Skip to content

Build the autonomous repeat/supervising loop (discover → attempt → manage → repeat) #5135

Description

@JSONbored

Problem: loop-reentry.js and loop-closure.js — the real-IO pieces that "automatically re-invoke discovery on a resolved outcome" and "summarize a completed cycle before considering re-entry" — are both CLI-unreachable (confirmed via import-graph traced from bin/gittensory-miner.js). The Dockerfile's own CMD is ["doctor"], a one-shot batch command, and its own comment concedes the loop only runs "unless an operator wraps the container in their own supervising loop." Nothing anywhere autonomously repeats the discover → attempt → manage cycle today — the "repeat" step of the system's own stated end-goal (discover → analyze → plan → prepare → create → manage → repeat) does not exist.

Area: AMS / Miner

Proposal: Build a real supervising loop (a CLI command or long-running process mode) that: runs discovery, attempts a claimed item via the CLI dispatch from the sibling issue, waits for its outcome via the existing manage poll, invokes loop-closure then loop-reentry, and repeats — checking the governor (persisted per the sibling issue) and the existing kill-switch at every iteration boundary.

Deliverables:

  • A working autonomous loop entrypoint that runs unattended across multiple discover→attempt→close→reenter cycles.

Acceptance criteria:

Boundaries:


Part of #5130.

Metadata

Metadata

Assignees

Labels

gittensor:featureGittensor-scored feature linked to a feature issue — scores a 0.25x multiplier.maintainer-onlyOwner-only work — yields no Gittensor points.

Projects

Status
Done

Relationships

None yet

Development

No branches or pull requests

Issue actions