Skip to content
All flows

rlar ​

RLAR, Ralph Loop with Agentic Review, has every round of work reviewed and stops when the reviewer agrees it is done. An actor works in one session that remembers; after each round a fresh reviewer holds what the actor said to the repository, and its review is the actor's next prompt. Use it when "is it done?" should not be answered by the agent that did the work.

flowverse v0.1.0

In hmz, open /flow → Flowverses → official and install rlar. Or run the release without installing anything:

hmz exec -f 'git+https://github.com/humanfia/rlar-flow@v0.1.0#rlar' …
release
v0.1.0 · commit 7694ab0
licence
Apache-2.0

Read from humanfia/flowverse when this site was built.

rlarsimulated

The whole run, at rest. Step through it with the buttons, or drag the bar.

Every review answers two things: whether the task is done, and what the actor hears next. The actor remembers every round; each review starts from nothing but the task, what the actor said and the repository.

The run, turn by turn
  1. actor — adds an undo stack; a session opened for this turn; hands reviewer in the tree: undo, no redo
  2. reviewer — reads the repo; a session opened for this turn; hands actor as words: false: no redo
  3. actor — adds redo; another turn of the session it already had; hands reviewer in the tree: undo and redo
  4. reviewer — reads it fresh; a session opened for this turn; hands actor as words: false: a stubbed test
  5. actor — writes a real test; another turn of the session it already had; hands reviewer in the tree: a real test
  6. reviewer — reads it fresh; a session opened for this turn; hands the finish as words: done: true

Then round again: while the review says there is more to do.

It ends when the reviewer says done, a turn fails, or the budget runs out.

See also: humanize1 · flame_chase · goal