Intro to Gen AI Observability - LLM Tracing with LangSmith and Phoenix

Observability is crucial for Generative AI (Gen AI) applications not just in production, but also during development and early iteration.

Observability in Gen AI applications is more complex and multifaceted compared to traditional distributed applications due to the additional layers of monitoring required for model performance, data quality, bias, and explainability. While the foundational principles of observability remain the same, the specific needs and focus areas differ significantly between the two.

In this session, we will have a brief presentation to introduce the need for and salient characteristics of observability for Gen AI applications. This will be followed by a demo/hands-on session on LangServe, Arize Phoenix, and LangFuse which are some of the popular solutions for Gen AI observability.

If you wish to participate in the hands-on session, please create the following secrets in your Google Colab environment:

• OPENAI_API_KEY - Requires a paid OpenAI subscription
• SERPAPI_API_KEY - Free, limited quota (Register for free account at https://serpapi.com/users/sign_up)
• LANGCHAIN_API_KEY - Free, limited quota (Sign up for free account at https://www.langchain.com/)
• LANGFUSE_SECRET_KEY - Free limited quota (Sign up for free account at https://langfuse.com/)