Troubleshooting

Discord Message Too Long?

If your agent thinks for a while and then sends nothing (or replies cut off), you may be hitting Discord’s message length limit.

Recurring pattern: users ask for a long output, the bot appears to work, then either fails silently or returns partial text. This is usually not model failure — it’s output format mismatch with Discord limits.

Reality check: Discord has strict per-message character limits. Big responses must be split, summarized, or turned into a file.

Fast diagnosis (60 seconds)

  1. Ask for a tiny test: “Reply with exactly one sentence.”
  2. If tiny replies work but long ones fail, this is likely output-size related.
  3. Retry with: “Give me 5 bullets max, no long paragraphs.”

Known-good recovery flow

  1. Switch to chunked output: ask for Part 1/Part 2/Part 3.
  2. Limit each reply: “Max 8 bullets, max 120 characters per bullet.”
  3. For long drafts, ask for a file output instead of one giant chat message.
  4. Request a short executive summary first, then expand section-by-section.

Copy/paste anti-failure prompt

Output format rules (strict): - Do NOT send one giant message. - Split into numbered parts and wait after each part. - Keep each part under Discord-safe length. - If output is still too large, create a file and share the path. Now give me Part 1 only: a concise summary in 6 bullets max.

Avoid this trap: asking for “everything in one response” creates silent failures and trust breaks. Use staged output on purpose.

If you need the full long-form result

High-friction edge cases

Known-good "safe output" prompt

You are writing for Discord. Follow these constraints exactly: - No markdown tables. - No giant code blocks over 40 lines. - Max 900 characters per message. - Split into Part 1 / Part 2 / Part 3 and stop after each part. - If any part would exceed 900 chars, write full output to a file and return: file path + 5-bullet summary. Start with Part 1 now.

When to escalate to #help

Support packet

Discord long-message issue Where it fails (DM/server/thread): Tiny 1-sentence test result: Long output request used: Did chunked output work (yes/no): Did file-output fallback work (yes/no): Did DM vs server behavior differ (yes/no): Was markdown table/code-block output requested (yes/no): Exact error text (if any): Screenshot/message link:

Bottom line: if short responses work and long responses fail, switch to chunked or file-based output immediately. Don’t debug models first.