DEV3 Technologies home

DEV3 Faultsnare – Self-Hosted Error Tracking with Source Maps, Alerts & SDKs (Next.js) with Gemini AI admin tools

Self-hosted error tracking on Next.js 16 and MongoDB with Gemini AI admin tools: grouped issues, source maps, releases, alert rules, browser, Node and PHP SDKs, Google Translate language switcher. Bring your own Gemini key. Setup wizard and demo data included. No mobile apps.
by DEV3 TechnologiesVerified Author(0) 0 sales 1 viewsUpdated 15 Nov 2026

Video tour. Pick a section below to browse 44 screenshots in 8 sections.

App & phone tours

Tap play to watch each app. Nothing downloads until you do.

  • Mobile web0:13

Faultsnare is an error tracking and log monitoring service that you host yourself, built by DEV3 Technologies on Next.js 16, React 19, TypeScript, Tailwind 4 and MongoDB, with a separate Node worker that processes events. Your apps send errors to a project with a DSN key; Faultsnare scrubs personal data, groups the errors into issues, shows de-minified stack traces, tracks releases and deploys, and alerts your team. Each customer gets a workspace with a monthly event quota; you, the operator, run the platform from an operator console. The demo shows fictional companies: an online shop, a harbour logistics firm and a small studio.

No live demo: 44 screenshots and two video tours show the marketing site, the issue pages, releases, alerts, projects and the operator console. This product is a web application: there are no mobile apps.

For the people fixing bugs

  • Projects and DSN keys: browser keys are public and limited to the origins you allow; server keys carry a secret. Rotate a key and the old one keeps working for seven days, so a deploy never breaks.
  • Issues: errors are grouped by fingerprint (exception type plus normalised in-app frames, or the message template). Search with is:unresolved release:4.19.0 level:error browser:Firefox, save searches, and resolve, ignore, merge or assign in bulk.
  • Issue detail: a de-minified stack trace with context lines and a switch for library frames, a breadcrumb trail, tag distribution over the last 200 events, the event payload, activity, user feedback and older/newer event links.
  • Source maps: upload maps from CI with an organisation token. They are parsed in the worker with time and memory limits, kept private, and used to show original files, lines and functions.
  • Releases and deploys: record a deploy, mark an issue resolved in the next release, and see regressions when it comes back. Compare two releases and see the sessions that ended without a crash.
  • Alert rules: new issue, regression, spike and frequency rules with filters, cooldowns, muting and a preview of how often the rule would have fired in the last seven days. Channels: email, signed webhook and a Slack-compatible webhook.
  • Report dialog: a browser SDK call that lets a user describe what happened; the comment is stored with the event.
  • SDKs in the download: @faultsnare/browser, @faultsnare/node and a PHP package with a Laravel service provider, each with its own tests.

For the operator

  • Plans and quotas: monthly event quotas per plan with spike protection, overage packs, per-plan retention (30, 60 or 90 days) and a usage page that shows accepted and rejected events by reason.
  • Operator console: platform overview, ingest throughput, worker lag, driver health, workspaces, users and plans.
  • Workspaces and roles: owner, admin, developer, member and viewer, with teams, invitations, organisation tokens, a REST API with an OpenAPI document, webhooks, an audit log, two-factor sign-in and magic links.
  • Privacy by default: emails, IP addresses, tokens, JWTs, card numbers and the values of keys such as password and secret are replaced with [Filtered] on the server before anything is stored. Request bodies, cookies and headers never survive.
  • Setup wizard, demo data, installer lock and documentation (a guide inside the app and a page in the download). Payments are simulated until you connect your own gateway (Stripe and Razorpay adapters included).

The demo data

Six workspaces, 9 projects, 19 DSN keys, 78 releases with 156 deploys, 10 source-map files, 297 issues and 15,816 events, 12 alert rules with 207 incidents, and 36 user feedback items. Dates are relative to the day you load it. Every person, company, host and address in the data is fictional.

