Second Brain Demo

A canvas as an LLM agent's persistent memory, outside the chat window itself — each memory is one node, tagged by type (user/feedback/project/ reference). The constraint below enforces one rule an agent actually needs: a feedback/project memory always carries a **Why:** line, so a future session can judge an edge case instead of blindly following the rule with no sense of when it doesn't apply. Run meshfox check examples/second-brain.canvas.md to see it evaluate (all four nodes below are meant to pass).

for n in self.descendants():
    if "feedback" in n.tags or "project" in n.tags:
        if "**Why:**" not in n.text:
            fail(n.id + " (" + ",".join(n.tags) + "): missing a '**Why:**' line")

New to this repo's Rust half

user

Frontend engineer, ten years of React; this is their first time touching the Rust backend half of this repo. No **Why:** required — a user memory is a fact about the person, not a rule to apply selectively.

Edit .canvas.md via the meshfox MCP tools

feedback

Edit *.canvas.md files via the meshfox node_* MCP tools, not by hand-editing the raw Markdown.

Why: a hand-edit risks corrupting node ids, parent=, or meshfox:edge bookkeeping that the MCP tools keep consistent automatically. How to apply: for any *.canvas.md file, reach for canvas_open/ node_add/node_body/etc. first; hand-edit only prose inside an existing node's body.

Reporting pipeline rewrite is retention-driven

project

The reporting pipeline rewrite is driven by a data-retention deadline, not a performance goal.

Why: legal set a cutoff for how long raw event logs may be kept, and the current pipeline reads directly from that raw table. How to apply: any design choice in the rewrite should favor finishing before the retention cutoff over squeezing out extra throughput.

Incidents tracked in Linear, not GitHub Issues

reference

Incidents and outages are tracked in the "OPS" Linear project, not GitHub Issues. No **Why:** required — a reference memory is a pointer to an external system, not a rule.