marketing-shibata50/cmux
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge pull request #972 from manaflow-ai/task-browser-open-trailing-json-flags

Browse source 
Support trailing browser output flags
This commit is contained in:
Lawrence Chen 2026-03-05 22:39:14 -08:00 • committed by GitHub
commit 43f42b730c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 92 additions and 26 deletions
Download patch file Download diff file Expand all files Collapse all files

61
CLI/cmux.swift
View file

@ -2622,7 +2622,34 @@ struct CMUXCLI {
throw CLIError(message: "browser requires a subcommand")
}
let (surfaceOpt, argsWithoutSurfaceFlag) = parseOption(commandArgs, name: "--surface")
var effectiveJSONOutput = jsonOutput
var effectiveIDFormat = idFormat
var browserArgs = commandArgs
// Browser-skill examples often place output flags at the end of the command.
// Strip trailing display flags so they don't become part of a URL or selector.
while !browserArgs.isEmpty {
if browserArgs.last == "--json" {
effectiveJSONOutput = true
browserArgs.removeLast()
continue
}
if browserArgs.count >= 2,
browserArgs[browserArgs.count - 2] == "--id-format" {
let raw = browserArgs.last!
guard let parsed = try CLIIDFormat.parse(raw) else {
throw CLIError(message: "--id-format must be one of: refs, uuids, both")
}
effectiveIDFormat = parsed
browserArgs.removeLast(2)
continue
}
break
}
let (surfaceOpt, argsWithoutSurfaceFlag) = parseOption(browserArgs, name: "--surface")
var surfaceRaw = surfaceOpt
var args = argsWithoutSurfaceFlag
@ -2651,8 +2678,8 @@ struct CMUXCLI {
}
func output(_ payload: [String: Any], fallback: String) {
if jsonOutput {
print(jsonString(formatIDs(payload, mode: idFormat)))
if effectiveJSONOutput {
print(jsonString(formatIDs(payload, mode: effectiveIDFormat)))
return
}
print(fallback)
@ -2808,8 +2835,8 @@ struct CMUXCLI {
}
}
let payload = try client.sendV2(method: "browser.open_split", params: params)
let surfaceText = formatHandle(payload, kind: "surface", idFormat: idFormat) ?? "unknown"
let paneText = formatHandle(payload, kind: "pane", idFormat: idFormat) ?? "unknown"
let surfaceText = formatHandle(payload, kind: "surface", idFormat: effectiveIDFormat) ?? "unknown"
let paneText = formatHandle(payload, kind: "pane", idFormat: effectiveIDFormat) ?? "unknown"
let placement = ((payload["created_split"] as? Bool) == true) ? "split" : "reuse"
output(payload, fallback: "OK surface=\(surfaceText) pane=\(paneText) placement=\(placement)")
return
@ -2854,8 +2881,8 @@ struct CMUXCLI {
if subcommand == "url" || subcommand == "get-url" {
let sid = try requireSurface()
let payload = try client.sendV2(method: "browser.url.get", params: ["surface_id": sid])
if jsonOutput {
print(jsonString(formatIDs(payload, mode: idFormat)))
if effectiveJSONOutput {
print(jsonString(formatIDs(payload, mode: effectiveIDFormat)))
} else {
print((payload["url"] as? String) ?? "")
}
@ -2872,8 +2899,8 @@ struct CMUXCLI {
if ["is-webview-focused", "is_webview_focused"].contains(subcommand) {
let sid = try requireSurface()
let payload = try client.sendV2(method: "browser.is_webview_focused", params: ["surface_id": sid])
if jsonOutput {
print(jsonString(formatIDs(payload, mode: idFormat)))
if effectiveJSONOutput {
print(jsonString(formatIDs(payload, mode: effectiveIDFormat)))
} else {
print((payload["focused"] as? Bool) == true ? "true" : "false")
}
@ -2906,8 +2933,8 @@ struct CMUXCLI {
}
let payload = try client.sendV2(method: "browser.snapshot", params: params)
if jsonOutput {
print(jsonString(formatIDs(payload, mode: idFormat)))
if effectiveJSONOutput {
print(jsonString(formatIDs(payload, mode: effectiveIDFormat)))
} else {
print(displaySnapshotText(payload))
}
@ -3118,7 +3145,7 @@ struct CMUXCLI {
let sid = try requireSurface()
let (outPathOpt, _) = parseOption(subArgs, name: "--out")
let localJSONOutput = hasFlag(subArgs, name: "--json")
let outputAsJSON = jsonOutput || localJSONOutput
let outputAsJSON = effectiveJSONOutput || localJSONOutput
var payload = try client.sendV2(method: "browser.screenshot", params: ["surface_id": sid])
func fileURL(fromPath rawPath: String) -> URL {
@ -3233,7 +3260,7 @@ struct CMUXCLI {
}
if outputAsJSON {
let formattedPayload = formatIDs(payload, mode: idFormat)
let formattedPayload = formatIDs(payload, mode: effectiveIDFormat)
if var outputPayload = formattedPayload as? [String: Any] {
if hasText(screenshotPath) || hasText(screenshotURL) {
outputPayload.removeValue(forKey: "png_base64")
@ -3307,8 +3334,8 @@ struct CMUXCLI {
"styles": "browser.get.styles",
]
let payload = try client.sendV2(method: methodMap[getVerb]!, params: params)
if jsonOutput {
print(jsonString(formatIDs(payload, mode: idFormat)))
if effectiveJSONOutput {
print(jsonString(formatIDs(payload, mode: effectiveIDFormat)))
} else if let value = payload["value"] {
if let str = value as? String {
print(str)
@ -3347,8 +3374,8 @@ struct CMUXCLI {
throw CLIError(message: "Unsupported browser is subcommand: \(isVerb)")
}
let payload = try client.sendV2(method: method, params: ["surface_id": sid, "selector": selector])
if jsonOutput {
print(jsonString(formatIDs(payload, mode: idFormat)))
if effectiveJSONOutput {
print(jsonString(formatIDs(payload, mode: effectiveIDFormat)))
} else if let value = payload["value"] {
print("\(value)")
} else {

