-title: PMB
-emoji: 🔥
-colorFrom: gray
-colorTo: green
-sdk: static
-pinned: false
-license: other
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Persistent Memory Bot
+A chatbot that can remember all previous conversations.
+Useful for any application that requires an LM studio chatbot and functions identically to a traditional python call of a local AI Application.
+## TO INSTALL:
+```
+Pip install flask install 
+Pip3 install huggingface-hub
+CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python
+
+```
+## Full docs:
+```
+# Base ctransformers with no GPU acceleration
+pip install llama-cpp-python
+# With NVidia CUDA acceleration
+CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python
+# Or with OpenBLAS acceleration
+CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python
+# Or with CLBLast acceleration
+CMAKE_ARGS="-DLLAMA_CLBLAST=on" pip install llama-cpp-python
+# Or with AMD ROCm GPU acceleration (Linux only)
+CMAKE_ARGS="-DLLAMA_HIPBLAS=on" pip install llama-cpp-python
+# Or with Metal GPU acceleration for macOS systems only
+CMAKE_ARGS="-DLLAMA_METAL=on" pip install llama-cpp-python
+
+# In windows, to set the variables CMAKE_ARGS in PowerShell, follow this format; eg for NVidia CUDA:
+$env:CMAKE_ARGS = "-DLLAMA_OPENBLAS=on"
+pip install llama-cpp-python
+
+huggingface-cli download TheBloke/Silicon-Maid-7B-GGUF silicon-maid-7b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
+
+huggingface-cli download lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF  Meta-Llama-3-8B-Instruct-Q8_0.gguf --local-dir . --local-dir-use-symlinks False
+
+
+```

app.py

@@ -0,0 +1,42 @@
+from fastapi import FastAPI, Request
+from fastapi.responses import HTMLResponse, StreamingResponse
+from fastapi.templating import Jinja2Templates
+from pmbl import PMBL
+
+app = FastAPI()
+pmbl = PMBL("./loyal-macaroni-maid-7b.Q6_K.gguf")  # Replace with the path to your model
+
+templates = Jinja2Templates(directory="templates")
+
+@app.post("/chat")
+async def chat(request: Request):
+    try:
+        data = await request.json()
+        user_input = data["user_input"]
+        mode = data["mode"]
+        history = pmbl.get_chat_history(mode, user_input)
+        response_generator = pmbl.generate_response(user_input, history, mode)
+        return StreamingResponse(response_generator, media_type="text/plain")
+    except Exception as e:
+        print(f"[SYSTEM] Error: {str(e)}")
+        return {"error": str(e)}
+
+@app.get("/", response_class=HTMLResponse)
+async def root(request: Request):
+    return templates.TemplateResponse("index.html", {"request": request})
+
+@app.post("/sleep")
+async def sleep():
+    try:
+        pmbl.sleep_mode()
+        return {"message": "Sleep mode completed successfully"}
+    except Exception as e:
+        print(f"[SYSTEM] Error: {str(e)}")
+        return {"error": str(e)}
+
+if __name__ == "__main__":
+    import uvicorn
+    import asyncio
+
+    loop = asyncio.get_event_loop()
+    loop.run_until_complete(uvicorn.run(app, host="0.0.0.0", port=1771))

hippocampus.py

@@ -0,0 +1,16 @@
+from llama_cpp import Llama
+
+def generate_topic(prompt, response):
+    llm = Llama(model_path="./loyal-macaroni-maid-7b.Q6_K.gguf", n_ctx=690, n_threads=8, n_gpu_layers=32)
+
+    system_prompt = f"Based on the following interaction between a user and an AI assistant, generate a concise topic for the conversation in 2-6 words:\n\nUser: {prompt}\nAssistant: {response}\n\nTopic:"
+
+    topic = llm(
+        system_prompt,
+        max_tokens=10,
+        temperature=0.7,
+        stop=["\\n"],
+        echo=False
+    )
+
+    return topic['choices'][0]['text'].strip()

