feat(muse-proxy): tolerate per-CLI flag differences via --help capability probe by dezren39 · Pull Request #775 · developing-today/code

dezren39 · 2026-10-24T17:41:04Z

Stacked on #769 (base: warm/muse-proxy-serve-pool). Left open for review; not merged.

Problem

The warm muse-proxy from #769 spawns muse serve --provider meta --disable-request-user-input and its exec fallback passes --disable-reminders. The fleet runs two muse CLI releases side by side whose flag surfaces differ (both --help outputs captured and compared 2026-10-24):

flag	muse 1.4.2 (laptop)	muse 1.4.13 (sandbox)
serve `--disable-request-user-input`	absent	present
exec `--disable-reminders`	absent	present
everything else the proxy passes	present	present

An unknown option kills the child before it serves anything, so on the laptop every muse turn failed instantly (HTTP 500) while the sandbox worked.

Fix (commit `ccb2cac`)

At startup the proxy probes the resolved binary's serve --help and exec --help once, caches the advertised --flag sets, and builds both spawn argvs from them — only advertised flags are passed. No version strings are compared, so future CLIs land wherever their help says.
Probe failure degrades instead of dying: optional flags (the two above) are omitted, core flags are still passed.
/health now reports cliFlags: { serveProbed, execProbed, dropped: { serve, exec } }.
Behavior on 1.4.13 is unchanged: every planned flag is advertised there, so the argvs are byte-identical to feat(muse-proxy): warm sessions on a persistent muse serve (MSP) host (#480) #769's.

Also in this PR (commit `3955bc5`)

The laptop deployment carried one local adaptation the sandbox build lacked: the egress pin was overridable (MUSE_PROXY_EGRESS, empty = no pinning; CA bundle pinned only when the file exists). That is folded into the fleet code verbatim so both machines run the same file; sandbox behavior is unchanged (same default pin, CA present).

Tests

node --test bin/muse-proxy.test.mjs — 13/13 pass (9 pre-existing + 4 new). The stub muse gains canned --help surfaces (1.4.2-shaped by default, 1.4.13-shaped with STUB_HELP_ALL_FLAGS=1) and argv logging; the new tests pin omission/pass-through for both the exec spawn and the warm serve spawn, plus the /health dropped-flag report.

^{Need help on this PR? Tag @codesmith-bot with what you need.}

The sandbox can only reach Meta through its unauthenticated egress proxy + CA bundle, but fleet machines with direct network access (the laptop) must not pin either for the CLI children. MUSE_PROXY_EGRESS now overrides the pinned proxy URL; setting it empty disables the pinning entirely, and the CA bundle is only pinned when the file exists. Sandbox behaviour is unchanged (same default pin, CA present). This folds the laptop deployment's local adaptation into the fleet code so both machines run the same file.

The fleet runs muse CLIs with different flag surfaces side by side: 1.4.2 (laptop) advertises neither serve --disable-request-user-input nor exec --disable-reminders, and an unknown option kills the child before it can serve anything — the warm proxy deployed there failed every turn instantly. 1.4.13 (sandbox) advertises both. Probe `serve --help` / `exec --help` on the resolved binary once at startup, cache the advertised flag sets, and build both spawn argvs from them. No version strings are compared. If a probe fails, optional flags are omitted and core flags are still passed, so the proxy degrades instead of dying. /health reports the probe state and the dropped flags per subcommand. Tests: the stub muse gains canned --help surfaces (1.4.2-shaped by default, 1.4.13-shaped with STUB_HELP_ALL_FLAGS=1) and argv logging; four new tests pin the omission/pass-through behaviour for both the exec and warm serve spawns. 13/13 pass.

coderabbitai · 2026-10-24T17:41:13Z

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 57 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

Run configuration

Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: defaults
Review profile: CHILL
Plan: Advanced
Run ID: 9b6cdf33-f013-41ea-bfa3-70a17ce9c076

Commits

Reviewing files that changed from the base of the PR and between ae8cc79 and ccb2cac.

Files selected for processing: 2

bin/muse-proxy.mjs
bin/muse-proxy.test.mjs

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

chatgpt-codex-connector · 2026-10-24T17:41:13Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-24T17:43:01.077193Z	`ccb2cac`	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ccb2cacebb

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-10-24T17:42:59Z

+  child.on('close', () => {
+    clearTimeout(timer);
+    const flags = new Set(out.match(/--[a-z][a-z0-9-]*/g) || []);
+    resolve(flags.size > 0 ? flags : null);


Reject failed capability probes before caching their flags

If either help process exits unsuccessfully or is killed after emitting a diagnostic containing a --flag, this handler treats that output as a successful capability probe. For example, a failed probe mentioning --help produces a capability set containing only that flag; the exec argument builder then removes every planned option, including --prompt-file and --json, so subsequent requests launch bare muse exec without the user's prompt. This incorrect capability set is cached until restart. Check the exit code and signal before parsing, and return null for failed or timed-out probes so the existing core-preserving fallback applies.

Useful? React with 👍 / 👎.

bin/muse-proxy.mjs conflict resolved by taking main's version: the live-tree commit a936e16 already carries this PR's warm serve pool plus the follow-up CLI flag-probe and MUSE_PROXY_EGRESS override (the superset later stacked as #775). This PR's test file merges cleanly and passes 9/9 against it.

Muse Sweep added 2 commits October 24, 2026 17:39

dezren39 requested a review from a team as a code owner October 24, 2026 17:41

github-actions Bot assigned dezren39 Oct 24, 2026

chatgpt-codex-connector Bot reviewed Oct 24, 2026

View reviewed changes

Base automatically changed from warm/muse-proxy-serve-pool to main October 25, 2026 11:45

dezren39 merged commit a91b6b3 into main Oct 25, 2026
4 checks passed

dezren39 deleted the warm/muse-proxy-cli-flag-tolerance branch October 25, 2026 11:46

Navigation Menu

Repository navigation

feat(muse-proxy): tolerate per-CLI flag differences via --help capability probe - #775

feat(muse-proxy): tolerate per-CLI flag differences via --help capability probe#775
dezren39 merged 2 commits into
mainfrom
warm/muse-proxy-cli-flag-tolerance

dezren39 commented Oct 24, 2026 •

edited by blacksmith-sh Bot

Loading

Uh oh!

coderabbitai Bot commented Oct 24, 2026 •

edited

Loading

Review limit reached

Uh oh!

chatgpt-codex-connector Bot commented Oct 24, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Oct 24, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

dezren39 commented Oct 24, 2026 • edited by blacksmith-sh Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Fix (commit ccb2cac)

Also in this PR (commit 3955bc5)

Tests

Uh oh!

coderabbitai Bot commented Oct 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review limit reached

Uh oh!

chatgpt-codex-connector Bot commented Oct 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Oct 24, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

dezren39 commented Oct 24, 2026 •

edited by blacksmith-sh Bot

Loading

Fix (commit `ccb2cac`)

Also in this PR (commit `3955bc5`)

coderabbitai Bot commented Oct 24, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 24, 2026 •

edited

Loading