Building Generative AI Applications

Prerequisite

Download Visual Studio Code compatible with your operating system.

To use ElevenLabs, we need API key. Go to ElevenLabs and create an account. It's free! 🎉. And in the upper right corner click on your profile picture > profile. Next click on the eye icon and copy/save your API key.

To use LLMs, we need API key(s). During demo we are using Azure OpenAI and AWS Bedrock models. Both require a separate subscription of a cloud provider. You can use OpenAI API or any other LLM provider, just remember to update the blocks in langflow.

Local development

Python related

We are using poetry to manage packages, and conda to manage python virtual environments

Follow this guide to install conda

Once you have conda installed, we can create a new venv on conda using

conda env create -f environment.yaml

Remember to keep the conda's environment.yml file clean, we keep dependencies on poetry. If we need to upgrade python or poetry version - upgrade those in the environment.yml and recreate the environment.

We store the project python configuration in pyproject.toml Especially python version

[tool.poetry.dependencies]
python = "^3.12"

Run below command to install all dependencies

poetry install

Make sure the right python interpreter is selected

conda activate your_env_name
which python
poetry run which python
poetry shell
which python

Install dependencies

poetry install

Running in Docker

To run langflow you need to have three containers minimum:

• langflow
• postgres
• streamlit-app

streamlit-app is a custom application, hence you will need to build it first.

cd app/
docker build -t streamlit-app .

First two container uses .env file, that can be copied from .env.template, and should be filled accordingly.

Run then the below command

docker compose up -d

Langflow

Installing packages

Connect to the langflow container

docker exec -it app_langflow /bin/bash

Now run

which python
which pip3

If your answer is like below, we need to reinstall pip

/app/.venv/bin/python
/usr/local/bin/pip3

You need to make sure there is a right pip

source activate /app/.venv/bin/activate
python -m ensurepip --upgrade
which pip3

Now we should get

/app/.venv/bin/pip3

Now you can install packages, for example to run tavily search

pip3 install tavily-python

Ollama

We need to mamually pull models we would like to use on ollama, so once the container is up we can run

docker exec -it app_ollama /bin/bash

and now we can pull the models

ollama pull gemma2:2b
ollama pull nomic-embed-text

More about ollama models is on their website