Update app.py

app.py

@@ -1,92 +1,195 @@
-# app.py
-
 import gradio as gr
-from transformers import AutoTokenizer
-from groq import Groq
+import json
 import os
-from huggingface_hub import login
-import logging
-
-# Setup logging
-logging.basicConfig(level=logging.DEBUG)
-
-# Initialize Groq API client
-try:
-    client = Groq(api_key=os.environ["GROQ_API_KEY"])
-    logging.info("Groq API client initialized.")
-except KeyError:
-    raise ValueError("GROQ_API_KEY environment variable not set.")
-
-# Load the Hugging Face token from the environment variable
-hf_token = os.getenv('HF_TOKEN')
-if hf_token is None:
-    raise ValueError("Hugging Face token not found. Please set it as an environment variable.")
-
-# Authentication token for Hugging Face
-login(token=hf_token)
-
-# Model identifier for Groq API (you can replace it with your HF model if needed)
-model_name = "asthaa30/nomiChroma3.1"
-
-# Load tokenizer (model will be accessed via Groq API)
-try:
-    tokenizer = AutoTokenizer.from_pretrained(model_name)
-    logging.info(f"Tokenizer for model '{model_name}' loaded successfully.")
-except Exception as e:
-    raise ValueError(f"Failed to load tokenizer: {e}")
-
-def respond(
-    message: str,
-    history: list[tuple[str, str]],
-    system_message: str,
-    max_tokens: int,
-    temperature: float,
-    top_p: float,
-) -> str:
-    messages = [{"role": "system", "content": system_message}]
-
-    for user_msg, assistant_msg in history:
-        if user_msg:
-            messages.append({"role": "user", "content": user_msg})
-        if assistant_msg:
-            messages.append({"role": "assistant", "content": assistant_msg})
-
-    messages.append({"role": "user", "content": message})
-
-    # Use Groq API to get the model's response
+from groq import Groq
+from groq.types.chat.chat_completion_tool_param import ChatCompletionToolParam
+
+# Use the fine-tuned maritime legal model
+MODEL = "asthaa30/nomiChroma3.1"
+client = Groq(api_key=os.environ["GROQ_API_KEY"])
+
+# Define your tools if needed (e.g., legal research, document retrieval)
+def legal_tool_function(arguments):
+    # Implement specific legal functions here
+    # Placeholder for legal research or similar functionality
+    return {"result": "Legal tool function response here"}
+
+# Define your tools
+legal_tool: ChatCompletionToolParam = {
+    "type": "function",
+    "function": {
+        "name": "legal_tool_function",
+        "description": "Legal assistant tool: use this for various maritime legal tasks.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "arguments": {
+                    "type": "string",
+                    "description": "Arguments for the legal function.",
+                },
+            },
+            "required": ["arguments"],
+        },
+    },
+}
+
+tools = [legal_tool]
+
+def call_function(tool_call, available_functions):
+    function_name = tool_call.function.name
+    if function_name not in available_functions:
+        return {
+            "tool_call_id": tool_call.id,
+            "role": "tool",
+            "content": f"Function {function_name} does not exist.",
+        }
+    function_to_call = available_functions[function_name]
+    function_args = json.loads(tool_call.function.arguments)
+    function_response = function_to_call(**function_args)
+    return {
+        "tool_call_id": tool_call.id,
+        "role": "tool",
+        "name": function_name,
+        "content": json.dumps(function_response),
+    }
+
+def get_model_response(messages, inner_messages, message, system_message):
+    messages_for_model = []
+    for msg in messages:
+        native_messages = msg.get("metadata", {}).get("native_messages", [msg])
+        if isinstance(native_messages, list):
+            messages_for_model.extend(native_messages)
+        else:
+            messages_for_model.append(native_messages)
+
+    messages_for_model.insert(
+        0,
+        {
+            "role": "system",
+            "content": system_message,
+        },
+    )
+    messages_for_model.append(
+        {
+            "role": "user",
+            "content": message,
+        }
+    )
+    messages_for_model.extend(inner_messages)
+
     try:
-        response = client.chat.completions.create(
-            model=model_name,
-            messages=messages,
-            max_tokens=max_tokens,
-            temperature=temperature,
-            top_p=top_p,
+        return client.chat.completions.create(
+            model=MODEL,
