Problem

When hitting API rate limits (HTTP 429), Hermes currently displays a full traceback error to the user. This is a poor UX for a transient condition that the system should handle gracefully.

Current Behavior

ERROR:anthropic:Rate limit exceeded (429)
Traceback (most recent call last):
  File "/hermes/path/to/file.py", line 123, in function_name
    response = await client.chat(...)
anthropic.RateLimitError: Rate limit exceeded

Desired Behavior

1. User-friendly message: Display a clear, concise message like:

   • "Rate limit reached. Waiting 30s before retry..." (with countdown)
   • "Rate limit active. Retry in 15s."
   • Or similar clean messaging

2. Configurable automatic retry: Add config option(s) for rate limit handling:

   • retry.on_rate_limit: true (default)
   • retry.max_rate_limit_retries: 5 (default)
   • retry.rate_limit_backoff_base: 30 (seconds, default - exponential backoff)
   • retry.rate_limit_wait_max: 300 (max seconds to wait)

3. Show progress: Display countdown or progress indicator during wait:

   • CLI: "Rate limit hit. Waiting 30s... [████░░░░░░] 12/30"
   • Gateway platforms: "Rate limit hit. Retrying in 30s..." (update as countdown progresses)

4. Graceful failure: If retries exhausted or disabled, show clean error:

   • "Rate limit persisted after 5 retries. Please try again later."
   • No traceback for transient errors

Additional Notes

• This applies to all providers, not just Anthropic
• Current code already has some 429 retry logic (see PR fix(anthropic): retry 429/529 errors and surface error details to users #1585), but user-facing messaging still shows traces
• Should respect provider-specific Retry-After headers when available

Priority

Medium - UX improvement, system already attempts retries internally