npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@titan-design/review-panel

v0.9.1

Published

Review-panel types and the reviewer ports a caller satisfies

Readme

@titan-design/review-panel

Review-panel types and the reviewer ports a caller satisfies

Tier 2 of the titan-platform DAG. May import only packages in the same tier or below; the package-layers rule in .codewatch/check.json enforces this in CI.

Status: types, ports, classifyPr, planPanel, the reviewer briefs, the verdict acceptor, the aggregate and reviewPullRequest. Tracked by TP-1916.

acceptVerdict(input, messages) decides whether a reviewer's final message is its verdict for this PR at this head: the dispatched agent and session, written after dispatch, a block that names the target, and at least one investigative call (isInvestigativeCall) unless the reader could not count. Anything else is { kind: "none" }, with a malformed record (readMalformed) when the block was refused or named another target. A FIX_FIRST carries its findings, bounded by boundedFindings; a verdict carries the reviewer's OWNER-BRIEF block as parseOwnerBrief reads it.

A reviewer may propose non-blocking work in a Follow-ups: section, one - <title> | <done_when> | low|medium|high line per task, at most five, before or after the verdict lines. parseFollowUps(text) reads it outside fences and comments; any fault drops the whole block with one named warning (FOLLOW_UPS_WARNINGS). acceptVerdict carries the items as followUps and a dropped block as followUpsWarning, on MERGE and FIX_FIRST alike, and neither ever changes the verdict. reviewerBrief teaches the section through FOLLOW_UPS_RULE.

classifyPr(facts, rules?) returns the review class (g10 or standard) and the touch flags, from signals alone. DEFAULT_CLASS_RULES holds the glob tables and the large threshold (400 lines; more than 12 files when a file lacks line counts); pass rules to override them. changedLineCount(files, rules?) is the size it compares, with the generated globs (registry files a script rewrites) left out, unless those lines alone pass largeLines or a rename brought the file in from a non-generated path. pathTouches(files, rules?) is the path-only part of those touches, with no kind or caller flag, so a caller can tell a sensitive path from a kind: "security" that classifyPr still forces. A rename's old path counts too, so renaming a guard file away keeps its touch. A test file (the test globs) never counts toward the authority, policy or security touches; it still counts toward the others, such as perf.

planPanel(cls, policy, headroom) plans the panel for one head: each member's shape, profile, brief id and blocking flag, and a spend estimate in points. It plans at most 3 members and 1 premium member, and drops none on cost. Premium is the price class sonnetFor lists (bd-reviewer and fable-reviewer by default), and every premium member costs points.opus. A rule with uncapped: true (a hold's adversary) keeps its premium profile beside the capped seat. When headroom.opus is false, a premium member is planned at its sonnet profile with degraded: true. With policy.installed, a lowering to a profile that is not installed throws UninstalledProfileError. DEFAULT_PANEL_POLICY has no panel table, so it plans the correctness member alone at the g10 or standard profile (bd-reviewer, reviewer); set panel: DEFAULT_PANEL_TABLE for the class table.

aggregate(plan, results, input) is the panel verdict for input.head. Any blocking FIX_FIRST blocks; MERGE needs every blocking member's MERGE at that head; a blocking member with no verdict is no-verdict (or timeout), never MERGE. Advisory findings ride along only when every path:line they cite exists in input.source; blocking findings are kept as written. satisfiesG10 is false when any member is degraded.

shapeBrief(shape, input) is the brief for an adversary, tests, visual or perf member: its overlay from SHAPE_BRIEFS, then the unchanged reviewerBrief. Each overlay has a stable id (the shape name, which is planPanel's default brief id) and a sha256 hash, so an edited overlay is a new variant. Every overlay points at the base brief's test rule for where checks run instead of restating it. The tests brief reads input.fixProof, a fix-proof/v1 result line: it uses the verdict, counts and flags for this head, never test names or paths, and ignores a line that does not parse or names another head. An error verdict reads as blocking. reviewerBrief also takes fixProofUnverified (error, untrusted or unread), which Shepherd sets when it could not verify the proof: the brief then says FIX_FIRST unless the reviewer names an added test that fails on the base, and the reason picks a fixed sentence rather than reaching the prompt.

reviewPullRequest(request, ports, opts?) is the one entry point over those parts: classify, plan, spawn every member in parallel, wait through acceptVerdict, aggregate. Members are named by memberName(target, shape), so a repeat call with the same input reuses the roster rows and spawns nothing. A refused opus spawn is retried once at sonnet and marked degraded; a malformed reply gets one correction per member through resume; a silent or uncorrected member gets one fresh reviewer under the -r2 name, then counts as timeout; a usage-limit notice is final. Every side effect is a port: dispatch, reader, clock and headroom.