stateful_ralph
Leave one agent on a task it has to remember. One session holds the whole run and is sent the task again every round, so the agent keeps every approach it has already ruled out, and --resume carries that session on. Reach for it when the work is exploratory and what has been tried is expensive to find out again; when the work is simply long, ralph_loop is the better choice.
In hmz, open /flow → Flowverses → official and install stateful_ralph. Or run the release without installing anything:
hmz exec -f 'git+https://github.com/humanfia/stateful-ralph-flow@v0.1.0#stateful_ralph' …- repository
- humanfia/stateful-ralph-flow
- release
- v0.1.0 · commit
afdfaa0 - licence
- Apache-2.0
Read from humanfia/flowverse when this site was built.
The whole run, at rest. Step through it with the buttons, or drag the bar.
Every round is sent the task again, in the one session that remembers what it has ruled out. One session is one conversation, so its context grows with every round — the thread thickens — and that is the other limit a long run of this reaches.
The run, turn by turn
- agent — profiles the parser; a session opened for this turn
- agent — rules out the cache; another turn of the session it already had
- agent — rules out the lexer; another turn of the session it already had
- agent — finds a held buffer; another turn of the session it already had
- agent — frees it, re-checks; another turn of the session it already had
Then round again: the task again, one round longer.
It ends when the budget runs out, or a turn fails.
See also: continue_loop · ralph_loop