Good to know

  • Self-hosted on your server with MongoDB 6+ (a replica set is recommended; a single node works) and two long-lived Node processes, the web app and the worker. How many events it can take depends on your hardware; plan quotas protect it and retention is configurable.
  • Includes browser, Node and PHP/Laravel SDKs. There is no performance tracing and no session replay in this version.
  • Personal data is scrubbed on ingest by default; you can add your own scrub keys per project.
  • The Sentry-compatible ingest endpoint is off by default and is switched on per project. It accepts the common envelope format for errors and sessions; it is not an official integration of any other vendor.
  • Email needs a real driver (SMTP) in production. The default log driver stores messages in the Outbox and refuses to run in production, so set MAIL_DRIVER before you go live. Payment, notification and other third-party connections run in demo mode until you configure your own provider; demo drivers refuse to run in production.
  • Gemini AI admin tools need your own Google Gemini API key (Settings → AI); no key or AI usage is included. Without a key the AI buttons stay disabled. You add your own Gemini key (GEMINI_API_KEY) after purchase; the product ships with it blank and never uses ours.
  • A Google Translate language switcher is on the public site and admin; it needs Google's widget to load in the visitor's browser, and private areas (stack traces, event payloads, breadcrumbs, tags, user feedback, DSN keys and tokens, issue titles and release names, the search bar) are excluded from translation so that your error data is never sent to Google.
  • Starting codebase for an error tracking service, not a finished business. Extended licence needed to sell the service to paying customers.
  • Back up your APP_SECRET and your storage folder (source maps live there).

Gemini AI admin tools

Add your own GEMINI_API_KEY after purchase to switch on suggestion-only AI actions that you preview and edit before accepting: explain the dashboard, what to fix first, what happened in an issue, explain a stack trace, fix hints, suggest a severity, plain-English issue search that is checked against an allow-list of fields, release notes, alert rule suggestions, fingerprint rule suggestions, feedback replies, usage explanations and operator actions. Inputs are built on the server from scrubbed data only: never request bodies, cookies, headers, keys or source-map contents, and identifiers such as hosts, DSNs and ids are redacted before every call. Results are never saved or applied automatically, and accepted ones are audited as AI-assisted. A Google Translate language switcher sits in the site header and the app top bar. Google bills your own key.

How it's built

Architecture, stack, key decisions and the quality gates DEV3 Faultsnare passed before release.

The engineering behind Faultsnare: an ingest endpoint with size caps, rate limits and quotas, server-side scrubbing, fingerprint grouping, a worker that de-minifies stack traces, and AI tools that only suggest, with limits stated.

The problem it solves

An error tracker is a service that other people's broken software calls, often all at once. Requests arrive from browsers anyone can inspect, bodies can be compressed bombs, a bad release can multiply traffic by a hundred, and every payload may contain emails, tokens or card numbers that must never be stored. Stack traces from minified code are useless until source maps are applied, and source maps are themselves untrusted files. On top of that, one spike should not hide the next problem, and an AI helper must not leak customer data.

Architecture and approach

Faultsnare is a Next.js 16 application on MongoDB, on a shared DEV3 core for auth, billing and tenancy, with a Node worker. Every tenant model is scoped by organisation through a fail-closed plugin, and cross-tenant reads live in one reviewed module.

Ingest. A route handler authenticates by DSN key, reads the body through a streaming decompressor with a byte counter, validates it against a schema, then applies a per-key rate limit, a per-project spike guard and the organisation's monthly quota before anything is stored. Refused requests are counted by reason and shown on the usage page.

Scrubbing. Emails, addresses, tokens, card numbers and the values of secret-looking keys are replaced on the server; request bodies, cookies and headers never survive.

Grouping and state. A fingerprint is a hash of the exception type and normalised in-app frames, or the message template, with per-project rules. Issues move between unresolved, resolved, resolved in the next release, ignored and regressed through conditional single-document updates, so two events arriving together cannot double-create an issue.

Source maps. A hand-written Source Map v3 and Base64 VLQ decoder runs only in the worker with time and memory limits, per release and file, and context lines come from the embedded sources.

Alerts. Rules for new issues, regressions, spikes and frequency use windows, cooldowns and de-duplication; channels go through a client that pins DNS and refuses private addresses.

AI. The Gemini AI admin tools build their input on the server from scrubbed frames only, redact hosts, keys and ids, show an editable preview and save nothing until accepted. Under tests, seeds and the demo a deterministic fake driver is used; the real driver is never constructed.

