update

app.py

@@ -1,47 +1,65 @@
 import gradio as gr
-import asyncio
-from openai import AsyncOpenAI
+import ollama
+# import asyncio
+# from openai import AsyncOpenAI
 from RAGModule import RAGModule
 
 # Instantiate the RAG module
 RAG_Trwira = RAGModule()
 
-# Configure the async OpenAI client
-client = AsyncOpenAI(api_key="34.69.9.203", base_url="http://34.69.9.203:11434/v1")
+# # Configure the async OpenAI client
+# client = AsyncOpenAI(api_key="34.69.9.203", base_url="http://34.69.9.203:11434/v1")
 
-settings = {
-    "model": "MarcoAland/llama3.1-rag-indo",
-    "temperature": 0.3,
-    "max_tokens": 2048,
-}
+# settings = {
+#     "model": "MarcoAland/llama3.1-rag-indo",
+#     "temperature": 0.3,
+#     "max_tokens": 2048,
+# }
 
-async def generate_response(user_input: str) -> str:
-    message = "Namamu adalah Mitrakara.\n\n" + user_input
-    # Call documents options or not
+# async def generate_response(user_input: str) -> str:
+#     message = "Namamu adalah Mitrakara.\n\n" + user_input
+#     # Call documents options or not
+#     if "dokumen" in message.lower() or "document" in message.lower() or "documents" in message.lower():
+#         prompt = RAG_Trwira.main(message[10:])
+#     else:
+#         prompt = message
+        
+#     # Format the messages as a list of message dictionaries
+#     message_formated = [
+#         {"role": "user", "content": prompt}
+#     ]
+
+#     # Use streaming to handle partial responses
+#     stream = await client.chat.completions.create(messages=message_formated, stream=True, **settings)
+    
+#     response = ""
+#     async for part in stream:
+#         if token := part.choices[0].delta.content or "":
+#             response += token
+
+#     return response
+
+# def chat(user_input: str):
+#     # Call the asynchronous response generation function
+#     response = asyncio.run(generate_response(user_input))
+#     return response
+
+def chat(message: str, chat_history: str):
     if "dokumen" in message.lower() or "document" in message.lower() or "documents" in message.lower():
         prompt = RAG_Trwira.main(message[10:])
     else:
         prompt = message
-        
-    # Format the messages as a list of message dictionaries
-    message_formated = [
-        {"role": "user", "content": prompt}
-    ]
-
-    # Use streaming to handle partial responses
-    stream = await client.chat.completions.create(messages=message_formated, stream=True, **settings)
-    
-    response = ""
-    async for part in stream:
-        if token := part.choices[0].delta.content or "":
-            response += token
 
-    return response
+    stream = ollama.chat(
+                            model='MarcoAland/llama3.1-rag-indo',
+                            messages=[{'role': 'user', 'content': prompt}],
+                            stream=True,
+                        )
 
-def chat(user_input: str):
-    # Call the asynchronous response generation function
-    response = asyncio.run(generate_response(user_input))
-    return response
+    response_text = ''
+    for chunk in stream:
+        response_text += chunk['message']['content']
+        yield response_text
 
 # Define the Gradio interface
 iface = gr.Interface(

requirements.txt

@@ -1,5 +1,6 @@
 openai
 gradio
+ollama
 llama-cloud==0.0.13
 llama-index==0.10.64
 llama-index-embeddings-huggingface==0.2.3