Retrieve local evidence and generate crystallized answers for knowledge questions.
npx skills add https://github.com/mengsi16/brain-base --skill qa-workflow-mengsi16
Please help me install this Agent Skill. Skill: qa-workflow Source: https://github.com/mengsi16/brain-base/tree/main/skills/qa-workflow Command: npx skills add https://github.com/mengsi16/brain-base --skill qa-workflow-mengsi16
The QA workflow orchestrates end-to-end knowledge QA by prioritizing locally stored data and crystallized answers, ensuring traceability to source chunks and reducing dependency on external fetches.
Ask a knowledge question that should be answered from local data; the workflow will first check crystallized answers, rewrite queries, retrieve local evidence, assess sufficiency, trigger Get-Info if needed, generate an evidence-based answer, and hand off to organize-agent for crystallization.
High-intent search queries and answers about installing and using this skill.
The QA workflow solves local-first knowledge QA by rewriting queries, retrieving local evidence chunks, and validating sufficiency before answering. It prioritizes existing data to ensure traceability and reduces dependency on external fetches.
Knowledge crystallization is a process that saves successful reasoning and execution traces for future reuse. After generating an evidence-based answer, the workflow delegates to an organize-agent to crystallize the reasoning, enabling faster retrieval of validated answers for subsequent queries.
Yes, the workflow features dynamic decision-making that triggers a Get-Info Agent to augment local knowledge when retrieved evidence is insufficient or outdated, ensuring accurate question answering even if existing local data lacks necessary information.
Yes, the QA workflow handles questions about definitions, steps, comparisons, and configurations where local evidence suffices. It normalizes these inquiries and retrieves local raw data to validate evidence before generating a crystallized answer.
The primary limitation of local-first reasoning is that it depends entirely on the completeness of stored local data. If local evidence is outdated or missing, the workflow must trigger external fetches via the Get-Info Agent to augment knowledge before answering.
mengsi16Chunk documents, generate QA pairs, and persist to Milvus.
PieroSierraSynthesize answers to natural-language questions from a personal knowledge base.
mengsi16Query and persist local documents through a structured knowledge base workflow.
agilebydesignQuery the abd-answers Pinecone vector index for grounding chunks with similarity scores.
