app.py

from fastapi import FastAPI
from pydantic import BaseModel
from transformers import pipeline, AutoTokenizer
import nltk
import os
import uvicorn
import time

from chunker import chunk_by_token_limit

# Setup NLTK directory
NLTK_DATA_DIR = "/app/nltk_data"
os.makedirs(NLTK_DATA_DIR, exist_ok=True)
nltk.data.path.append(NLTK_DATA_DIR)

print("📦 Downloading NLTK 'punkt' tokenizer...")
nltk.download("punkt", download_dir=NLTK_DATA_DIR, quiet=True)

app = FastAPI()

HF_AUTH_TOKEN = os.getenv("HF_TOKEN")

MODEL_NAME = "VincentMuriuki/legal-summarizer"
print(f"🚀 Loading summarization pipeline: {MODEL_NAME}")
start_model_load = time.time()
summarizer = pipeline("summarization", model=MODEL_NAME, token=HF_AUTH_TOKEN)
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, token=HF_AUTH_TOKEN)
print(f"✅ Model loaded in {time.time() - start_model_load:.2f}s")

class SummarizeInput(BaseModel):
    text: str

class ChunkInput(BaseModel):
    text: str
    max_tokens: int = 1024

@app.post("/summarize")
def summarize_text(data: SummarizeInput):
    print("📥 Received summarize request.")
    start = time.time()
    summary = summarizer(data.text, max_length=150, min_length=30, do_sample=False)
    duration = time.time() - start
    print(f"📝 Summary generated in {duration:.2f}s.")
    return {"summary": summary[0]["summary_text"], "time_taken": f"{duration:.2f}s"}

@app.post("/chunk")
def chunk_text(data: ChunkInput):
    print(f"📥 Received chunking request with max_tokens={data.max_tokens}")
    start = time.time()
    chunks = chunk_by_token_limit(data.text, data.max_tokens, tokenizer)
    duration = time.time() - start
    print(f"🔖 Chunking completed in {duration:.2f}s. Total chunks: {len(chunks)}")
    return {"chunks": chunks, "chunk_count": len(chunks), "time_taken": f"{duration:.2f}s"}

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)