Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 14 additions & 7 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,19 @@
# Changelog

## Pending
## [1.0.0] - 2026-09-21

1.0 marks the command surface as complete — apps, metrics, endpoints, background
jobs, traces, errors, anomalies, insights, usage and billing — and this release
is about making that surface dependable rather than larger: consistent `--json`
contracts, input validated before it reaches the server, and charts that
represent the data behind them honestly. The
[README](https://github.com/scoutapp/scout-cli#usage) covers what each command does.

### Breaking changes

- `--toon` and the `toon-format/toon-go` dependency are removed. `--json` is now the CLI's only structured output format, and **piped output auto-enables `--json` instead of TOON** — anything piping `scout` and parsing TOON needs to parse JSON instead (#28)
- `scout usage --billing-period --json` now wraps the per-app result in an object carrying `billing_period` and `server_total`, so a script can tell which window it got and what the billed figure was. Without `--billing-period` the shape is unchanged (#29)
- `scout jobs metrics --type latency --json` renames the API's `total` sub-series to `execution_time_total`, which is what it actually holds; other metric types are unchanged (#29)

### Added

Expand All @@ -12,8 +25,6 @@
### Changed

- `scout usage --app <id>` (without `--by-day`) now filters to that one app instead of silently reporting every app, matching `--by-day --app <id>` (#29)
- `scout usage --billing-period --json` now wraps the per-app result in an object carrying `billing_period` and `server_total`, so a script can tell which window it got and what the billed figure was. Without `--billing-period` the shape is unchanged (#29)
- `scout jobs metrics --type latency --json` renames the API's `total` sub-series to `execution_time_total`, which is what it actually holds; other metric types are unchanged (#29)
- `scout errors show` omits the `Count:` line when the API reports 0, which it does for every error group even when `errors list` shows a real count (#29)
- `scout insights show` prints nested fields as indented `key: value` pairs and empty values as `—`, instead of leaking Go's `map[...]` and `<nil>` formatting (#29)
- Table cells are capped at 60 display columns with an ellipsis, so one long URI no longer stretches a table past a normal terminal width; `--json` output is uncapped (#29)
Expand Down Expand Up @@ -46,10 +57,6 @@
- String truncation in tables, the span tree and API error bodies counts graphemes rather than bytes, so it cannot split a multibyte character (#29)
- `scout setup <name>` matches framework names case-insensitively and its error points at `scout setup` with no arguments for the valid list (#29)

### Removed

- `--toon` flag and the `toon-format/toon-go` dependency — `--json` is now the CLI's only structured output format, and piped output auto-enables `--json` instead of TOON (#28)

## [0.5.0] - 2026-09-10

This release rounds out command coverage across the whole Scout API surface
Expand Down
Loading