Fix request body and add api_key field

.env.example

@@ -1 +0,0 @@
-OPENAI_API_KEY=

app.py

@@ -1,7 +1,6 @@
 import gradio as gr
 import time
 from openai import OpenAI
-from dotenv import load_dotenv
 from tenacity import (
     retry,
     stop_after_attempt,
@@ -9,25 +8,13 @@ from tenacity import (
 )
 import prompts
 
-load_dotenv()
-
-# Initialize OpenAI client
-client = OpenAI(
-    base_url="https://openai-proxy.replicate.com/v1"
-)
-
 
 previous_thought = ""
 previous_answer = ""
 is_clearing = False
 
 
-@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
-def completion_with_backoff(**kwargs):
-    return client.chat.completions.create(**kwargs)
-
-
-def ai_response(input_text, shared_text, temperature):
+def ai_response(api_key, input_text, shared_text, temperature):
     global previous_thought
 
     in_context_learning = [*prompts.continue_skill_example, *prompts.boilerplate_example, *prompts.continue_complete_skill_example,
@@ -40,19 +27,29 @@ def ai_response(input_text, shared_text, temperature):
          "content": str({"shared_context": shared_text, "previous_thought": previous_thought})},
         {"role": "user", "content": input_text}
     ]
-    messages = [*prompts.system_prompt, *in_context_learning, *context]
+    messages = [prompts.system_prompt, *in_context_learning, *context]
     print(messages)
 
     shared_text = ""
 
     print(messages)
 
+    # Initialize OpenAI client
+    client = OpenAI(
+        base_url="https://openai-proxy.replicate.com/v1",
+        api_key=api_key
+    )
+
+    @retry(wait=wait_random_exponential(min=1, max=120), stop=stop_after_attempt(6))
+    def completion_with_backoff(**kwargs):
+        return client.chat.completions.create(**kwargs)
+
     stream = completion_with_backoff(
         model='meta/llama-2-70b-chat',
         temperature=temperature,
         messages=messages,
         response_format={"type": "json_object"},
-        stream=True
+        stream=True,
     )
 
     thought = ""
@@ -79,13 +76,15 @@ def ai_response(input_text, shared_text, temperature):
 
 
 with gr.Blocks() as demo:
+    api_input = gr.Textbox(label="Your OpenAI API key")
+
     user_input = gr.Textbox(lines=2, label="User Input")
     cot_textbox = gr.Textbox(label="CoT etc.")
     shared_textbox = gr.Textbox(label="Shared Textbox", interactive=True)
     temperature = gr.Slider(label="Temperature", minimum=0, maximum=2, step=0.01, value=0.01)
     # n_shots = gr.Slider(label="N-shots (~150 tokens each. It should not work 0-shot)", minimum=0, maximum=5, step=1, value=1)
     ai_btn = gr.Button("Generate AI Response")
-    generation = ai_btn.click(fn=ai_response, inputs=[user_input, shared_textbox, temperature],
+    generation = ai_btn.click(fn=ai_response, inputs=[api_input, user_input, shared_textbox, temperature],
                               outputs=[shared_textbox, cot_textbox])