We'll want full end-to-end tests for each interface mode that's supported. We likely want to have different versions of the same test for different providers (OpenAI/AWS Bedrock), different databases (SQLite/PostgreSQL), and orchestration patterns (agent-as-tool/magentic/sequential).
We'll want full end-to-end tests for each interface mode that's supported. We likely want to have different versions of the same test for different providers (OpenAI/AWS Bedrock), different databases (SQLite/PostgreSQL), and orchestration patterns (agent-as-tool/magentic/sequential).