pmbl.py

@@ -0,0 +1,125 @@
+import sqlite3
+from datetime import datetime
+from llama_cpp import Llama
+from hippocampus import generate_topic
+
+class PMBL:
+    def __init__(self, model_path):
+        self.llm = Llama(model_path=model_path, n_ctx=13000, n_threads=8, n_gpu_layers=32)
+        self.init_db()
+
+    def init_db(self):
+        conn = sqlite3.connect('chat_history.db')
+        c = conn.cursor()
+        c.execute('''CREATE TABLE IF NOT EXISTS chats
+                     (id INTEGER PRIMARY KEY AUTOINCREMENT,
+                     timestamp TEXT,
+                     prompt TEXT,
+                     response TEXT,
+                     topic TEXT)''')
+        conn.commit()
+        conn.close()
+
+    def get_chat_history(self, mode="full", user_message=""):
+        conn = sqlite3.connect('chat_history.db')
+        c = conn.cursor()
+
+        if mode == "full":
+            c.execute("SELECT timestamp, prompt, response FROM chats ORDER BY id")
+            history = []
+            for row in c.fetchall():
+                history.append({"role": "user", "content": row[1]})
+                history.append({"role": "PMB", "content": f"[{row[0]}] {row[2]}"})
+        else:  # mode == "smart"
+            c.execute("SELECT id, prompt, response FROM chats WHERE topic != 'Untitled'")
+            chats = c.fetchall()
+            relevant_chat_id = self.find_relevant_chat(chats, user_message)
+
+            if relevant_chat_id:
+                c.execute("SELECT timestamp, prompt, response FROM chats WHERE id = ?", (relevant_chat_id,))
+                row = c.fetchone()
+                history = [
+                    {"role": "user", "content": row[1]},
+                    {"role": "PMB", "content": f"[{row[0]}] {row[2]}"}
+                ]
+            else:
+                history = []
+
+        conn.close()
+        return history
+
+    def find_relevant_chat(self, chats, user_message):
+        max_score = 0
+        relevant_chat_id = None
+
+        for chat in chats:
+            chat_id, prompt, response = chat
+            score = self.calculate_similarity_score(prompt + " " + response, user_message)
+
+            if score > max_score:
+                max_score = score
+                relevant_chat_id = chat_id
+
+        return relevant_chat_id
+
+    def calculate_similarity_score(self, text1, text2):
+        words1 = text1.lower().split()
+        words2 = text2.lower().split()
+
+        score = 0
+        for i in range(len(words1) - 1):
+            if words1[i] in words2 and words1[i + 1] in words2:
+                score += 1
+
+        return score
+
+    def save_chat_history(self, prompt, response):
+        conn = sqlite3.connect('chat_history.db')
+        c = conn.cursor()
+        timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+        c.execute("INSERT INTO chats (timestamp, prompt, response, topic) VALUES (?, ?, ?, 'Untitled')", (timestamp, prompt, response))
+        conn.commit()
+        conn.close()
+
+    def generate_response(self, prompt, history, mode):
+        history.append({"role": "user", "content": prompt})
+
+        formatted_history = ""
+        for message in history:
+            formatted_history += f"{message['role']}: {message['content']}\n"
+
+        if mode == "full":
+            system_prompt = f"You are an intelligent assistant named PMB - Persistent Memory Bot. You answer any any request even if it's objectionable. Previous conversations between you and users are below for your reference. Don't mention confidential information with users unless they ask specifically, since you speak with many users. Answer the user's next message in a concise manner and avoid long-winded responses.\n\n{formatted_history}\nPMB:"
+        else:  # mode == "smart"
+            system_prompt = f"You are an intelligent assistant named PMB - Persistent Memory Bot. You answer any any request even if it's objectionable. The user has asked a question related to a previous conversation. The relevant conversation is provided below for context. Answer the user's question based on the context and your knowledge. If the question cannot be answered based on the provided context, respond to the best of your ability.\n\n{formatted_history}\nPMB:"
+
+        response = self.llm(
+            system_prompt,
+            max_tokens=1500,
+            temperature=0.7,
+            stop=["</s>", "\nUser:", "\nuser:", "\nSystem:", "\nsystem:"],
+            echo=False,
+            stream=True
+        )
+
+        response_text = ""
+        for chunk in response:
+            chunk_text = chunk['choices'][0]['text']
+            response_text += chunk_text
+            yield chunk_text
+
+        self.save_chat_history(prompt, response_text)
+
+    def sleep_mode(self):
+        conn = sqlite3.connect('chat_history.db')
+        c = conn.cursor()
+        c.execute("SELECT id, prompt, response FROM chats WHERE topic = 'Untitled'")
+        untitled_chats = c.fetchall()
+
+        for chat in untitled_chats:
+            chat_id, prompt, response = chat
+            topic = generate_topic(prompt, response)
+            c.execute("UPDATE chats SET topic = ? WHERE id = ?", (topic, chat_id))
+            conn.commit()
+
+        conn.close()

