Bubbletea TUI for verifying tripbot's chat command surface end-to-end. Sends commands through Twitch IRC as a configured test user, captures the bot's replies, and prompts the human for Y/N on commands with stream-visible (overlay) effects.
Designed for one command per keypress — no batch mode, so it doesn't elbow into Twitch chat rate limits during a test session.
You need a Twitch developer application registered as a public client (no client secret) so the device-code OAuth flow works. Visit dev.twitch.tv/console/apps, register one with category "Chat Bot" and any redirect URL (the device flow doesn't use it). Copy the client ID.
go build -o command-tester .
TWITCH_CLIENT_ID=<your client id> ./command-testerFirst run prints a https://twitch.tv/activate URL and a user code — open the URL logged into the Twitch account you want to send chat as, paste the code, authorize. Token is cached at .token.json (gitignored, mode 0600); subsequent runs skip the auth flow until the cached token is invalidated.
OAuth scopes requested: chat:read chat:edit user:read:follows user:read:subscriptions. The follow / subscriber checks power the role badges in the TUI header so you can tell at a glance whether a follower-gated command is expected to pass.
| Flag | Default | Purpose |
|---|---|---|
-client-id |
$TWITCH_CLIENT_ID |
Twitch app client ID |
-channel |
adanalife_staging |
channel to send commands in |
-bot |
tripbot4000 |
bot username whose replies to capture |
-log |
logs/command-tester-<date>.log |
JSON-lines session log |
-timeout |
5 |
seconds to wait per command |
-token-cache |
.token.json |
OAuth token cache location |
List view
↑/↓move cursor ·g/Gfirst / lastspacetoggle whether the row runs ·a/ninclude all / noneenterorrrun the focused commandtabjump to summaryqquit
Review screen (shown after each run)
enteraccept the auto-resolved status and advancey/noverride as pass / fail ·sskipcenter comment mode — type freely,escto exit; notes are saved with the resultrre-run this commandescback to list without saving
JSON-lines at the configured log path. Each entry:
{
"ts": "2026-05-03T17:55:08Z",
"session_id": "ab12cd34ef567890",
"trigger": "!leaderboard",
"status": "manual-fail",
"bot_reply": "Right now only followers of the channel can run unlimited commands :)",
"notes": "failed because test account isn't following",
"latency_ms": 142,
"login": "mathgaming",
"channel": "adanalife_staging",
"bot": "tripbot4000",
"command_tester_version": "0ac5590-dirty",
"follower": false,
"subscriber": false
}Resume-aware: re-launching reads the existing log on startup and pre-marks rows that already have results, so you can pick up where you left off.
The list of commands and their metadata lives in catalog.go, populated by hand from tripbot's pkg/chatbot/handlers.go (~28 commands, ~70 aliases). Each entry carries category, expected-reply hint, onscreen-effect tag, and sample params.
Catalog drift is a known cost of this living in its own repo. When tripbot grows a new chat command, the catalog here won't auto-track it; you'll notice the gap on the next session. There's open work to either auto-generate the catalog or add a CI check comparing triggers against tripbot's handlers.
Extracted from adanalife/tripbot's streamtest/ subdir on 2026-05-03 via git subtree split, preserving the original four-commit history. The rename commit on top retitles the project from streamtest to command-tester.