main.py

from haystack.nodes import PreProcessor, PDFToTextConverter, EmbeddingRetriever
from haystack.document_stores import InMemoryDocumentStore
from haystack.pipelines import DocumentSearchPipeline
import gradio as gr

preprocessor = PreProcessor(
    clean_empty_lines=True,
    clean_whitespace=True,
    clean_header_footer=True,
    split_by="word",
    split_length=100,
    split_respect_sentence_boundary=True,
    split_overlap=3
)
document_store = InMemoryDocumentStore()
retriever = EmbeddingRetriever(document_store=document_store, embedding_model="sentence-transformers/all-MiniLM-L6-v2")
pipeline = DocumentSearchPipeline(retriever)

def print_answers(results):
    fields = ["answer", "score"] # "context"
    answers = results["answers"]
    filtered_answers = []
    for ans in answers:
        filtered_ans = {
            field: getattr(ans, field) for field in fields if getattr(ans, field) is not None
        }
        filtered_answers.append(filtered_ans)
    return filtered_answers

def write_pdf(pdf_file):
    converter = PDFToTextConverter(remove_numeric_tables=True, valid_languages=["en"])
    document = converter.convert(file_path=pdf_file.name, meta=None)[0]
    preprocessed_docs = preprocessor.process(document)
    document_store.write_documents(preprocessed_docs)

def predict(question, pdf_file):
    print("Start processing pdf")
    write_pdf(pdf_file)
    print("Processing done.")
    result = pipeline.run(query=question, params={"Retriever": { "top_k": 2 }})
    answers = print_answers(result)
    return answers

title = "Search"
interface = gr.Interface(
    fn=predict,
    inputs=[gr.components.Textbox(lines = 3, label="Ask an open question!"),gr.components.File(file_count="single", type="file", label="Upload a pdf")],
    outputs="text",
    title=title, 
    flagging_options=["top", "medium", "bad"],
    interpretation="default",
    theme="default" # “default", “huggingface", “dark-grass", “peach"
)

interface.launch()