LLMs

LLMs are no longer a standalone Technology Radar item because they are a foundational capability rather than an implementation decision. The Radar instead tracks concrete AI technology choices.

We are using multiple LLM providers, evaluating performance for each use case:

• OpenAI
  • Auto-linking
  • Themes
  • Websearch and context scraping
  • AI Analyze
• Anthropic (Provided by both Anthropic directly and through Amazon Bedrock as fallback)
  • Spark agent
• Vertex AI
  • Rolling summaries (condensing agent's context window) with Gemini

Resources

• OpenAI Platform
• Anthropic
• Vertex AI

We are using Large Language Models (LLMs) for Productboard AI. Our default vendor for LLMs is OpenAI. We can use Anthropic Claude as a fallback or for any future needs.

https://platform.openai.com/ https://claude.ai/