Update main.py

main.py

@@ -1,59 +1,39 @@
 from fastapi import FastAPI
 from pydantic import BaseModel
-from huggingface_hub import InferenceClient
-from fastapi.responses import StreamingResponse
-import uvicorn
+from transformers import pipeline
 
 app = FastAPI()
 
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-class Item(BaseModel):
-    prompt: str
-    history: list
-    system_prompt: str
-    temperature: float = 0.0
-    max_new_tokens: int = 1048
-    top_p: float = 0.15
-    repetition_penalty: float = 1.0
-
-def format_prompt(message, history):
-    print("````")
-    print(message)
-    print("++++")
-    print(history)
-    print("````")
-    prompt = "<s>"
-    for user_prompt, bot_response in history:
-        prompt += f"[INST] {user_prompt} [/INST]"
-        prompt += f" {bot_response}</s> "
-    prompt += f"[INST] {message} [/INST]"
-    return prompt
-
-async def generate_stream(item: Item):
-    temperature = float(item.temperature)
-    if temperature < 1e-2:
-        temperature = 1e-2
-    top_p = float(item.top_p)
-
-    generate_kwargs = dict(
-        temperature=temperature,
-        max_new_tokens=item.max_new_tokens,
-        top_p=top_p,
-        repetition_penalty=item.repetition_penalty,
-        do_sample=True,
-        seed=42,
+# Load model once
+classifier = pipeline(
+    "zero-shot-classification",
+    model="MoritzLaurer/mDeBERTa-v3-base-mnli-xnli"
+)
+
+# Your classes
+CANDIDATE_LABELS = [
+    "Garbage issue",
+    "Streetlight not working",
+    "Road damage",
+    "Water supply issue",
+    "Noise pollution",
+    "Flooding",
+    "Corruption",
+    "Other"
+]
+
+class Query(BaseModel):
+    text: str
+
+@app.post("/predict")
+def predict(query: Query):
+    result = classifier(
+        query.text,
+        candidate_labels=CANDIDATE_LABELS,
+        multi_label=False   # <-- single best class
     )
 
-    formatted_prompt = format_prompt(f"{item.system_prompt} [/INST] Ok..! </s> [INST] {item.prompt}", item.history)
-    print(formatted_prompt)
-    print("=======")
-    print(item.history)
-    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
-
-    for response in stream:
-        yield response.token.text  # Stream each token as it's received
+    # Top-1 predicted label
+    predicted_label = result["labels"][0]
 
-@app.post("/generate/")
-async def generate_text(item: Item):
-    return StreamingResponse(generate_stream(item), media_type="text/plain")
+    return {"label": predicted_label}