Update app.py

app.py

@@ -1,177 +1,168 @@
+import os
 import gradio as gr
+
 from openai import OpenAI
-import os
-from datetime import datetime
-
-# App title and description
-APP_TITLE = "NO GPU, Multi LLMs Uses"
-APP_DESCRIPTION = "Access and chat with multiple language models without requiring a GPU"
-
-# Load environment variables
-ACCESS_TOKEN = os.getenv("HF_TOKEN")
-client = OpenAI(
-    base_url="https://api-inference.huggingface.co/v1/",
-    api_key=ACCESS_TOKEN,
-)
-
-# Model categories for better organization
-MODEL_CATEGORIES = {
-    "Qwen": [
-        "Qwen/Qwen2.5-72B-Instruct",
-        "Qwen/Qwen2.5-3B-Instruct",
-        "Qwen/Qwen2.5-0.5B-Instruct",
-        "Qwen/Qwen2.5-Coder-32B-Instruct",
-    ],
-    "Meta LLaMa": [
-        "meta-llama/Llama-3.3-70B-Instruct",
-        "meta-llama/Llama-3.1-70B-Instruct",
-        "meta-llama/Llama-3.0-70B-Instruct",
-        "meta-llama/Llama-3.2-3B-Instruct",
-        "meta-llama/Llama-3.2-1B-Instruct",
-        "meta-llama/Llama-3.1-8B-Instruct",
-    ],
-    "Mistral": [
-        "mistralai/Mistral-Nemo-Instruct-2407",
-        "mistralai/Mixtral-8x7B-Instruct-v0.1",
-        "mistralai/Mistral-7B-Instruct-v0.3",
-        "mistralai/Mistral-7B-Instruct-v0.2",
-    ],
-    "Microsoft Phi": [
-        "microsoft/Phi-3.5-mini-instruct",
-        "microsoft/Phi-3-mini-128k-instruct",
-        "microsoft/Phi-3-mini-4k-instruct",
-    ],
-    "Other Models": [
-
-        "NousResearch/Hermes-3-Llama-3.1-8B",
-        "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
-        "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
-        "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
-        "HuggingFaceH4/zephyr-7b-beta",
-        "HuggingFaceTB/SmolLM2-360M-Instruct",
-        "tiiuae/falcon-7b-instruct",
-        "01-ai/Yi-1.5-34B-Chat",
-    ]
-}
-
-# Flatten the model list
-ALL_MODELS = [m for models in MODEL_CATEGORIES.values() for m in models]
-
-def get_model_info(model_name):
-    parts = model_name.split('/')
-    if len(parts) != 2:
-        return f"**Model:** {model_name}\n**Format:** Unknown"
-    org, model = parts
-    import re
-    size_match = re.search(r'(\d+\.?\d*)B', model)
-    size = size_match.group(1) + "B" if size_match else "Unknown"
-    return f"**Organization:** {org}\n**Model:** {model}\n**Size:** {size}"
-
-def respond(
-    message,
-    history,
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    frequency_penalty,
-    seed,
-    selected_model
-):
-    # Prepare messages
-    if seed == -1:
-        seed = None
-    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})
-
-    model_to_use = selected_model or ALL_MODELS[0]
-
-    new_history = list(history) + [(message, "")]
-    current_response = ""
+
+from optillm.cot_reflection import cot_reflection
+from optillm.rto import round_trip_optimization
+from optillm.z3_solver import Z3SymPySolverSystem
+from optillm.self_consistency import advanced_self_consistency_approach
+from optillm.plansearch import plansearch
+from optillm.leap import leap
+from optillm.reread import re2_approach
+
+
+API_KEY = os.environ.get("OPENROUTER_API_KEY")
+
+def compare_responses(message, model1, approach1, model2, approach2, system_message, max_tokens, temperature, top_p):
+    response1 = respond(message, [], model1, approach1, system_message, max_tokens, temperature, top_p)
+    response2 = respond(message, [], model2, approach2, system_message, max_tokens, temperature, top_p)
+    return response1, response2
+
+def parse_conversation(messages):
+    system_prompt = ""
+    conversation = []
+    
+    for message in messages:
+        role = message['role']
+        content = message['content']
+        
+        if role == 'system':
+            system_prompt = content
+        elif role in ['user', 'assistant']:
+            conversation.append(f"{role.capitalize()}: {content}")
+    
+    initial_query = "\n".join(conversation)
+    return system_prompt, initial_query
+
+def respond(message, history, model, approach, system_message, max_tokens, temperature, top_p):
     try:
-        for chunk in client.chat.completions.create(
-            model=model_to_use,
-            max_tokens=max_tokens,
-            stream=True,
-            temperature=temperature,
-            top_p=top_p,
-            frequency_penalty=frequency_penalty,
-            seed=seed,
-            messages=messages,
