DocAILab Critic Component
Provide the critic capability for an Agentic RAG slot.
Interface
Input CritiqueRequest:
query: original questiondocuments: selected evidence documentsanswer: generated answer- optional
max_tokens
Output CritiqueResult:
approved: whether the answer is supported and responsivescore: quality score between 0 and 1feedback: concise overall critiqueissues: list of concrete problems, possibly empty
Execution
Run scripts/component.py:run(inputs, context). The runtime must provide the frozen context.call_model(...). This Component critiques an answer only; it does not retrieve documents, generate a replacement answer, or select another Skill.