Update main.py

main.py

@@ -1,50 +1,65 @@
-from fastapi import FastAPI, Response
+from fastapi import FastAPI
 from pydantic import BaseModel
 from huggingface_hub import InferenceClient
-import graphviz
+import uvicorn
 
 app = FastAPI()
 
-# Initialize the inference client for the AI model
+# Initialize the InferenceClient with the specified model
 client = InferenceClient("nvidia/Llama-3.1-Nemotron-70B-Instruct-HF")
 
+# Define the structure of the request body
 class CourseRequest(BaseModel):
     course_name: str
+    history: list = []  # Keeping history optional
+    temperature: float = 0.0
+    max_new_tokens: int = 1048
+    top_p: float = 0.15
+    repetition_penalty: float = 1.0
+
+# Format the prompt for the model
+def format_prompt(course_name, history):
+    prompt = "<s>"
+    for user_prompt, bot_response in history:
+        prompt += f"[INST] {user_prompt} [/INST] {bot_response} </s> "
+    prompt += f"[INST] Generate a roadmap for the course: {course_name} [/INST]"
+    return prompt
+
+# Generate text using the specified parameters
+def generate(course_request: CourseRequest):
+    temperature = max(float(course_request.temperature), 1e-2)
+    top_p = float(course_request.top_p)
 
-def format_prompt(course_name: str):
-    return f"As an expert in education, please generate a detailed roadmap for the course '{course_name}'. Include key topics."
 
-def generate_roadmap(item: CourseRequest):
-    prompt = format_prompt(item.course_name)
-    stream = client.text_generation(prompt, max_new_tokens=200)
-    output = ""
 
-    for response in stream:
-        output += response.token.text
 
-    return output
 
-def create_diagram(roadmap_text: str):
-    dot = graphviz.Digraph()
 
-    # Split the roadmap text into lines or sections for diagram creation
-    lines = roadmap_text.split('\n')
-    for i, line in enumerate(lines):
-        dot.node(str(i), line.strip())  # Create a node for each topic
 
-        if i > 0:
-            dot.edge(str(i - 1), str(i))  # Connect nodes sequentially
 
-    return dot
+    generate_kwargs = {
+        'temperature': temperature,
+        'max_new_tokens': course_request.max_new_tokens,
+        'top_p': top_p,
+        'repetition_penalty': course_request.repetition_penalty,
+        'do_sample': True,
+        'seed': 42,
+    }
 
-@app.post("/generate/")
-async def generate_roadmap_endpoint(course_request: CourseRequest):
-    roadmap_text = generate_roadmap(course_request)
-    diagram = create_diagram(roadmap_text)
+    formatted_prompt = format_prompt(course_request.course_name, course_request.history)
+    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
+    output = ""
+
+    for response in stream:
+        output += response.token.text
+
+    return output
 
-    # Render the diagram to a PNG image
-    diagram_path = "/tmp/roadmap.png"
-    diagram.render(diagram_path, format='png', cleanup=True)
+# Define the API endpoint for generating course roadmaps
+@app.post("/generate-roadmap/")
+async def generate_roadmap(course_request: CourseRequest):
+    return {"roadmap": generate(course_request)}
 
-    with open(diagram_path, "rb") as f:
-        return Response(content=f.read(), media_type="image/png")
+# Run the application (uncomment the next two lines if running this as a standalone script)
+# if __name__ == "__main__":
+#     uvicorn.run(app, host="0.0.0.0", port=8000)