I'm working on a Python application that utilizes the OpenAI API to process and generate responses based on the text extracted from PDF documents.
def main():load_dotenv()st.set_page_config(page_title="Ask your PDF")st.header("Ask your PDF 💬")# upload filepdf = st.file_uploader("Upload your PDF", type="pdf")# extract the textif pdf is not None: pdf_reader = PdfReader(pdf) text = "" for page in pdf_reader.pages: text += page.extract_text() # split into chunks text_splitter = CharacterTextSplitter( separator="\n", chunk_size=1000, chunk_overlap=200, length_function=len ) chunks = text_splitter.split_text(text) # create embeddings embeddings = OpenAIEmbeddings() knowledge_base = FAISS.from_texts(chunks, embeddings) # show user input user_question = 'Write a cover letter using file' if user_question: docs = knowledge_base.similarity_search(user_question) llm = OpenAI() chain = load_qa_chain(llm, chain_type="stuff") with get_openai_callback() as cb: response = chain.run(input_documents=docs, question=user_question) print(cb) st.write(response) print(response)if __name__ == '__main__':main()I was under the impression that the token limit for a single request to the OpenAI API (including both the input and the response) is 4096 tokens. However, I noticed that the responses I'm getting are incomplete and capped at 1074 tokens, which seems to be significantly under the limit.