nerdalert/docbot

Uses Langchain, ChatGPT and Flask to upload your own docs via a web UI to be indexed and and queried

Docbot 🤖

Upload your own docs to be queried by the docbot using Langchain and the language model gpt-3.5-turbo. Exploring how project docs are low-hanging fruit for how natural language processing (NLP) will have immediate impact on software documentation as the tiny tip of the iceberg.

• For a practical implementation of chatting with your project docs, check out nerdalert/docbot-nexodus 🤖

docbot.mp4

Docbot Pre-Requisites

• Python
• Flask

pip install -U Flask

• LLM modules

pip install llama_index openai langchain

# Or with requirements.txt

pip install -r requirements.txt

Add your ChatGPT key in docbot.py. Get a key from OpenAI. You get $5 free in API usage from OpenAI. If you upload a small file like the example, gpt-3.5-turbo model doesn't incur much, if any cost. GAI competition should drive small usage to zero.

os.environ["OPENAI_API_KEY"] = "insert_your_key_here"

Run

Run the app with python3 docbot.py which starts the flask server running the web UI.

$ git clone https://github.com/nerdalert/docs-chatbot.git
$ python3 docbot.py
 * Serving Flask app 'docbot'
 * Debug mode: on
INFO:werkzeug:WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on all addresses (0.0.0.0)
 * Running on http://127.0.0.1:5000
 * Running on http://192.168.1.203:5000
INFO:werkzeug:Press CTRL+C to quit

Docbot Usage

Point your browser at the link flask gives you and upload a document. I uploaded this markdown file. You could also add txt files (and probably lots of other filetypes). If you ask about something completely irrelevant to the docs it will tell you it doesn't know about that topic.

Bugs/Incomplete:

• The upload button doesn't show what file you chose and once you hit upload, it doesn't list what file has been indexed in json.index but you can see the uploaded file in the ./data directory and view the index.json file in the root directory. The file gets chunked and uploaded to ChatGPT then indexed locally.
• This doesn't support PDFs, but it's easy to add with from langchain.document_loaders import PyPDFLoader
• Langchain is awesome! It is changing daily, for example, with I need to swap LLMPredictor(model_name="gpt-3.5-turbo") to from langchain.chat_models import ChatOpenAI for gpt-3.5-turbo so you will see a deprecation warning until I fix it.