Repository navigation
Releases
v0.1.0
The first tagged release of Geko: a failure-first dashboard for OpenTelemetry traces. It is one Go binary with SQLite storage, built for your machine and for a shared dev or staging server.
Install
Binary. Download the archive for your platform below, check it against checksums.txt, and run it:
sha256sum --check --ignore-missing checksums.txt
./geko serveDocker. The image is published for linux/amd64 and linux/arm64:
docker run --rm --network host -v geko-data:/data ghcr.io/yunusemrealps/geko:0.1.0Go 1.26.11 or later:
go install github.com/YunusEmreAlps/geko/cmd/geko@v0.1.0Open http://127.0.0.1:4318 and point your OpenTelemetry exporter at http://127.0.0.1:4318/v1/traces. For a shared server with TLS, see Run for a team.
What's in it
Find what broke
- The first screen lists requests whose spans recorded an exception, an HTTP 5xx or an error status. A request that returned 200 after a retried failure still shows that failure.
- Error groups with trace counts, first and last receipt times, and live updates. You can resolve, mute or reopen a group, and a group resolved in a version reopens only for traces from that version.
- Receives OTLP/HTTP (protobuf and JSON) and, optionally, OTLP/gRPC.
Investigate
- Trace pages with a service map, a span waterfall and the evidence spans.
- Compare a failing request with the latest successful one, and compare failure rate and duration between two
service.versionvalues. - VS Code links from recorded stack traces, and export and restore of OTLP JSON.
Work as a team
- Access tokens for ingest, read and triage, and roles:
viewer,triagerandadmin. You can also sign in through an authenticating proxy such as oauth2-proxy. - Notes and issue links on error groups. Copy as bug evidence produces Markdown to paste into Jira, Linear or GitHub. An optional Open in Tempo link opens the same trace in another tool.
- Webhooks for new and reopened groups (Slack, Discord, ntfy or JSON), and a JSON audit log of sign-ins and changes.
Coding agents
- MCP over stdio and at
/mcp: list failures and groups, compare traces and versions, and, if you allow it, change group state. geko run -- <command>lists the failures received while a command ran.geko failures --jsonprovides the same failure list for scripts.
Data stays careful
- An attribute allowlist drops headers, cookies, bodies and full URLs before storage.
- E-mail addresses, card numbers, IBANs and Turkish identity numbers are masked.
strictredaction, the default on a token-protected server, keeps only exception types and message hashes.
Upgrading from an untagged build
- The database migrates to schema 11 on first start. Back it up first with
geko backup; older builds can't open it afterwards. - A token-protected server without an explicit
redactionsetting switches tostrictfor new spans. SetGEKO_REDACTION=best_effortto keep messages and stack traces. - Read token and triage token names must differ.
Scope
Geko is meant for local development, side projects, and dev and staging environments. It doesn't collect logs or metrics, send threshold alerts or keep a user database. See the roadmap for what's next.
Full list of changes: CHANGELOG.md