Skip to content

Discord user-token smoke testing (jellybot)

Full dev-bot smoke suite: SMOKE.md (make smoke, make smoke-ci, GitHub Smoke workflow on self-hosted runner).

Optional individual live checks using a dedicated test user account and the local discord.py-self checkout.

Skill reference: discord-user-token-smoke-testing (~/coding/skills/discord-user-token-smoke-testing/SKILL.md).

Prerequisites

Requirement Where
discord.py-self checkout DISCORD_PY_SELF_ROOT=~/coding/discord.py-self
User token (throwaway test account) ~/coding/discord.py-self/.env → DISCORD_USER_TOKEN
Test channel DISCORD_TEST_CHANNEL_ID in discord.py-self/.env
Bot application id jellybot/.env → DISCORD_CLIENT_ID
Running prod (or dev) bot default log source: docker logs jellybot-dev (override JELLYBOT_SMOKE_LOG_CMD)

Commands

export DISCORD_PY_SELF_ROOT=~/coding/discord.py-self

# /quote match autocomplete (log-correlated pass gate)
bun run smoke:discord:quote

# /clip happy path (uses configured JELLYBOT_SMOKE_ITEM_ID)
bun run smoke:discord

# Both
bun run smoke:discord:all

# Log correlation unit tests (offline)
python3 scripts/test_discord_smoke_support.py

Makefile equivalents: make smoke-discord-quote, make smoke-discord, make smoke-discord-all.

Pass criteria: /quote autocomplete

The user client (discord.py-self) often reports client_choices=0 even when the bot responded correctly. Do not use client choices as the pass gate.

Authoritative check (implemented in scripts/discord_smoke_support.py):

  1. Fire APPLICATION_COMMAND_AUTOCOMPLETE for /quote option match from the test channel.
  2. Read structured bot logs (JELLYBOT_SMOKE_LOG_CMD, default docker logs jellybot).
  3. Pass when logs contain, in order:
  4. quote.autocomplete with matching interactionId + query
  5. quote.autocomplete.responded with responded: true and resultCount > 0
  6. Fail on quote.autocomplete.respond_skipped / respond_skip / autocomplete_failed, or missing outcome.

Override query: JELLYBOT_SMOKE_QUOTE_QUERY=looking.

docs/DISCORD_SMOKE_TESTING.md

When agents should run this

After changing quote autocomplete, Discord gateway, or deploy verification:

  1. bun run ci (required)
  2. bun run smoke:discord:quote when prod bot is up and credentials exist
  3. Treat failure + green CI as gateway/token/duplicate-bot ops issue until logs prove otherwise

Policy

  • User-token automation violates Discord ToS — dedicated test account only.
  • Never commit tokens.
  • Do not run bun run start from dev with the production bot token (duplicate gateway risk). See issue #60.