+            messages=messages_for_model,
+            tools=tools,
+            temperature=0.5,
+            top_p=0.65,
+            max_tokens=4096,
         )
-        assistant_message = response.choices[0].message['content']
-        logging.info(f"Received response from model: {assistant_message}")
     except Exception as e:
-        logging.error(f"An error occurred while getting model response: {str(e)}")
-        assistant_message = "An error occurred. Please try again later."
+        print(f"An error occurred while getting model response: {str(e)}")
+        print(messages_for_model)
+        return None
+
+def respond(message, history, system_message):
+    inner_history = []
+
+    available_functions = {
+        "legal_tool_function": legal_tool_function,
+    }
+
+    assistant_content = ""
+    assistant_native_message_list = []
+
+    while True:
+        response_message = (
+            get_model_response(history, inner_history, message, system_message)
+            .choices[0]
+            .message
+        )
+
+        if not response_message.tool_calls and response_message.content is not None:
+            break
+
+        if response_message.tool_calls is not None:
+            assistant_native_message_list.append(response_message)
+            inner_history.append(response_message)
+
+            assistant_content += (
+                "```json\n"
+                + json.dumps(
+                    [
+                        tool_call.model_dump()
+                        for tool_call in response_message.tool_calls
+                    ],
+                    indent=2,
+                )
+                + "\n```\n"
+            )
+            assistant_message = {
+                "role": "assistant",
+                "content": assistant_content,
+                "metadata": {"native_messages": assistant_native_message_list},
+            }
+
+            yield assistant_message
+
+            for tool_call in response_message.tool_calls:
+                function_response = call_function(tool_call, available_functions)
+                assistant_content += (
+                    "```json\n"
+                    + json.dumps(
+                        {
+                            "name": tool_call.function.name,
+                            "arguments": json.loads(tool_call.function.arguments),
+                            "response": json.loads(function_response["content"]),
+                        },
+                        indent=2,
+                    )
+                    + "\n```\n"
+                )
+                native_tool_message = {
+                    "tool_call_id": tool_call.id,
+                    "role": "tool",
+                    "content": function_response["content"],
+                }
+                assistant_native_message_list.append(
+                    native_tool_message
+                )
+                tool_message = {
+                    "role": "assistant",
+                    "content": assistant_content,
+                    "metadata": {"native_messages": assistant_native_message_list},
+                }
+                yield tool_message
+                inner_history.append(native_tool_message)
+
+    assistant_content += response_message.content
+    assistant_native_message_list.append(response_message)
+
+    final_message = {
+        "role": "assistant",
+        "content": assistant_content,
+        "metadata": {"native_messages": assistant_native_message_list},
+    }
+    
+    yield final_message
 
-    return assistant_message
+# Update the system prompt to be more relevant to maritime legal assistance
+system_prompt = "You are a maritime legal assistant with expertise in maritime law. Provide detailed legal advice and information based on maritime legal principles and regulations."
 
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
+        gr.Textbox(
+            value=system_prompt,
+            label="System message",
         ),
     ],
-    title="Maritime Legal Assistant",
-    description="This chatbot provides legal assistance related to maritime laws using a fine-tuned model hosted on Hugging Face and integrated with Groq API.",
+    type="messages",
+    title="Maritime Legal Assistant Chat",
+    description="This chatbot uses the fine-tuned maritime legal model to provide legal assistance and information related to maritime law.",
 )
 
 if __name__ == "__main__":
-    demo.launch(share=False)  # Set share to False or remove it
+    demo.launch()