├── .gitignore ├── app.py ├── media └── ScreenshotAskYourPDF.png ├── readme.md └── requirements.txt /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | wheels/ 23 | share/python-wheels/ 24 | *.egg-info/ 25 | .installed.cfg 26 | *.egg 27 | MANIFEST 28 | 29 | # PyInstaller 30 | # Usually these files are written by a python script from a template 31 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 32 | *.manifest 33 | *.spec 34 | 35 | # Installer logs 36 | pip-log.txt 37 | pip-delete-this-directory.txt 38 | 39 | # Unit test / coverage reports 40 | htmlcov/ 41 | .tox/ 42 | .nox/ 43 | .coverage 44 | .coverage.* 45 | .cache 46 | nosetests.xml 47 | coverage.xml 48 | *.cover 49 | *.py,cover 50 | .hypothesis/ 51 | .pytest_cache/ 52 | cover/ 53 | 54 | # Translations 55 | *.mo 56 | *.pot 57 | 58 | # Django stuff: 59 | *.log 60 | local_settings.py 61 | db.sqlite3 62 | db.sqlite3-journal 63 | 64 | # Flask stuff: 65 | instance/ 66 | .webassets-cache 67 | 68 | # Scrapy stuff: 69 | .scrapy 70 | 71 | # Sphinx documentation 72 | docs/_build/ 73 | 74 | # PyBuilder 75 | .pybuilder/ 76 | target/ 77 | 78 | # Jupyter Notebook 79 | .ipynb_checkpoints 80 | 81 | # IPython 82 | profile_default/ 83 | ipython_config.py 84 | 85 | # pyenv 86 | # For a library or package, you might want to ignore these files since the code is 87 | # intended to run in multiple environments; otherwise, check them in: 88 | # .python-version 89 | 90 | # pipenv 91 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 92 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 93 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 94 | # install all needed dependencies. 95 | #Pipfile.lock 96 | 97 | # poetry 98 | # Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. 99 | # This is especially recommended for binary packages to ensure reproducibility, and is more 100 | # commonly ignored for libraries. 101 | # https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control 102 | #poetry.lock 103 | 104 | # pdm 105 | # Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. 106 | #pdm.lock 107 | # pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it 108 | # in version control. 109 | # https://pdm.fming.dev/#use-with-ide 110 | .pdm.toml 111 | 112 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm 113 | __pypackages__/ 114 | 115 | # Celery stuff 116 | celerybeat-schedule 117 | celerybeat.pid 118 | 119 | # SageMath parsed files 120 | *.sage.py 121 | 122 | # Environments 123 | .env 124 | .venv 125 | env/ 126 | venv/ 127 | ENV/ 128 | env.bak/ 129 | venv.bak/ 130 | 131 | # Spyder project settings 132 | .spyderproject 133 | .spyproject 134 | 135 | # Rope project settings 136 | .ropeproject 137 | 138 | # mkdocs documentation 139 | /site 140 | 141 | # mypy 142 | .mypy_cache/ 143 | .dmypy.json 144 | dmypy.json 145 | 146 | # Pyre type checker 147 | .pyre/ 148 | 149 | # pytype static type analyzer 150 | .pytype/ 151 | 152 | # Cython debug symbols 153 | cython_debug/ 154 | 155 | # PyCharm 156 | # JetBrains specific template is maintained in a separate JetBrains.gitignore that can 157 | # be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore 158 | # and can be added to the global gitignore or merged into this file. For a more nuclear 159 | # option (not recommended) you can uncomment the following to ignore the entire idea folder. 160 | #.idea/ 161 | 162 | *.pdf 163 | *.bin 164 | .vscode -------------------------------------------------------------------------------- /app.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | from PyPDF2 import PdfReader 3 | 4 | import langchain 5 | from langchain.text_splitter import CharacterTextSplitter 6 | from langchain.chains.question_answering import load_qa_chain 7 | from langchain.llms import LlamaCpp 8 | from langchain.vectorstores import Qdrant 9 | from langchain.embeddings import SentenceTransformerEmbeddings 10 | from langchain.callbacks.manager import CallbackManager 11 | from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler 12 | 13 | # Fix for some new weird "no attribute 'verbose'" bug https://github.com/hwchase17/langchain/issues/4164 14 | langchain.verbose = False 15 | 16 | 17 | def main(): 18 | # Callback just to stream output to stdout, can be removed 19 | callback_manager = CallbackManager([StreamingStdOutCallbackHandler()]) 20 | 21 | # stable-vicuna through LlamaCpp 22 | # Download model manually at https://huggingface.co/TheBloke/stable-vicuna-13B-GGML/tree/main 23 | llm = LlamaCpp( 24 | model_path="./stable-vicuna-13B.ggml.q4_2.bin", 25 | stop=["### Human:"], 26 | callback_manager=callback_manager, 27 | verbose=True, 28 | n_ctx=2048, 29 | n_batch=512, 30 | ) 31 | 32 | # Load question answering chain 33 | chain = load_qa_chain(llm, chain_type="stuff") 34 | 35 | # Patching qa_chain prompt template to better suit the stable-vicuna model 36 | # see https://huggingface.co/TheBloke/stable-vicuna-13B-GGML#prompt-template 37 | if "Helpful Answer:" in chain.llm_chain.prompt.template: 38 | chain.llm_chain.prompt.template = ( 39 | f"### Human:{chain.llm_chain.prompt.template}".replace( 40 | "Helpful Answer:", "\n### Assistant:" 41 | ) 42 | ) 43 | 44 | # Page setup 45 | st.set_page_config(page_title="Ask your PDF") 46 | st.header("Ask your PDF 💬") 47 | pdf = st.file_uploader("Upload a PDF", type=["pdf"]) 48 | 49 | if pdf: 50 | pdf_reader = PdfReader(pdf) 51 | 52 | # Collect text from pdf 53 | text = "" 54 | for page in pdf_reader.pages: 55 | text += page.extract_text() 56 | 57 | # Split the text into chunks 58 | text_splitter = CharacterTextSplitter( 59 | separator="\n", chunk_size=1000, chunk_overlap=200, length_function=len 60 | ) 61 | chunks = text_splitter.split_text(text) 62 | 63 | # Use https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2 as embedding 64 | # (downloaded automatically) 65 | embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2") 66 | 67 | # Create in-memory Qdrant instance 68 | knowledge_base = Qdrant.from_texts( 69 | chunks, 70 | embeddings, 71 | location=":memory:", 72 | collection_name="doc_chunks", 73 | ) 74 | 75 | user_question = st.text_input("Ask a question about your PDF:") 76 | 77 | if user_question: 78 | docs = knowledge_base.similarity_search(user_question, k=4) 79 | 80 | # Calculating prompt (takes time and can optionally be removed) 81 | prompt_len = chain.prompt_length(docs=docs, question=user_question) 82 | st.write(f"Prompt len: {prompt_len}") 83 | if prompt_len > llm.n_ctx: 84 | st.write( 85 | "Prompt length is more than n_ctx. This will likely fail. Increase model's context, reduce chunk's \ 86 | sizes or question length, or retrieve less number of docs." 87 | ) 88 | 89 | # Grab and print response 90 | response = chain.run(input_documents=docs, question=user_question) 91 | st.write(response) 92 | 93 | 94 | if __name__ == "__main__": 95 | main() 96 | -------------------------------------------------------------------------------- /media/ScreenshotAskYourPDF.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wafflecomposite/langchain-ask-pdf-local/a4f188c1d01d507b83f7b33799718f2971003336/media/ScreenshotAskYourPDF.png -------------------------------------------------------------------------------- /readme.md: -------------------------------------------------------------------------------- 1 | # Ask Your PDF, locally 2 | 3 | | ![UI screenshot of Ask Your PDF](media/ScreenshotAskYourPDF.png) | 4 | |:--:| 5 | | Answering question about [2303.12712 paper](https://arxiv.org/pdf/2303.12712.pdf) 7mb pdf file | 6 | 7 | This is an attempt to recreate [Alejandro AO's langchain-ask-pdf](https://github.com/alejandro-ao/langchain-ask-pdf) (also check out [his tutorial on YT](https://www.youtube.com/watch?v=wUAUdEw5oxM)) using open source models running locally. 8 | 9 | It uses [all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) instead of OpenAI Embeddings, and [StableVicuna-13B](https://huggingface.co/CarperAI/stable-vicuna-13b-delta) instead of OpenAI models. 10 | 11 | It runs on the CPU, is impractically slow and was created more as an experiment, but I am still fairly happy with the results. 12 | 13 | ## Requirements 14 | GPU is not used and is not required. 15 | 16 | You can squeeze it into 16 GB of RAM, but I recommend 24 GB or more. 17 | 18 | ## Installation 19 | 20 | - Install requirements (preferably to `venv`): `pip install -r requirements.txt` 21 | 22 | - Download `stable-vicuna-13B.ggml.q4_2.bin` from [TheBloke/stable-vicuna-13B-GGML](https://huggingface.co/TheBloke/stable-vicuna-13B-GGML/tree/main) and place it in project folder. 23 | 24 | ## Usage 25 | 26 | Run `streamlit run .\app.py` 27 | 28 | This should launch the UI in your default browser. Select a PDF file, send the question, wait patiently. 29 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | langchain==0.0.161 2 | PyPDF2==3.0.1 3 | streamlit==1.22.0 4 | qdrant-client==1.1.6 5 | sentence-transformers==2.2.2 6 | llama-cpp-python==0.1.44 --------------------------------------------------------------------------------