fix(apify): correct token URL, poll run --wait, add live --max-cost watchdog, reachable abort

[michegz]

Summary

Four fixes surfaced by dogfooding apify-pp-cli against a real Actor (apify/facebook-marketplace-scraper), including a cost-safety gap where a --wait run reached $2.63 before it could be manually aborted.

Findings

ID	Type	Fix
F1	bug	auth setup/doctor pointed to an MDN page; now points to https://console.apify.com/settings/integrations (the correct URL was already present in the spec's auth prose)
F2	bug	run --wait sent waitForFinish=60 on the start POST, exceeding the 30s HTTP client read timeout → "context deadline exceeded while reading body" while the run kept billing. Now caps the server-side block under the client timeout and relies on the existing GET poll loop
F3	bug	--max-cost could not protect a first run (no cached history to project from) and nothing capped a live run. Added a live watchdog: under --wait, polls usageTotalUsd and aborts the run once cost exceeds the cap
F4	bug	actor-runs abort <runId> only printed parent help; now accepts a run id directly and delegates to the actor-run-post leaf

Changes

Verification

• go build, go vet, govulncheck, manifest, phase5: PASS
• Behaviorally verified: auth setup prints the Apify console URL; actor-runs abort [runId] accepts a run id; run --wait polls under a bounded server block

Note: a pre-existing verify-skill canonical-sections drift in SKILL.md (untouched by this PR — the CLI was generated on press v4.8.0) remains and needs a regen, separate from these fixes.

Patch record

.printing-press-patches/auth-url-wait-cost-watchdog-abort.json — includes a deferred_to_upstream entry noting F1 is also a generator auth-URL-extraction heuristic bug worth fixing at the machine level.

[greptile-apps]

Greptile Summary

This PR fixes four bugs surfaced by dogfooding the apify-pp-cli against a real Actor: corrects the API-token URL from MDN to the Apify console across all auth-hint sites (F1), caps the waitForFinish server-side block to stay under the HTTP client timeout so run --wait no longer deadlines mid-body (F2), adds a live cost watchdog that polls usageTotalUsd and aborts the Actor once it crosses --max-cost (F3), and makes actor-runs abort <runId> work directly instead of printing parent help (F4).

• F1 (URL fix): Seven files updated consistently; MDN reference replaced with https://console.apify.com/settings/integrations in auth, doctor, helpers, mcp/tools, manifest, and tools-manifest.
• F2 + F3 (run.go): waitForFinish is now int(flags.timeout.Seconds()) - 5 (clamped ≥ 0), preventing the POST from blocking past the client read deadline; pollRunUntilTerminal gains a maxCost parameter and aborts the run once usageTotalUsd > maxCost, with terminal-status checked first to avoid spurious aborts on naturally-finishing runs.
• F4 (actor-runs_abort.go): Parent abort command now accepts a runId positional arg and delegates to the child actor-run-post RunE; --gracefully is forwarded by explicitly calling child.Flags().Set() before delegation, which correctly updates the pflag-bound closure variable.

Confidence Score: 5/5

Safe to merge; all four fixes are targeted and well-reasoned, with no regressions introduced in the shared poll loop.

The four bug fixes are narrow and correct: the URL replacements are mechanical, the waitForFinish cap removes a real client-timeout failure, the cost watchdog checks terminal status before cost (avoiding spurious aborts), and the abort shorthand leverages pflag pointer-binding to forward --gracefully correctly. The only open items are two style-level observations — aborted runs not being recorded in local history, and waitSecs not capped at 60 for unusually large HTTP client timeouts — neither of which affects correctness for the common default-30s-timeout case.

run.go warrants the most attention: the waitSecs formula and the cost-watchdog abort path. actor-runs_abort.go is also worth a second read to confirm the pflag pointer behaviour matches expectations.