-        ):
-            delta = chunk.choices[0].delta.content
-            if delta:
-                current_response += delta
-                new_history[-1] = (message, current_response)
-                yield new_history
+        client = OpenAI(api_key=API_KEY, base_url="https://openrouter.ai/api/v1")
+        messages = [{"role": "system", "content": system_message}]
+        
+        for val in history:
+            if val[0]:
+                messages.append({"role": "user", "content": val[0]})
+            if val[1]:
+                messages.append({"role": "assistant", "content": val[1]})
+        
+        messages.append({"role": "user", "content": message})
+        
+        if approach == "none":
+            response = client.chat.completions.create(
+                extra_headers={
+                    "HTTP-Referer": "https://github.com/codelion/optillm",
+                    "X-Title": "optillm"
+                },
+                model=model,
+                messages=messages,
+                max_tokens=max_tokens,
+                temperature=temperature,
+                top_p=top_p,
+            )
+            return response.choices[0].message.content
+        else:
+            system_prompt, initial_query = parse_conversation(messages)
+            
+            if approach == 'rto':
+                final_response, _ = round_trip_optimization(system_prompt, initial_query, client, model)
+            elif approach == 'z3':
+                z3_solver = Z3SymPySolverSystem(system_prompt, client, model)
+                final_response, _ = z3_solver.process_query(initial_query)
+            elif approach == "self_consistency":
+                final_response, _ = advanced_self_consistency_approach(system_prompt, initial_query, client, model)
+            elif approach == "cot_reflection":
+                final_response, _ = cot_reflection(system_prompt, initial_query, client, model)
+            elif approach == 'plansearch':
+                response, _ = plansearch(system_prompt, initial_query, client, model)
+                final_response = response[0]
+            elif approach == 'leap':
+                final_response, _ = leap(system_prompt, initial_query, client, model)
+            elif approach == 're2':
+                final_response, _ = re2_approach(system_prompt, initial_query, client, model)
+            
+            return final_response
+            
     except Exception as e:
-        err = f"Error: {e}"
-        new_history[-1] = (message, err)
-        yield new_history
-
-with gr.Blocks(title=APP_TITLE, theme=gr.themes.Soft()) as demo:
-    gr.Markdown(f"## {APP_TITLE}\n\n{APP_DESCRIPTION}")
-
+        error_message = f"Error in respond function: {str(e)}\nType: {type(e).__name__}"
+        print(error_message)
+
+def create_model_dropdown():
+    return gr.Dropdown(
+        [ "meta-llama/llama-3.1-8b-instruct:free", "nousresearch/hermes-3-llama-3.1-405b:free","meta-llama/llama-3.2-1b-instruct:free",
+         "mistralai/mistral-7b-instruct:free","mistralai/pixtral-12b:free","meta-llama/llama-3.1-70b-instruct:free",
+         "qwen/qwen-2-7b-instruct:free", "qwen/qwen-2-vl-7b-instruct:free", "google/gemma-2-9b-it:free", "liquid/lfm-40b:free", "meta-llama/llama-3.1-405b-instruct:free",
+         "openchat/openchat-7b:free", "meta-llama/llama-3.2-90b-vision-instruct:free", "meta-llama/llama-3.2-11b-vision-instruct:free",
+         "meta-llama/llama-3-8b-instruct:free", "meta-llama/llama-3.2-3b-instruct:free", "microsoft/phi-3-medium-128k-instruct:free",
+         "microsoft/phi-3-mini-128k-instruct:free", "huggingfaceh4/zephyr-7b-beta:free"],
+        value="meta-llama/llama-3.2-1b-instruct:free", label="Model"
+    )
+
+def create_approach_dropdown():
+    return gr.Dropdown(
+        ["none", "leap", "plansearch", "cot_reflection", "rto", "self_consistency", "z3", "re2"],
+        value="none", label="Approach"
+    )
+
+html = """<iframe src="https://ghbtns.com/github-btn.html?user=codelion&repo=optillm&type=star&count=true&size=large" frameborder="0" scrolling="0" width="170" height="30" title="GitHub"></iframe>
+"""
+
+with gr.Blocks() as demo:
+    gr.Markdown("# optillm - Optimizing LLM Inference")
+    gr.HTML(html)
+    
     with gr.Row():
-        with gr.Column(scale=2):
-            # Model selection via Dropdown
-            selected_model = gr.Dropdown(
