Epure
Changelog · Apache 2.0

What we ship. What comes next.

Exception tracking on your box. Keep your Sentry SDK. Swap the DSN. The latest tag tells you when HTTP sign-in cannot keep the Secure cookie.

GitHub

Latest v0.1.7 · 2026-10-04

v0.1.7 · 2026-10-04

Gzip ingest, and HTTP sign-in explains the dropped cookie

v0.1.7 still accepts gzip envelopes. A correct password on http://your-server:8080 no longer looks like success. Production can start Caddy with --profile tls.

  • Login stays on the form

    When the cookie is Secure and the page is HTTP, sign-in is disabled and the page says the browser will drop it.

  • session_secure on /auth/config

    The login page reads that flag. Local Compose still uses an insecure cookie, so localhost is unchanged.

  • Caddy on the prod overlay

    --profile tls publishes ports 80 and 443. ./configure --prod requires an https DNS name, not a raw IP.

  • No migration

    git pull, set EPURE_IMAGE=ghcr.io/epure-sh/epure:v0.1.7 in .env, then recreate. Open the HTTPS URL.

GitHub release notes · Compare v0.1.7 · Docs changelog · Image ghcr.io/epure-sh/epure:v0.1.7

Since v0.1.0 · launch cut

Core stack since launch.

Exception ingest into grouped issues, Postgres + RLS, keyboard triage, webhooks, and RBAC. Patch releases add setup and agent tooling on top.

  • Ingest + spike valve

    Envelope and store paths. 202 responses. Caps stop a runaway fingerprint from melting the box.

  • Sourcemaps + grouping

    JS/TS demangle. SHA-256 fingerprints. Breadcrumbs. PII scrub at ingest.

  • Postgres 16 + RLS

    Org isolation in the database. Monthly partitions. TTL drop for old events.

  • Keyboard triage

    j/k/e/i/x. Query syntax. Merge/split. Bulk resolve. Markdown export for LLMs.

  • Lifecycle signals

    Regression, snooze, velocity alerts. Slack / Discord / generic webhooks.

  • Admin + RBAC

    Multi-project orgs. DSN rotate/revoke. Owner / Admin / Member.

Full changelog in docs

Roadmap

Self-host live. Public next. Cloud after.

Exception tracking only. You can see what shipped, what is next, and what we refuse to build.

  1. GRND

    Self-host

    Live

    Open source error tracking on your box. Issues, stack traces, grouping.

    • Spike valve + sourcemaps + RLS
    • Keyboard triage + webhooks
    • Apache 2.0 · clone and run
  2. OPEN

    Public launch

    Next

    Public GitHub, migration guides, and the dashboard you triage in.

    • Public GitHub publish
    • Migration + compatibility guides
    • Dashboard polish before launch
  3. CLD

    Cloud

    Queued

    Same binary, our ops. v2 flat tiers. Tiered soft-sample on spike. No auto-overage invoice.

    • Managed Postgres + TLS
    • Stripe checkout → first event
    • Count 100% · store 10% on overage
  4. HZN

    Scale

    Later

    Error tracking on the same binary. Heavy tenants still skip Redis on OSS.

    • Scale SSO / audit logs
    • Cloud split modes for hot orgs
    • ClickHouse for raw payloads (Cloud only)

Out of scope

Error tracking stays the product. We will not add tracing, replay, or profiling later.

  • Tracing / APM
  • Session replay
  • Profiling
  • Generic logs
  • iOS / Android
  • Redis on OSS
Contribute

Three ways to help the next release.

Stack gate stays fixed: Postgres 16 with RLS. No Redis or Kafka in the OSS compose.

  • Report a bug

    Use Bug or SDK compatibility templates. Search existing issues first.

  • Improve docs

    Install steps, compatibility notes, and migration guides help the next clone.

  • Send a focused PR

    Fork, branch off main, run ./scripts/test.sh. Open an issue first for large changes.

Read CONTRIBUTING

Out of scope for PRs: Redis, ClickHouse, Kafka, tracing, replay, profiling, logs, iOS/Android, or “100% Sentry parity.”

License

OSS is the product. Cloud is hosting.

Core lives in the self-hosted binary under Apache 2.0. Cloud is managed hosting of that same path.

  • Apache 2.0

    Use, modify, redistribute. Attribution and patent terms follow the license in the repo.

  • Self-host stays free

    Full triage in OSS. You pay the VPS. No seat lock on the core path.

  • Cloud is flat

    Pro $29 / Scale $89. Not an EE fork that strips triage.

License
Apache 2.0
Self-host
Free forever. Install path
Cloud
Flat Pro $29 / Scale $89. Pricing
FAQ

Before you star the repo.

What shipped?
Since v0.1.0: exception ingest, spike valve, JS/TS sourcemaps, grouping, PostgreSQL 16 with RLS, keyboard triage, regressions, webhooks, and RBAC. v0.1.2 adds the Agent API, epure-cli, and scoped PATs. v0.1.4 adds EPURE_REGISTRATION=false so a public self-host URL can stop new workspaces. v0.1.5 accepts public-key-only DSN auth. v0.1.6 ships epure-cli in the GHCR image and raises PAT/MCP rate limits. v0.1.7 adds gzip envelope ingest and clearer HTTP login when Secure cookies cannot be stored. Apache 2.0. Details: docs changelog and GitHub releases.
Will you cripple the OSS build later?
The product is the self-hosted binary under Apache 2.0. Cloud is managed hosting of that same path, with triage staying in OSS.
How often do you ship?
Watch releases for dated notes.
What license is Epure under?
Apache 2.0. You can use, modify, and redistribute the binary. Attribution and patent terms follow the license in the repo.
Is this 100% Sentry compatible?
No. We keep official Sentry SDKs on the wire and accept envelope/store for exceptions. Transactions are discarded. No replay, profiling, sessions, or mobile symbolication. Honest split: Epure vs Sentry.
How do I propose a large feature?
Open a GitHub issue first. PRs that assume Redis, Kafka, or ClickHouse get closed. Focused fixes and docs land faster. Start from the contributing guide.
Where do I report a security issue?
Follow SECURITY.md. Prefer a private GitHub advisory, or email security@news.epure.sh. Do not file public issues for unpatched vulnerabilities.

Read the code. Run the binary.

Start Cloud from the waitlist or star the repo for self-host.

GitHub

Apache 2.0 · open source on GitHub