Quality gates and results

Faultsnare 1.0.0 passed type-checking and linting with no errors, 560 app tests and 59 core tests (including Mongo race tests for ingest, grouping and quotas), a 72-route production build with 200.1 KB of gzipped first-load JavaScript and CSS, seeding of the empty and demo datasets, and a headless-browser pass over 42 routes and 14 happy paths. Lighthouse on the home page, mobile: performance 93, accessibility 100.

Tech stack

Area Technologies
Application Next.js 16 App Router, Server actions, React 19, Tailwind CSS 4, Route handlers for ingest and REST
Engine Node worker process, Mongo inbox with leases, Source Map v3 decoder, Fingerprint grouping, Alert windows and channels
Data MongoDB 6+, Mongoose 9, Quota counters with conditional updates
SDKs @faultsnare/browser, @faultsnare/node, PHP package with a Laravel provider
Quality Vitest, Mongo race tests, Headless browser smoke tests

Inside the build

Home: Errors grouped, de-minified and alerted on your own server, with a flat snare line catching red dots. Home — Errors grouped, de-minified and alerted on your own server, with a flat snare line catching red dots.

What you get: SDKs, grouping, source maps, releases, alerts and privacy defaults, each in one card. What you get — SDKs, grouping, source maps, releases, alerts and privacy defaults, each in one card.

Guides & articles

Practical guides and engineering write-ups for DEV3 Faultsnare.

Self-hosted error tracking: 7 things to check before you pick one

Last updated
Self-hosted error tracking: 7 things to check before you pick one: DEV3 Faultsnare

An error tracker is only useful if the noise is low, the stack traces are readable and the right person hears about a problem in time. Here are seven things to check before you pick error tracking software you host yourself, with DEV3 Faultsnare as the worked example.

An issue with a de-minified stack trace, context lines and a management panel

1. Does it group errors sensibly?

Look for grouping by exception type and the shape of the stack, not by the raw message. Faultsnare normalises frames and message numbers, lets you add fingerprint rules per project, and lets you merge issues by hand.

The issue list with search and sparklines

2. Can you read the stack trace?

Minified code is useless in a report. Check that source maps can be uploaded from CI, that they stay private, and that frames show original files, lines and a few lines of context. In Faultsnare the maps are parsed in a separate worker with time and memory limits.

3. Does it know about releases?

You want to mark an issue as fixed in the next release and see it reopen as a regression if it returns. Faultsnare records deploys, compares two releases and shows the sessions that ended without a crash.

Releases with deploys and session health

4. Are the alerts tunable?

Check for new-issue, regression, spike and frequency rules, cooldowns, muting and a way to preview a rule before it pages anyone. Faultsnare shows how often a rule would have fired in the last seven days and sends through email, signed webhooks or a Slack-compatible webhook.

The alert rule editor

5. What happens to personal data?

Scrubbing should happen on the server before storage, not as an option in the client. Faultsnare removes emails, addresses, tokens, card numbers and the values of keys such as password and secret by default, and never keeps request bodies, cookies or headers.

6. What stops it from falling over?

Ask about size caps, rate limits per key, monthly quotas and spike protection. Faultsnare ships plan quotas, overage packs and per-plan retention, and its usage page shows what was accepted and what was rejected, by reason.

Usage with quota and rejected events

7. How is the AI wired?

The AI tools should suggest, never act. In Faultsnare, optional Gemini AI admin tools use your own Google Gemini key, build their input on the server from scrubbed frames only, show an editable preview and save nothing until you accept. Without a key the buttons stay disabled.

An AI suggestion shown as a preview

Video tour of Faultsnare

Limits to weigh

There is no performance tracing and no session replay in this version. Volume depends on your hardware. The Sentry-compatible ingest endpoint is off by default. Email needs a real driver in production. Check these against your needs.

Frequently asked questions

The complete source code (version 1.0.0) as a download from your account, with the setup and usage guides shown on this page. You can modify the code to suit your product.

Questions about this product (0)

Answers from the DEV3 team and the community.

Ask a question

No questions yet.

Wondering about setup, licensing or customisation? Ask, and the DEV3 team will help.

More from this author

View all

Related in Enterprise SaaS Starters

View all