Manav0603 commited on
Commit
bdfdb09
·
verified ·
1 Parent(s): a61ca29

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +48 -50
app.py CHANGED
@@ -1,7 +1,4 @@
1
- from fastapi import FastAPI, UploadFile, File
2
- from fastapi.middleware.cors import CORSMiddleware
3
- from pydantic import BaseModel
4
- import shutil
5
  import os
6
  import requests
7
  from langchain_community.document_loaders import PyPDFLoader
@@ -10,33 +7,34 @@ from langchain_community.embeddings import HuggingFaceEmbeddings
10
  from langchain_community.vectorstores import Chroma
11
  from langchain.chains import RetrievalQA
12
 
13
- # Hugging Face API setup
14
  HF_API_URL = "https://api-inference.huggingface.co/models/your-username/your-model"
15
  HF_API_TOKEN = "hf_xxxYOUR_TOKENxxx" # Replace with your HF access token
16
  headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
17
 
18
- app = FastAPI(title="PDF QA API via HF Model")
19
-
20
- app.add_middleware(
21
- CORSMiddleware,
22
- allow_origins=["*"],
23
- allow_methods=["*"],
24
- allow_headers=["*"],
25
- )
26
-
27
- UPLOAD_DIR = "./uploads"
28
- os.makedirs(UPLOAD_DIR, exist_ok=True)
29
-
30
  qa = None
 
 
31
 
32
- class QueryRequest(BaseModel):
33
- query: str
 
 
 
 
 
 
 
 
34
 
35
- # ---------------- PDF Processing ----------------
36
- def process_pdf(pdf_file_path):
37
  global qa
 
 
38
 
39
- loader = PyPDFLoader(pdf_file_path)
40
  documents = loader.load()
41
 
42
  text_splitter = RecursiveCharacterTextSplitter(
@@ -46,41 +44,41 @@ def process_pdf(pdf_file_path):
46
  docs = text_splitter.split_documents(documents)
47
 
48
  embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
49
- vectorstore = Chroma.from_documents(docs, embeddings, persist_directory="./chroma_cache")
50
 
51
- # RetrievalQA using HF API for completion
52
- def hf_llm_call(prompt):
53
- payload = {"inputs": prompt}
54
- response = requests.post(HF_API_URL, headers=headers, json=payload)
55
- if response.status_code == 200:
56
- output = response.json()
57
- # HF API response format varies; usually output[0]["generated_text"]
58
- return output[0]["generated_text"]
59
- else:
60
- return f"Error: {response.status_code} - {response.text}"
61
 
62
  qa = RetrievalQA.from_chain_type(
63
- llm=hf_llm_call,
64
  retriever=vectorstore.as_retriever()
65
  )
66
 
67
- return "PDF processed and QA ready via HF API!"
68
-
69
- # ---------------- Endpoints ----------------
70
- @app.post("/upload_pdf")
71
- async def upload_pdf(file: UploadFile = File(...)):
72
- file_path = os.path.join(UPLOAD_DIR, file.filename)
73
- with open(file_path, "wb") as f:
74
- shutil.copyfileobj(file.file, f)
75
-
76
- msg = process_pdf(file_path)
77
- return {"message": msg}
78
 
79
- @app.post("/ask")
80
- def ask_question(request: QueryRequest):
81
  global qa
82
  if qa is None:
83
- return {"error": "No PDF uploaded yet!"}
 
 
 
 
 
 
 
 
 
 
84
 
85
- answer = qa.run(request.query)
86
- return {"query": request.query, "answer": answer}
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
 
 
 
2
  import os
3
  import requests
4
  from langchain_community.document_loaders import PyPDFLoader
 
7
  from langchain_community.vectorstores import Chroma
8
  from langchain.chains import RetrievalQA
9
 
10
+ # ---------------- HF Model API Setup ----------------
11
  HF_API_URL = "https://api-inference.huggingface.co/models/your-username/your-model"
12
  HF_API_TOKEN = "hf_xxxYOUR_TOKENxxx" # Replace with your HF access token
13
  headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
14
 
15
+ # Global QA object
 
 
 
 
 
 
 
 
 
 
 
16
  qa = None
17
+ UPLOAD_FOLDER = "uploads"
18
+ os.makedirs(UPLOAD_FOLDER, exist_ok=True)
19
 
20
+ # ---------------- Functions ----------------
21
+ def hf_llm_call(prompt: str):
22
+ """Call HF hosted model API"""
23
+ payload = {"inputs": prompt}
24
+ response = requests.post(HF_API_URL, headers=headers, json=payload)
25
+ if response.status_code == 200:
26
+ output = response.json()
27
+ return output[0]["generated_text"]
28
+ else:
29
+ return f"Error: {response.status_code} - {response.text}"
30
 
31
+ def process_pdf(pdf_file):
32
+ """Load PDF, create embeddings, setup RetrievalQA with HF API"""
33
  global qa
34
+ file_path = os.path.join(UPLOAD_FOLDER, pdf_file.name)
35
+ pdf_file.save(file_path)
36
 
37
+ loader = PyPDFLoader(file_path)
38
  documents = loader.load()
39
 
40
  text_splitter = RecursiveCharacterTextSplitter(
 
44
  docs = text_splitter.split_documents(documents)
45
 
46
  embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
47
+ vectorstore = Chroma.from_documents(docs, embeddings)
48
 
49
+ # Setup RetrievalQA but use HF API as LLM
50
+ def llm_wrapper(prompt):
51
+ return hf_llm_call(prompt)
 
 
 
 
 
 
 
52
 
53
  qa = RetrievalQA.from_chain_type(
54
+ llm=llm_wrapper,
55
  retriever=vectorstore.as_retriever()
56
  )
57
 
58
+ return f"{pdf_file.name} uploaded and processed successfully!"
 
 
 
 
 
 
 
 
 
 
59
 
60
+ def ask_question(query):
 
61
  global qa
62
  if qa is None:
63
+ return "No PDF uploaded yet!"
64
+ return qa.run(query)
65
+
66
+ # ---------------- Gradio UI ----------------
67
+ with gr.Blocks() as demo:
68
+ gr.Markdown("## PDF QA System using HF Hosted Model")
69
+
70
+ with gr.Row():
71
+ pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
72
+ upload_btn = gr.Button("Upload & Process PDF")
73
+ output_msg = gr.Textbox(label="Status", interactive=False)
74
 
75
+ upload_btn.click(fn=process_pdf, inputs=pdf_input, outputs=output_msg)
76
+
77
+ query_input = gr.Textbox(label="Ask a question")
78
+ query_btn = gr.Button("Get Answer")
79
+ answer_output = gr.Textbox(label="Answer", interactive=False)
80
+
81
+ query_btn.click(fn=ask_question, inputs=query_input, outputs=answer_output)
82
+
83
+ if __name__ == "__main__":
84
+ demo.launch(server_name="0.0.0.0", server_port=7860)