ForgeRank
GitHub

Transparent repository intelligence

Rank what you watch.
Keep the why.

ForgeRank compares momentum, engineering activity, and maturity inside a disclosed observed cohort. Every score keeps its evidence, freshness, and confidence attached.

Explore sample data Run locally
  • 01 No GitHub token or OAuth
  • 02 Missing evidence stays missing
  • 03 Every rank names its cohort
EVIDENCE OBSERVATORY BOUNDARY VISIBLE
0 GitHub API
credentials
Selected pages HTTPS Git Own snapshots
Credentials0GitHub tokens required
Score model6inspectable dimensions
Evidence4confidence levels
Ranking scope1disclosed observed cohort

Product / not another star counter

Ask what is moving—and inspect the answer.

Start with a ranked cohort, move into the Momentum Matrix, then open the exact score reasons and retained history behind a repository.

ForgeRank / isolated sample mode SYNTHETIC VALUES
ForgeRank home dashboard showing repository rankings and discovery

The guided demo uses fictional repositories and synthetic values. It demonstrates the interface and evidence model; it is not a current measurement of any real project.

Three questions / one evidence trail

Reach. Motion. Engineering.
Never collapsed into one mystery number.

01 / MOMENTUM

Is attention compounding—or merely large?

The Momentum Matrix separates observed reach from versioned movement. Comparable snapshot windows power every point; incomplete history is withheld.

  • Fixed 0–100 momentum and logarithmic observed popularity
  • Shareable cohort filters and visible denominators
  • No popularity fallback when history is insufficient
ForgeRank Momentum Matrix plotting repository popularity against momentum
02 / SCORE REASONS

A score you can disagree with intelligently.

Each dimension retains the observed inputs, evidence status, score version, confidence multiplier, and calculation time.

Inspect the formula →
03 / IDENTITY BOUNDARY

Project evidence without identity shortcuts.

Git authors are not silently relabeled as public accounts. Developer views require confirmed public profiles and explicit non-fork portfolio evidence.

Read the boundary →

Method / evidence first

Observe less.
Explain more.

ForgeRank deliberately trades universal coverage for a bounded, reviewable pipeline. A rank always describes the repositories observed at that calculation time.

  1. 01Identify

    Validated public identifiers enter a deduplicated queue.

  2. 02Observe

    Public pages that pass the current exact-URL robots rules and bounded HTTPS Git reads provide evidence.

  3. 03Preserve

    Append-only snapshots retain time, parser, source hash, and confidence.

  4. 04Explain

    Versioned formulas persist one reason per score dimension.

0

Credential boundary

No REST, GraphQL, token, OAuth, or hidden endpoint.

Collection is cached, budgeted, and stopped on robots denial, throttling, or uncertainty. Avoiding API credentials does not mean unlimited access or complete GitHub coverage.

Read the source boundary →

Repository score / v1

Six dimensions. Fixed scale.

100
Impact25
Momentum20
Health20
Community15
Engineering10
Trust10

The sum is multiplied by observed confidence. Missing evidence contributes nothing.

First useful screen / about two minutes

Start with the isolated demo.

The demo builds a dedicated PGlite database with four fictional repositories and synthetic history. It never touches normal local data and performs no network collection.

Install the operations CLI from npm →
DEMO QUICK START 
git clone https://github.com/OthmaneBlial/ForgeRank.git
cd ForgeRank
corepack enable
pnpm install --frozen-lockfile
pnpm demo
Opens on 127.0.0.1:3001 · Node.js 22+ · no GitHub token

Coverage truth

A real fresh install begins with identifiers, not analytics. Rankings cover ForgeRank’s observed corpus—not every repository on GitHub. Blocked, stale, or insufficient evidence is shown as unavailable instead of estimated.

Try the decision surface

See the rank.
Interrogate the evidence.

Open the guided demo Star on GitHub