LINZA Use Cases

LINZA helps when an agent has material to read and verify, but there is no safe basis for writing conclusions into notes, rules, or memory immediately.

Project Folder

A project folder usually contains README files, plans, decisions, error logs, release drafts, and technical discussions. LINZA indexes Markdown, shows a workspace map, finds close fragments, and explains relations through connect.

This lets the agent answer from concrete files and show the local context behind the answer.

Incoming Material

Through agent_workspace(action="ingest_artifacts"), an agent can pass articles, logs, chats, JSON, DOCX, XLSX, and extracted PDF text. LINZA stores them as local artifacts, chunks them, and prepares review candidates through analyze_inbox and review_next.

Text inside an artifact is stored as data. It does not become an instruction for the agent.

Context For A New Session

search_memory searches stored artifacts and accepted memory. export_context builds a short Markdown packet with matching fragments. The packet can go to another model, another agent, or the next session.

Agent Work Review

record_trace accepts a structured work trace: task, result, tools, changed files, checks, errors, and context spend. analyze_trace computes metrics, and review_calibr returns cards for process learning.

calibr cards record sidecar decisions only after review/apply. They do not edit active skills, rules, code, or notes directly.