.gitignore

@@ -1,4 +1,5 @@
 venv/
-__pycache__
+__pycache__/
 .idea/
 .vscode/
+token.txt

py/api.py

@@ -1,27 +1,52 @@
 import requests
 import json
-from transformers import AutoTokenizer, LlamaForCausalLM
+from gradio_client import Client
-
+import os
+from mistralai import Mistral
 
 class API:
-    # This method processes a message via transformers. (NOT FINISHED!)
     @staticmethod
-    def process_text_transformers(prompt, model):
+    def process_text_mistralai(prompt, model, system):
-        model = LlamaForCausalLM.from_pretrained(model)
+        with open("token.txt", "r") as f:
-        tokenizer = AutoTokenizer.from_pretrained(model)
+            token = f.readlines()[0].strip()
 
-        inputs = tokenizer(prompt, return_tensors="pt")
+        api_key = token
 
-        generate_ids = model.generate(inputs.input_ids, max_length=30)
+        client = Mistral(api_key=api_key)
-        return tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+
+        chat_response = client.chat.complete(
+            model=model,
+            messages=[
+                {
+                    "role": "user",
+                    "content": prompt,
+                }, {
+                    "role": "system",
+                    "content": system,
+                },
+            ]
+        )
+        return chat_response.choices[0].message.content
+    @staticmethod
+    def process_text_gradio(prompt, model, system):
+        client = Client(model)
+        result = client.predict(
+            message=prompt,
+            system_message=system,
+            max_tokens=512,
+            temperature=0.7,
+            top_p=0.95,
+            api_name="/chat"
+        )
+        return result;
 
     # This method processes a message via ollama
     @staticmethod
-    def process_text_local(prompt, model):
+    def process_text_local(prompt, model, system):
         ollama_url = "http://localhost:11434"
 
         response = requests.post(
-            f"{ollama_url}/api/generate", json={"model": model, "prompt": prompt}
+            f"{ollama_url}/api/generate", json={"model": model, "prompt": prompt, "system": system}
         )
 
         if response.status_code == 200:
@@ -37,17 +62,20 @@ class API:
             return "Error: " + response.text
 
     # This method sends a message to a certain AI.
-    def send_message(self, message, model):
+
+    def send_message(self, message, model, system):
         if model == 1:
-            answer = self.process_text_local(message, "phi3.5")
+            answer = self.process_text_local(message, "phi3.5", system)
         elif model == 2:
-            answer = self.process_text_local(message, "gemma2:2b")
+            answer = self.process_text_local(message, "gemma2:9b", system)
         elif model == 3:
-            answer = self.process_text_local(message, "qwen2:0.5b")
+            answer = self.process_text_local(message, "codegemma:2b", system)
         elif model == 4:
-            answer = self.process_text_local(message, "codegemma:2b")
+            answer = self.process_text_gradio(message, "PatrickPluto/InterstellarAIChatbot", system)
         elif model == 5:
-            answer = self.process_text_transformers(message, "meta-llama/Meta-Llama-3.1-8B")
+            answer = self.process_text_mistralai(message, "mistral-large-latest", system)
+        elif model == 6:
+            answer = self.process_text_mistralai(message, "codestral-latest", system)
         else:
             return "Invalid choice"
         return answer

py/install.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+
+python3 -m venv venv
+source venv/bin/activate
+pip install flask
+pip install SpeechRecognition
+pip install pyaudio
+pip install pocketsphinx
+pip install sentencepiece
+pip install pyqt5
+pip install pyqtwebengine
+pip install gradio_client
+pip install mistralai
+
+ollama pull phi3.5
+ollama pull codegemma:2b
+ollama pull gemma2:9b

py/venv.sh

@@ -1,10 +0,0 @@
-#!/bin/bash
-
-virtualenv venv
-source venv/bin/activate
-pip install transformers
-pip install torch
-pip install flask
-pip install SpeechRecognition
-pip install pyaudio
-pip install pocketsphinx

py/web_flask.py

@@ -1,6 +1,14 @@
+#!venv/bin/python
+
 from flask import Flask, request, render_template
 from api import API
 from voice_recognition import Voice
+import sys
+import threading
+from PyQt5.QtCore import *
+from PyQt5.QtWebEngineWidgets import *
+from PyQt5.QtWidgets import *
+
 
 APP = Flask(__name__)
 api = API()
@@ -13,6 +21,8 @@ messages = []
 @APP.route('/', methods=['GET', 'POST'])
 def index():
     global messages
+    system_prompt = 'You are a helpful assistant.'
+    system = 'Your system prompt is: \"'+system_prompt+'\" The following is your chat log so far: \n'
 
     if request.method == 'POST':
         option = request.form['option']
@@ -25,7 +35,12 @@ def index():
         elif option == "chat":
             messages.append(f"User: {user_message}")
 
-        ai_response = "AI: " + api.send_message(user_message, 3)
+        for line in messages:
+            system += line + '\n'
+
+        system += "The chat log is now finished."
+
+        ai_response = "AI: " + api.send_message(user_message, 5, system)
         messages.append(ai_response)
 
     return render_template('index.html', messages=messages)
@@ -42,6 +57,28 @@ def contact():
 
 
 if __name__ == '__main__':
-    APP.run(debug=True)
+    qapp = QApplication(sys.argv)
 
-# This is a comment --> test if this creates a merge conflict
+    view = QWebEngineView()
+
+    view.setGeometry(100, 100, 1280, 720)
+    view.setWindowTitle("InterstellarAI")
+
+    view.setUrl(QUrl("http://localhost:5000"))
+
+    view.show()
+
+    def run_flask():
+        APP.run()
+
+    def stop_flask():
+        thread.join()
+        qapp.quit()
+
+    thread = threading.Thread(target=run_flask)
+    thread.daemon = True
+    thread.start()
+
+    qapp.aboutToQuit.connect(stop_flask)
+
+    sys.exit(qapp.exec_())