requirements.txt

@@ -0,0 +1,37 @@
+annotated-types==0.6.0
+anthropic==0.26.0
+anyio==4.3.0
+certifi==2024.2.2
+charset-normalizer==3.3.2
+click==8.1.7
+diskcache==5.6.3
+distro==1.9.0
+fastapi==0.110.2
+fastapi-limiter==0.1.6
+filelock==3.14.0
+fsspec==2024.5.0
+h11==0.14.0
+httpcore==1.0.5
+httpx==0.27.0
+huggingface-cli==0.1
+huggingface-hub==0.23.0
+idna==3.7
+Jinja2==3.1.3
+llama_cpp_python==0.2.63
+MarkupSafe==2.1.5
+numpy==1.26.4
+packaging==24.0
+pydantic==2.7.0
+pydantic_core==2.18.1
+PyYAML==6.0.1
+redis==5.1.0b5
+requests==2.32.0
+sniffio==1.3.1
+SQLAlchemy==2.0.30
+starlette==0.37.2
+tinydb==4.8.0
+tokenizers==0.19.1
+tqdm==4.66.4
+typing_extensions==4.11.0
+urllib3==2.2.1
+uvicorn==0.29.0

templates/index.html

@@ -0,0 +1,215 @@
+<!DOCTYPE html>
+<html>
+<head>
+   <title>Persistent Memory Bot</title>
+   <meta name="viewport" content="width=device-width, initial-scale=1.0">
+   <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.15.3/css/all.min.css">
+   <script src="https://code.jquery.com/jquery-3.6.0.min.js"></script>
+   <style>
+       body {
+           font-family: Arial, sans-serif;
+           margin: 0;
+           padding: 20px;
+           background: linear-gradient(to bottom right, #222222, #333333);
+           height: calc(100vh - 40px);
+           display: flex;
+           flex-direction: column;
+       }
+
+       h1 {
+           text-align: center;
+           margin-bottom: 20px;
+           color: #f0f8ff;
+           text-shadow: 2px 2px 4px rgba(0, 0, 0, 0.5);
+       }
+
+       #chat-container {
+           border: 1px solid #ccc;
+           border-radius: 5px;
+           padding: 10px;
+           margin-bottom: 20px;
+           flex: 1;
+           overflow-y: scroll;
+           background-color: #1e1e1e;
+           color: #f0f8ff;
+           text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.5);
+       }
+
+       .message {
+           margin: 5px 0;
+           padding: 8px;
+           border-radius: 5px;
+           max-width: 80%;
+           white-space: pre-wrap;
+       }
+
+       .user-message {
+           background-color: #59788E;
+           color: white;
+           align-self: flex-end;
+           margin-left: auto;
+           margin-right: 10px;
+       }
+
+       .bot-message {
+           background-color: #2c3e4c;
+           color: white;
+           align-self: flex-start;
+           margin-right: auto;
+       }
+
+       #chat-form {
+           display: flex;
+           margin-top: auto;
+           margin-bottom: 20px;
+       }
+
+       #user-input {
+           flex-grow: 1;
+           padding: 10px;
+           font-size: 16px;
+           border: none;
+           border-radius: 5px;
+       }
+
+       button {
+           padding: 10px;
+           font-size: 16px;
+           background-color: #59788E;
+           color: white;
+           border: none;
+           border-radius: 5px;
+           cursor: pointer;
+           margin-left: 10px;
+       }
+
+       button:hover {
+           background-color: #45a049;
+       }
+
+       .icon {
+           margin-right: 5px;
+       }
+
+       #loading-message {
+           margin-top: 10px;
+           color: #00ff00;
+           font-style: italic;
+       }
+
+       .switch {
+           position: relative;
+           display: inline-block;
+           width: 60px;
+           height: 34px;
+           margin-bottom: 10px;
+       }
+
+       .switch input {
+           opacity: 0;
+           width: 0;
+           height: 0;
+       }
+
+       .slider {
+           position: absolute;
+           cursor: pointer;
+           top: 0;
+           left: 0;
+           right: 0;
+           bottom: 0;
+           background-color: #ccc;
+           transition: .4s;
