Skip to content

Add a queue-throughput quickstart for high-volume personal repos #61

Description

@coygeek

Summary

DeployBot's README explains the full durable delivery controller, but the fastest path for a personal repo with many independent PRs is not obvious. The docs would benefit from a focused "high-throughput personal repo" quickstart that starts with queue coordination before full production release tracking.

This is distinct from a full first-run onboarding skill: it is a concise operator recipe for the common problem "I create PRs faster than GitHub/manual merging can land them."

Problem

The README covers many advanced concepts:

  • durable deploy intent;
  • exact-head authorization;
  • event worker setup;
  • release admission;
  • integration PRs;
  • deployment workflow dispatch;
  • health checks;
  • native thread notifications;
  • unpause and repair ownership.

That completeness is useful, but a first-time user trying to solve queue stalls may not know which minimum settings matter for throughput.

The key decisions are scattered:

  • release_admission = "merged" is the low-latency mode.
  • integration.mode = "overlap" is a reasonable starting point before all.
  • queue.required_checks must be exact PR check display names.
  • pipeline.ci_workflows must be workflow names.
  • deployment workflows are optional for initial queue behavior but matter for merge-to-live metrics and full release receipts.

Proposed documentation

Add a short README section or doc page:

High-throughput personal repo quickstart

Suggested contents:

  1. Who this is for:
    • personal repos;
    • many small independent PRs;
    • user controls merge policy;
    • goal is avoiding accidental global queue stalls.
  2. Minimum pilot configuration:
[queue]
required_checks = ["<exact PR check name>"]
trusted_actors = ["@repository-owner"]
coordinator_actors = ["@repository-owner", "github-actions[bot]"]

[pipeline]
release_admission = "merged"
batch_settle_seconds = 0
promotion_workers = 4

[integration]
mode = "overlap"
max_batch_size = 3
  1. A warning that required_checks are PR check display names, not workflow names.
  2. A pilot procedure:
    • create 3 independent trivial PRs;
    • create 2 overlapping PRs;
    • prove independent PRs merge while overlaps route to integration/repair.
  3. A "do not use this if" section:
    • CI takes longer than PR production rate;
    • most PRs touch the same files;
    • every merge must be production-verified before the next merge;
    • deployment workflow identity is not configured yet but live receipts are expected.

Impact

This would help users evaluate DeployBot for the high-throughput queue problem without requiring them to absorb the entire release-controller model first.

It would also reduce misconfiguration risk by making the check-name/workflow-name split explicit in the first recipe users copy.

Acceptance criteria

  • README or docs include a high-throughput personal-repo quickstart.
  • The quickstart recommends release_admission = "merged" for independent-work throughput and explains the safety tradeoff.
  • The quickstart recommends integration.mode = "overlap" as the first high-throughput integration setting.
  • The quickstart explicitly distinguishes required check display names from workflow names.
  • The quickstart includes a small pilot plan with independent and overlapping PRs.
  • The quickstart states when DeployBot will not improve throughput.

Verification

  • Documentation tests pass.
  • Any docs-reference tests that require command/config coverage remain green.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions