Airbyte, Dagster, Langchain Example

References:

• Implement AI data pipelines with Langchain, Airbyte, and Dagster
• LLM training pipelines with Langchain, Airbyte, and Dagster
• Using Airbyte with Dagster
• Langchain Airbyte JSON

Tested on Ubuntu 20.04:

• Install docker and docker-compose
• Install and start Airbyte

    mkdir ~/build
    cd ~/build
    git clone https://github.com/airbytehq/airbyte.git
    cd airbyte
    # Edit BASIC_AUTH_PASSWORD in .env from the default
    ./run-ab-platform.sh 
  

• Once you see an Airbyte banner, the UI is ready to go at http://localhost:8000. It actually runs at 0.0.0.0:8000.
  • You will be asked for a username and password. By default, that’s username airbyte and password password.
  • Once you deploy airbyte to your servers, be sure to change these in your .env file.
  • Configure Airbyte to connect Sample Data source to Local JSON destination.
  • Pick test as destination path. This will result in three output files /tmp/airbyte_local/_airbyte_raw_{stream_name}.jsonl.
• Create the build folder:

    mkdir ~/build/airbyte-dagster-langchain
    cd ~/build/airbyte-dagster-langchain
  

• Set up venv, and enter the virtual environment:

    python3 -m venv ~/.venv/langchain
    . ~/.venv/langchain/bin/activate
    pip install openai faiss-cpu requests beautifulsoup4 tiktoken dagster_managed_elements langchain dagster dagster-airbyte dagit
  

• Download https://github.com/airbytehq/dagster-langchain/blob/main/ingest.py
• Edit the airbyte_loader to point it to one of the configured /tmp/airbyte_local/_airbyte_raw_{stream_name}.jsonl files
• Add the user and password here: python airbyte_instance = AirbyteResource( host="localhost", port="8000", username="airbyte", password="password", )
• Set up Dagster

    if [ ! -d $DAGSTER_HOME ]; then mkdir -p $DAGSTER_HOME ]; fi
    touch ${DAGSTER_HOME}/dagster.yaml
  

• Put your OPENAI_API_KEY and DAGSTER_HOME in your ~/.bashrc:

    export OPENAI_API_KEY=XXXg
    export DAGSTER_HOME=~/build/airbyte-dagster-langchain/dagster_home    
  

• Re-source the ~/.bashrc if necessary. Re-enter the virtual env.
• Start Dagster:

    dagster dev -f ingest.py -h 0.0.0.0
  

  The -h parameter controls the dagit host, and -p controls the port. Default port is 3000. Use ingest_all.py for ingesting multiple sources.

• Start Streamlit:

    . ~/.venv/streamlit/bin/activate
    streamlit run chatbot.py --server.port 8080
  

• Open the ui at port 8501