Kosmos E2E Testing Skill

End-to-end testing for an autonomous AI scientist project is automated by validating core research workflows, gap implementations, and provider configurations through a tiered testing approach using pytest. This ensures component reliability and functional integration across different environments.

Yes, E2E tests support local LLM providers like Ollama, external APIs such as Anthropic and OpenAI, or mock services. The testing framework auto-detects configured providers to validate research workflows across your chosen environment.

Docker is required for full E2E testing functionality to enable database integrations and sandboxing. The environment validation feature checks for the availability and proper setup of Docker, Ollama, and API keys before running tests.

A tiered testing approach for LLM applications supports different test suites ranging from quick sanity checks to full E2E validation. This allows you to run fast sanity tests using auto-detected providers or comprehensive checks before deployment.

Environment validation checks for the availability and proper setup of dependencies like Docker, Ollama, and API keys before executing tests. This ensures your local and external LLM provider configurations are correctly established for accurate E2E testing.

Using mock services for E2E testing validates workflow integration without actual LLM generation, limiting the verification of provider-specific configurations and research quality. Full functionality requires pytest, Docker, and configured LLM access.