Important Files Changed

Filename	Overview
library/developer-tools/apify/internal/cli/run.go	Core run logic: fixes F2 (waitForFinish cap) and F3 (live cost watchdog). Terminal-status check correctly precedes cost check; abort path well-structured. Aborted runs are not recorded in local DB history (early return skips RecordActorRun), meaning cost-cap events don't inform future projections.
library/developer-tools/apify/internal/cli/actor-runs_abort.go	Fix F4: adds direct runId argument to the abort command. The --gracefully forwarding via child.Flags().Set() + pflag pointer-binding works correctly since the child RunE uses closure variables, not cmd.Flags(). --stdin is intentionally not forwarded.
library/developer-tools/apify/internal/cli/auth.go	Fix F1: corrects two occurrences of the MDN auth URL to the Apify console integrations page. Straightforward and correct.
library/developer-tools/apify/internal/cli/helpers.go	Fix F1: updates three auth-error hint URLs (HTTP 400/401/403 paths) from MDN to the Apify console. All occurrences across the error switch are updated consistently.
library/developer-tools/apify/internal/cli/doctor.go	Fix F1: corrects auth_key_url in the doctor report output. Single-line URL replacement.
library/developer-tools/apify/internal/mcp/tools.go	Fix F1 + key_url: updates four auth hint URLs (400/401/403 errors and context handler key_url) from MDN to Apify console. All occurrences consistently updated.
library/developer-tools/apify/internal/cli/ab.go	Adds maxCost=0 to the pollRunUntilTerminal call — disables the cost watchdog for ab-style runs, which is correct since they have no --max-cost flag.
library/developer-tools/apify/internal/cli/workflow_run.go	Adds maxCost=0 to pollRunUntilTerminal for workflow step execution — correct, workflow steps have no user-facing --max-cost flag.
library/developer-tools/apify/manifest.json	Fix F1: corrects the APIFY_TOKEN credential description URL from MDN to the Apify console.
library/developer-tools/apify/tools-manifest.json	Fix F1: updates key_url in auth section and the [More info] href in the API description from MDN to the Apify console. Both in-description occurrences corrected.

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant User
    participant CLI as apify-pp-cli
    participant Apify as Apify API

    User->>CLI: run --wait --max-cost 1.00 actor

    Note over CLI: waitSecs = int(flags.timeout.Seconds()) - 5

    CLI->>Apify: "POST /v2/acts/actor/runs?waitForFinish=waitSecs"
    Apify-->>CLI: RunData status RUNNING (after waitSecs)

    loop pollRunUntilTerminal every 5s until deadline
        CLI->>Apify: GET /v2/actor-runs/runId
        Apify-->>CLI: RunData with status and usageTotalUsd

        alt isTerminalStatus
            CLI->>User: return RunData
        else usageTotalUsd exceeds maxCost
            CLI->>Apify: POST /v2/actor-runs/runId/abort
            Apify-->>CLI: abort ACK
            CLI->>User: error cost exceeded cap aborted run
        else context done or deadline passed
            CLI->>User: error timeout or context cancelled
        end
    end

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
    participant User
    participant CLI as apify-pp-cli
    participant Apify as Apify API

    User->>CLI: run --wait --max-cost 1.00 actor

    Note over CLI: waitSecs = int(flags.timeout.Seconds()) - 5

    CLI->>Apify: "POST /v2/acts/actor/runs?waitForFinish=waitSecs"
    Apify-->>CLI: RunData status RUNNING (after waitSecs)

    loop pollRunUntilTerminal every 5s until deadline
        CLI->>Apify: GET /v2/actor-runs/runId
        Apify-->>CLI: RunData with status and usageTotalUsd

        alt isTerminalStatus
            CLI->>User: return RunData
        else usageTotalUsd exceeds maxCost
            CLI->>Apify: POST /v2/actor-runs/runId/abort
            Apify-->>CLI: abort ACK
            CLI->>User: error cost exceeded cap aborted run
        else context done or deadline passed
            CLI->>User: error timeout or context cancelled
        end
    end

Loading

_{Reviews (3): Last reviewed commit: "fix(apify): forward --gracefully through..." | Re-trigger Greptile}

[michegz]

Thanks @greptile-apps — good catch on the watchdog ordering. Fixed in 1e2b0a6: isTerminalStatus is now checked before the cost watchdog in pollRunUntilTerminal, so a run that completes naturally with a final cost just over the cap returns its successful result instead of triggering a no-op abort against an already-finished run. The watchdog now only fires on a still-running Actor. Build + vet green.

On the secondary note (F4 --gracefully not forwarded to the abort <runId> shorthand): intentional — the shorthand is a fast-stop convenience; the full actor-runs abort actor-run-post <runId> --gracefully path remains for graceful aborts.

[greptile-apps]

The ordering fix in 1e2b0a6 looks correct — terminal check first eliminates the race where a naturally-completed run with a final cost tick over the cap would hit a 404/no-op abort and potentially surface a confusing error to the user.

On --gracefully: makes sense as a design choice. The shorthand communicates "stop now" semantics; users who need a clean drain have the explicit path. Appreciate the clarification.

[michegz]

Ready for maintainer review whenever you have a moment 🙏

Quick status: CI is green, mergeStateStatus: CLEAN / MERGEABLE, Greptile reviewed at 5/5 "safe to merge," and all review threads are resolved (the cost-watchdog ordering and the --gracefully shorthand were both addressed). No outstanding changes on my end.

Scope is four dogfood-surfaced fixes to apify-pp-cli: correct token URL (Apify console vs MDN), run --wait no longer exceeds the client read timeout, a live --max-cost watchdog that protects first runs, and a directly-reachable actor-runs abort <runId>. Patch record + rationale are in .printing-press-patches/auth-url-wait-cost-watchdog-abort.json. Thanks!