Personal project · On-prem · Local LLM · Human-approved apply

Find the nulls. Inside your network.

Gridnull is a full-stack on-prem application I built: sync GitHub or GitLab issues, surface stale and stuck work, and optionally run local Ollama suggestions — without sending repo data to a vendor cloud. This site showcases the product; evaluation access is available on request, at your own risk.

  • On-prem
  • No cloud AI
  • GitHub + GitLab
  • Human in the loop
  • Invite-only

See the journey

A short walkthrough: project dashboard, local LLM suggestions, and admin overview.

gridnull · walkthrough

Product UI

gridnull · dashboard
Project — suggestions to review
gridnull · dashboard
Admin — jobs and instance health

The problem

Backlog blind spots stay invisible until they hurt

Backlogs hide real risk

Open issues pile up. Stale tickets and decaying milestones are hard to see in the board view alone.

Backlog hygiene is manual and repetitive

Duplicate issues, empty descriptions, and stuck tickets eat lead time every sprint.

Cloud AI is a hard sell

Security and platform teams often block tools that send issue text to external APIs.

The approach

Sync. Measure. Assist.

Three pillars, one on-prem deployment. You stay in control of data and every applied change.

Sync

Unified issue mirror

Pull issues, labels, and milestones from GitHub or GitLab into Postgres on a schedule or on demand.

Measure

Backlog metrics that matter

Stale issues, stuck issues, milestone decay — configurable thresholds per project.

Assist

Local AI, your approval

Ollama suggests duplicates and description drafts. Apply updates VCS only when a human confirms.

Why Gridnull

Built for how teams already work

Most engineering teams already live in GitHub or GitLab. Gridnull does not replace your board — it mirrors issue metadata into infrastructure you control and surfaces what sprint boards hide: what is stale, what is stuck, and which milestones are decaying.

Engineering leads

Null signals become measurable — stale, stuck, and milestone-decay metrics across projects, not gut feel in standup.

Platform / DevOps

Standard Docker Compose stack. No vendor SaaS holding issue text. Fits intranet and reverse-proxy setups you already run.

Developers

Same OAuth login as GitHub/GitLab. Command palette, project favorites, per-project thresholds — fast navigation without a new issue tracker.

Security & compliance

Issue metadata stays on your network. LLM workers read Postgres only. PATs encrypted at rest. Closed registration + email allowlist in production.

Mixed VCS shops

One dashboard for GitHub and GitLab — useful when teams split across hosts but platform wants one backlog view.

Metrics that match backlog reality

Stale, stuck, and milestone decay with per-project thresholds — not one-size-fits-all dashboards.

Local AI without the procurement fight

Ollama on your hardware suggests duplicates and thin descriptions. Security gets no external LLM API instead of another exception request.

Human-approved write-back

Suggestions are drafts. Apply pushes to GitHub/GitLab with your credentials; dismiss costs nothing.

Ops-grade from day one

Admin overview, background jobs, audit log, rate limiting, bootstrap + EULA — built for a team instance.

Honest scope

Not a funded startup SKU. Full functionality today; future CE/Pro packaging is a roadmap idea, not enforced in code.

Your issues never leave your network for AI — but your team finally sees which ones are quietly rotting.

A full on-prem product build: workers, encryption, release pipeline — built to solve real backlog hygiene, not sold as SaaS.

Who it's for

Teams who self-host — and people hiring for product engineering

Engineering teams evaluating self-hosted backlog visibility beside GitHub/GitLab

Platform teams who want Docker Compose, OAuth, and closed registration patterns

Security-conscious orgs that block cloud AI on issue text

Recruiters and hiring managers reviewing a full-stack product build

How it works

From install to insights in four steps

01

Install

After you contact me and we agree on evaluation access, you receive the gridnull-install bundle and GHCR pull instructions. Configure .env, pull gridnull-web / gridnull-worker, run migrations — no product git clone.

02

Bootstrap

Open /setup, accept the EULA, first OAuth sign-in becomes admin; invite teammates with roles.

03

Connect

Add a GitHub or GitLab connection (PAT), register projects, sync on demand or on a schedule.

04

Act

Review dashboard metrics, run local analysis, dismiss or apply suggestions.

A serious on-prem product exercise: metrics, queues, encryption, install bundle, and human-in-the-loop AI — built end to end by one engineer.

Curious how it works in practice?

Gridnull is not sold as a commercial product with SLA or support. If you want to evaluate it on your own infrastructure, get in touch — use a test repo and review every Apply.