v0.1.1 — agent graphs

pip install waku-agent

What's new: waku/graph/

A stdlib graph engine, and the same job done two ways so you can read them against each other.

A loop discovers what to do next. A graph pre-determines what happens next.

waku brief	the morning briefing as a LOOP — one prompt, the model works out which tools to call
waku gather	the same job as a GRAPH — GitHub, web, calendar and memory fetched together, then one digest

Measured on a live run:

node_start  scan_github  scan_web  scan_calendar  scan_memory
node_end    scan_github    1500ms
node_end    scan_web        515ms
node_end    scan_calendar     5ms
node_end    scan_memory       6ms

Four starts before a single finish. The wave costs its slowest branch, not the sum.

The graph front door is off by default (WAKU_GRAPH_WORKFLOWS=1) and fails open to the plain loop, so it cannot make Waku worse. gather proposes and never acts — its model call passes no tool schemas, so it cannot send or merge anything; it drafts into the outbox for you to read.

Also: read-only GitHub via the gh CLI (WAKU_GH_TOOL=1, no token stored).

Fixed

v0.1.0's wheel shipped no skills — procedural memory was silently absent for anyone who installed from PyPI. Fixed and pinned by a test. Note that the v0.1.0 git tag predates the fix; v0.1.1 is the first tag whose code matches its wheel.

Thanks

@Tian-Timm, @wilenWang, @drakegriffith, @c436zhan, @likweitan, @luminary19, @epicnellson, @AllenHK, @neoneye.

Full changelog: v0.1.0...v0.1.1