Context
AMS's architecture already appears designed to work against any GitHub repository, independent of whether that repo has ORB installed or participates in gittensor at all:
packages/loopover-miner/lib/self-review-context.js's own header comment states it rebuilds the gate-prediction context "from live GitHub data instead of a DB round-trip, since the miner has no database" — reading a target repo's .loopover.yml directly from GitHub, with only two fields (bounties, issueQuality) degrading gracefully when unavailable.
loopover-miner discover <owner/repo> accepts an arbitrary owner/repo target directly, not just a gittensor-registered pool.
resolveGitHubToken (packages/loopover-miner/lib/github-token-resolution.js) checks a plain GITHUB_TOKEN env override before touching any loopover-hosted session/auth endpoint — a self-hoster can run AMS with zero calls to loopover's backend at all.
This all reads correctly from static code review, but nobody has actually verified it end-to-end against a real, non-gittensor GitHub repo — the existing docs (docs.miner-workflow, docs.quickstart) don't explicitly confirm or deny standalone usage either way, and no issue currently tracks this as a checked, working path. This issue is the verification-and-documentation pass to close that gap: confirm the standalone story actually works as designed, find any real friction, and make the "ORB is optional" story explicit and discoverable.
Requirements
- Pick 2-3 small, real, public, non-gittensor-participating GitHub repositories (your own test repos, or well-known small OSS projects you have legitimate contribution standing on — do not target a repo you don't have a genuine reason to open a PR against; a personal sandbox repo is the safest choice) that have no
.loopover.yml and are not gittensor-registered.
- Using only a plain
GITHUB_TOKEN (no loopover-mcp login, to test the fully-standalone path first), run the full AMS loop against at least one such repo: discover <owner/repo> → plan → code → open a PR (or a dry-run equivalent if you don't want to open a real PR against someone else's repo — a sandbox repo you own is the cleanest way to test the real open-PR path).
- Separately, verify the
loopover-mcp login device-flow path also works end-to-end for git operations against a non-gittensor repo (confirm resolveGitHubToken's session-fetch path, not just the GITHUB_TOKEN override path).
- Document every actual step taken, every command run, and the exact observed behavior — including anything that didn't work as the static code review above predicted.
- If you find a genuine blocker or bug during this process, do NOT attempt to redesign or architect a fix as part of this issue. File a separate, narrowly-scoped follow-up issue describing exactly what broke and where, and reference it from this issue's PR. This issue's own deliverable is the verification + documentation, not a fix for whatever you find.
- Update
apps/loopover-ui/src/routes/docs.miner-workflow.tsx and/or docs.quickstart.tsx (whichever fits better) to explicitly state that AMS works against any GitHub repository and does not require ORB, gittensor registration, or a .loopover.yml file to function — cite the actual verified behavior, not the architectural intent.
Test Coverage Requirements
This is primarily a manual verification + docs task. If the verification process surfaces a genuinely untested code path (e.g. the standalone .loopover.yml-absent branch of self-review-context.js), add a regression test for that specific gap — 99%+ Codecov patch coverage applies to any src/**/packages/** test additions, but this issue does not require new source code changes unless a real gap is found and a minimal, obviously-safe fix is the right size for this issue (a docs/comment fix, not a redesign).
Deliverables
Expected Outcome
AMS's "works on any repo, ORB and gittensor optional" story is verified against real usage (not just inferred from reading the code) and clearly documented for a first-time user who has never touched gittensor.
Links & Resources
packages/loopover-miner/lib/self-review-context.js, discover-cli.js, github-token-resolution.js
apps/loopover-ui/src/routes/docs.miner-workflow.tsx, docs.quickstart.tsx
Context
AMS's architecture already appears designed to work against any GitHub repository, independent of whether that repo has ORB installed or participates in gittensor at all:
packages/loopover-miner/lib/self-review-context.js's own header comment states it rebuilds the gate-prediction context "from live GitHub data instead of a DB round-trip, since the miner has no database" — reading a target repo's.loopover.ymldirectly from GitHub, with only two fields (bounties,issueQuality) degrading gracefully when unavailable.loopover-miner discover <owner/repo>accepts an arbitraryowner/repotarget directly, not just a gittensor-registered pool.resolveGitHubToken(packages/loopover-miner/lib/github-token-resolution.js) checks a plainGITHUB_TOKENenv override before touching any loopover-hosted session/auth endpoint — a self-hoster can run AMS with zero calls to loopover's backend at all.This all reads correctly from static code review, but nobody has actually verified it end-to-end against a real, non-gittensor GitHub repo — the existing docs (
docs.miner-workflow,docs.quickstart) don't explicitly confirm or deny standalone usage either way, and no issue currently tracks this as a checked, working path. This issue is the verification-and-documentation pass to close that gap: confirm the standalone story actually works as designed, find any real friction, and make the "ORB is optional" story explicit and discoverable.Requirements
.loopover.ymland are not gittensor-registered.GITHUB_TOKEN(noloopover-mcp login, to test the fully-standalone path first), run the full AMS loop against at least one such repo:discover <owner/repo>→ plan → code → open a PR (or a dry-run equivalent if you don't want to open a real PR against someone else's repo — a sandbox repo you own is the cleanest way to test the real open-PR path).loopover-mcp logindevice-flow path also works end-to-end for git operations against a non-gittensor repo (confirmresolveGitHubToken's session-fetch path, not just theGITHUB_TOKENoverride path).apps/loopover-ui/src/routes/docs.miner-workflow.tsxand/ordocs.quickstart.tsx(whichever fits better) to explicitly state that AMS works against any GitHub repository and does not require ORB, gittensor registration, or a.loopover.ymlfile to function — cite the actual verified behavior, not the architectural intent.Test Coverage Requirements
This is primarily a manual verification + docs task. If the verification process surfaces a genuinely untested code path (e.g. the standalone
.loopover.yml-absent branch ofself-review-context.js), add a regression test for that specific gap — 99%+ Codecov patch coverage applies to anysrc/**/packages/**test additions, but this issue does not require new source code changes unless a real gap is found and a minimal, obviously-safe fix is the right size for this issue (a docs/comment fix, not a redesign).Deliverables
GITHUB_TOKEN.loopover-mcp login's device-flow session instead.docs.miner-workflow.tsxand/ordocs.quickstart.tsxupdated to explicitly state AMS's standalone capability.Expected Outcome
AMS's "works on any repo, ORB and gittensor optional" story is verified against real usage (not just inferred from reading the code) and clearly documented for a first-time user who has never touched gittensor.
Links & Resources
packages/loopover-miner/lib/self-review-context.js,discover-cli.js,github-token-resolution.jsapps/loopover-ui/src/routes/docs.miner-workflow.tsx,docs.quickstart.tsx