57
tests_v2/test_browser_cli_agent_port.py
View file

@ -91,6 +91,32 @@ def _run_cli_text(cli: str, args: list[str], retries: int = 3) -> str:
raise cmuxError(f"CLI failed ({' '.join(args)}): {last_merged}")
def _run_cli_tail_json(cli: str, args: list[str], retries: int = 3) -> dict:
last_merged = ""
for attempt in range(1, retries + 1):
proc = subprocess.run(
[cli, "--socket", SOCKET_PATH] + args,
capture_output=True,
text=True,
check=False,
)
if proc.returncode == 0:
try:
return json.loads(proc.stdout or "{}")
except Exception as exc: # noqa: BLE001
raise cmuxError(f"Invalid CLI JSON output for {' '.join(args)}: {proc.stdout!r} ({exc})")
merged = f"{proc.stdout}\n{proc.stderr}".strip()
last_merged = merged
if "Command timed out" in merged and attempt < retries:
time.sleep(0.2)
continue
raise cmuxError(f"CLI failed ({' '.join(args)}): {merged}")
raise cmuxError(f"CLI failed ({' '.join(args)}): {last_merged}")
def _run_cli_expect_failure(cli: str, args: list[str], needles: list[str]) -> None:
proc = subprocess.run(
[cli, "--socket", SOCKET_PATH, "--json"] + args,
@ -144,15 +170,6 @@ def main() -> int:
cli = _find_cli_binary()
with _local_test_server() as page_url:
opened = _run_cli_json(cli, ["browser", "open", page_url])
surface = str(opened.get("surface_ref") or opened.get("surface_id") or "")
_must(bool(surface), f"browser open returned no surface handle: {opened}")
_must(surface.startswith("surface:"), f"Expected short surface ref from browser open, got: {opened}")
_run_cli_json(cli, ["browser", surface, "wait", "--load-state", "complete", "--timeout-ms", "15000"])
snapshot_text = _run_cli_text(cli, ["browser", surface, "snapshot", "--interactive"])
_must("ref=e" in snapshot_text, f"Expected snapshot text with refs from CLI: {snapshot_text!r}")
identify = _run_cli_json(cli, ["identify"])
focused = identify.get("focused") or {}
workspace = str(
@ -163,6 +180,28 @@ def main() -> int:
or ""
)
_must(bool(workspace), f"Expected workspace handle from identify: {identify}")
os.environ["CMUX_WORKSPACE_ID"] = workspace
opened_tail_json = _run_cli_tail_json(
cli,
["browser", "open", page_url, "--workspace", workspace, "--id-format", "both", "--json"],
)
tail_surface = str(opened_tail_json.get("surface_ref") or "")
_must(tail_surface.startswith("surface:"), f"Expected trailing --json browser open to return surface_ref: {opened_tail_json}")
_must(bool(opened_tail_json.get("surface_id")), f"Expected trailing --id-format both to preserve surface_id: {opened_tail_json}")
_must("--json" not in str(opened_tail_json.get("url") or ""), f"Trailing output flags leaked into browser open URL: {opened_tail_json}")
_run_cli_json(cli, ["browser", tail_surface, "wait", "--load-state", "complete", "--timeout-ms", "15000"])
tail_url_payload = _run_cli_json(cli, ["browser", tail_surface, "url"])
_must(str(tail_url_payload.get("url") or "").startswith(page_url), f"Expected trailing --json browser open to navigate: {tail_url_payload}")
opened = _run_cli_json(cli, ["browser", "open", page_url])
surface = str(opened.get("surface_ref") or opened.get("surface_id") or "")
_must(bool(surface), f"browser open returned no surface handle: {opened}")
_must(surface.startswith("surface:"), f"Expected short surface ref from browser open, got: {opened}")
_run_cli_json(cli, ["browser", surface, "wait", "--load-state", "complete", "--timeout-ms", "15000"])
snapshot_text = _run_cli_text(cli, ["browser", surface, "snapshot", "--interactive"])
_must("ref=e" in snapshot_text, f"Expected snapshot text with refs from CLI: {snapshot_text!r}")
blank_opened = _run_cli_json(cli, ["browser", "open", "about:blank", "--workspace", workspace])
blank_surface = str(blank_opened.get("surface_ref") or blank_opened.get("surface_id") or "")