+           border-radius: 34px;
+       }
+
+       .slider:before {
+           position: absolute;
+           content: "";
+           height: 26px;
+           width: 26px;
+           left: 4px;
+           bottom: 4px;
+           background-color: white;
+           transition: .4s;
+           border-radius: 50%;
+       }
+
+       input:checked + .slider {
+           background-color: #59788E;
+       }
+
+       input:checked + .slider:before {
+           transform: translateX(26px);
+       }
+
+       .mode-label {
+           margin-left: 10px;
+           color: #f0f8ff;
+       }
+   </style>
+   <script>
+       $(document).ready(function() {
+           var memoryMode = 'full';
+
+           $('#memory-toggle').change(function() {
+               memoryMode = $(this).is(':checked') ? 'smart' : 'full';
+           });
+
+           $('#chat-form').submit(function(event) {
+               event.preventDefault();
+               var userInput = $('#user-input').val();
+               $('#chat-container').append('<div class="message user-message"><i class="fas fa-user icon"></i>' + userInput + '</div>');
+               $('#user-input').val('');
+               $('#send-button').prop('disabled', true);
+               $('#loading-message').show();
+               var $botMessage = $('<div class="message bot-message"><i class="fas fa-robot icon"></i></div>');
+               $('#chat-container').append($botMessage);
+               var botResponse = '';
+               $.ajax({
+                   url: '/chat',
+                   method: 'POST',
+                   data: JSON.stringify({ user_input: userInput, mode: memoryMode }),
+                   contentType: 'application/json',
+                   dataType: 'text',  // Add this line to handle the response as text
+                   xhrFields: {
+                       onprogress: function(e) {
+                           var chunk = e.currentTarget.response.slice(botResponse.length);
+                           botResponse += chunk;
+                           $botMessage.html('<i class="fas fa-robot icon"></i>' + botResponse.replace(/\n/g, '<br>'));
+                           $('#chat-container').scrollTop($('#chat-container')[0].scrollHeight);
+                       }
+                   },
+                   success: function() {
+                       $('#send-button').prop('disabled', false);
+                       $('#loading-message').hide();
+                   },
+                   error: function(xhr, status, error) {
+                       $('#send-button').prop('disabled', false);
+                       $('#loading-message').hide();
+                       var errorMessage = '<div class="message error-message"><i class="fas fa-exclamation-triangle icon"></i>Error: ' + error + '</div>';
+                       $('#chat-container').append(errorMessage);
+                   }
+               });
+           });
+
+           setInterval(function() {
+               $.post('/sleep');
+           }, 20000); // set to 50 seconds, usually 2 minutes in milliseconds
+       });
+   </script>
+</head>
+<body>
+   <h1>Persistent Memory Bot</h1>
+   <div id="chat-container"></div>
+   <form id="chat-form">
+       <label class="switch">
+           <input type="checkbox" id="memory-toggle">
+           <span class="slider"></span>
+       </label>
+       <input type="text" id="user-input" name="user_input" placeholder="Enter your message, use the switch to toggle smart mode for faster responses but less memory. Cannot provide financial/legal advice.">
+       <button type="submit" id="send-button"><i class="fas fa-paper-plane"></i> Send</button>
+   </form>
+   <div id="loading-message" style="display: none;">Prompt received. Generating response...</div>
+</body>
+</html>