fabric-e2e

End-to-end integration testing for an LLM system is executed against a live environment to validate the complete operational pipeline, including API, CLI, data capture, triage, recall with scoring, storage, and error handling. This serves as a final quality gate to ensure system integrity before deployment.

E2e testing for a data pipeline and recall mechanisms validates that observation data flows correctly from capture to recall. It ensures proper data handling, XML injection, and accurate scoring, confirming the system behaves as expected when processing new user inputs.

This integration testing skill operates as a standalone quality gate with zero dependencies. It independently confirms the integrity of the observation data flow and recall accuracy by executing comprehensive tests against a live LLM system's API and CLI.

Full stack validation is needed because it catches integration issues that unit tests miss. While unit tests isolate components, full stack validation verifies the interaction between API, CLI, capture, triage, recall, and storage, ensuring the entire system functions correctly end-to-end.

To validate that captured observations are stored and recalled with accurate scoring, run comprehensive end-to-end tests against the live system. This verifies the complete data flow from capture to storage, ensuring recall mechanisms return accurate scores and handle errors correctly.