GitHub Models: Build and Deploy your first AI App in 5 minutes
Your launchpad for getting started to AI Engineering
GitHub team granted me access to their new Github Models which is still in public limited beta. My initial experience is really good (refer video) and it solves one problem really well - Lower the entry barrier for AI Engineers to build prototype using foundation models like GPT-4o Llama 3.1 .
Software Engineers often ask me how to get into AI space. I would recommend to learn by building sample AI application using foundation models. However, there are lot of initial hurdles for getting started. Some even give up in the middle. Some of the common problems:
๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐๐๐ฟ๐๐ฒ : Every model provider has their own way of creating account, generate API keys, reference code.
๐ฃ๐ฎ๐ถ๐ฑ : Most of the model providers charge for API access limiting to get started. Although people will be ready to invest, problem here is to know which suits them.
"๐๐ผ๐ฟ๐ธ๐ ๐ผ๐ป ๐บ๐ ๐บ๐ฎ๐ฐ๐ต๐ถ๐ป๐ฒ" ๐ถ๐๐๐๐ฒ: Since AI space is fast evolving so is the SDK, library dependency. Hence many a times following exactly as described in official site or tutorial video, might still end up failing.
Github models solves above problems really well:
๐ข๐ป๐ฒ ๐๐ฐ๐ฐ๐ผ๐๐ป๐ : Just one Github account is enough to get started. Github personal access token is sufficient to access all models.
๐๐ฅ๐๐: All the models are for prototyping. This means there will be severe rate limiting which makes it unusable for production. However for trying out, its more than sufficient.
๐๐ ๐๐๐๐ ๐๐ผ๐ฟ๐ธ๐: From the playground, just choose language(Python,JS), SDK and deploy in github codespace. It just works.
๐ช๐ต๐ ๐ถ๐ ๐บ๐ฎ๐๐๐ฒ๐ฟ๐ ?
- Lowers the entry barrier into AI Engineering field. Run your first AI application inย ๐ท๐๐๐ ๐ฑ ๐บ๐ถ๐ป๐๐๐ฒ๐.
- Give access to 100 million users in Github
- Works in your development environment in Codespaces and VS Code.
- Supports multiple models - From Llama 3.1, to GPT-4o and GPT-4o mini, to Phi 3 or Mistral Large 2