-                choices=ALL_MODELS,
-                value=ALL_MODELS[0],
-                label="Select Model"
-            )
-            model_info = gr.Markdown(get_model_info(ALL_MODELS[0]))
-
-            def update_info(model_name):
-                return get_model_info(model_name)
-            selected_model.change(
-                fn=update_info,
-                inputs=[selected_model],
-                outputs=[model_info]
-            )
-
-            # Conversation settings
-            system_message = gr.Textbox(
-                value="You are a helpful assistant.",
-                label="System Prompt",
-                lines=2
-            )
-
-            max_tokens = gr.Slider(1, 4096, value=512, label="Max New Tokens")
-            temperature = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature")
-            top_p = gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-P")
-            freq_penalty = gr.Slider(-2.0, 2.0, value=0.0, step=0.1, label="Frequency Penalty")
-            seed = gr.Slider(-1, 65535, value=-1, step=1, label="Seed (-1 random)")
-
-        with gr.Column(scale=3):
+        system_message = gr.Textbox(value="", label="System message")
+        max_tokens = gr.Slider(minimum=1, maximum=4096, value=1024, step=1, label="Max new tokens")
+        temperature = gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature")
+        top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
+    
+    with gr.Tabs():
+        with gr.TabItem("Chat"):
+            model = create_model_dropdown()
+            approach = create_approach_dropdown()
             chatbot = gr.Chatbot()
-            msg = gr.Textbox(placeholder="Type your message here...", show_label=False)
-            send_btn = gr.Button("Send")
-
-            send_btn.click(
-                fn=respond,
-                inputs=[
-                    msg, chatbot, system_message,
-                    max_tokens, temperature, top_p,
-                    freq_penalty, seed, selected_model
-                ],
-                outputs=[chatbot],
-                queue=True
+            msg = gr.Textbox()
+            with gr.Row():
+                submit = gr.Button("Submit")
+                clear = gr.Button("Clear")
+
+            def user(user_message, history):
+                return "", history + [[user_message, None]]
+
+            def bot(history, model, approach, system_message, max_tokens, temperature, top_p):
+                user_message = history[-1][0]
+                bot_message = respond(user_message, history[:-1], model, approach, system_message, max_tokens, temperature, top_p)
+                history[-1][1] = bot_message
+                return history
+
+            msg.submit(user, [msg, chatbot], [msg, chatbot]).then(
+                bot, [chatbot, model, approach, system_message, max_tokens, temperature, top_p], chatbot
+            )
+            submit.click(user, [msg, chatbot], [msg, chatbot]).then(
+                bot, [chatbot, model, approach, system_message, max_tokens, temperature, top_p], chatbot
             )
-            msg.submit(
-                fn=respond,
-                inputs=[
-                    msg, chatbot, system_message,
-                    max_tokens, temperature, top_p,
-                    freq_penalty, seed, selected_model
-                ],
-                outputs=[chatbot],
-                queue=True
+            clear.click(lambda: None, None, chatbot, queue=False)
+
+        with gr.TabItem("Compare"):
+            with gr.Row():
+                model1 = create_model_dropdown()
+                approach1 = create_approach_dropdown()
+                model2 = create_model_dropdown()
+                approach2 = create_approach_dropdown()
+            
+            compare_input = gr.Textbox(label="Enter your message for comparison")
+            compare_button = gr.Button("Compare")
+            
+            with gr.Row():
+                output1 = gr.Textbox(label="Response 1")
+                output2 = gr.Textbox(label="Response 2")
+            
+            compare_button.click(
+                compare_responses,
+                inputs=[compare_input, model1, approach1, model2, approach2, system_message, max_tokens, temperature, top_p],
+                outputs=[output1, output2]
             )
 
+if __name__ == "__main__":
     demo.launch()