Create app.py

app.py

@@ -0,0 +1,76 @@
+import os
+import gradio as gr
+from pypdf import PdfReader
+
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_community.vectorstores import FAISS
+from langchain_openai import OpenAIEmbeddings, ChatOpenAI
+from langchain.schema import Document
+
+APP_TITLE = "Election Memory Engine"
+APP_TAGLINE = "Ask your newsroom archive anything — with sources."
+
+def read_pdf(file_path: str) -> str:
+    reader = PdfReader(file_path)
+    pages = []
+    for p in reader.pages:
+        pages.append(p.extract_text() or "")
+    return "\n".join(pages)
+
+def build_index(files):
+    docs = []
+    for f in files:
+        text = read_pdf(f.name)
+        source_name = os.path.basename(f.name)
+        docs.append(Document(page_content=text, metadata={"source": source_name}))
+
+    splitter = RecursiveCharacterTextSplitter(chunk_size=900, chunk_overlap=120)
+    chunks = splitter.split_documents(docs)
+
+    embeddings = OpenAIEmbeddings()
+    index = FAISS.from_documents(chunks, embeddings)
+    return index
+
+def answer_question(index, question):
+    retriever = index.as_retriever(search_kwargs={"k": 4})
+    hits = retriever.get_relevant_documents(question)
+
+    context = ""
+    for i, d in enumerate(hits, start=1):
+        context += f"\n[Source {i}: {d.metadata['source']}]\n{d.page_content[:700]}\n"
+
+    llm = ChatOpenAI(model="gpt-4o-mini", temperature=0.2)
+
+    prompt = f"""
+You are an archive assistant for an election newsroom.
+Answer using ONLY the context below.
+Include which sources you used.
+
+Question: {question}
+
+Context:
+{context}
+"""
+
+    return llm.invoke(prompt).content
+
+with gr.Blocks() as demo:
+    gr.Markdown(f"# {APP_TITLE}")
+    gr.Markdown(APP_TAGLINE)
+
+    index_state = gr.State()
+
+    uploader = gr.File(file_types=[".pdf"], file_count="multiple")
+    build_btn = gr.Button("Build Memory")
+    question = gr.Textbox(label="Ask a question")
+    ask_btn = gr.Button("Ask")
+    output = gr.Markdown()
+
+    def build(files):
+        index = build_index(files)
+        return index
+
+    build_btn.click(build, uploader, index_state)
+    ask_btn.click(answer_question, [index_state, question], output)
+
+demo.launch()