[helpcodeai]

Summary

Sentry has accepted no spans for days: the organisation's span quota is used up every day. On 11-04 the Cloud backend sent about 714k spans (662k dropped by the SDK's rate-limit backoff, 52k refused with span_usage_exceeded) at the Cloud default SENTRY_TRACES_SAMPLE_RATE=0.05. So we paid the tracing CPU cost (Sentry Prisma tracing was about 12 % of backend CPU in the 11-03 profile) and saw no traces.

Changes

• The default of SENTRY_TRACES_SAMPLE_RATE in docker-compose.cloud.yml goes from 0.05 to 0.005. MCP calls keep a fifth of it (0.001). Errors are unchanged.
• The droplet .env does not set this variable, so the next Cloud deploy picks up the new value. Setting it in .env still overrides it.
• The website goes from 10 % to 2 % in HelpCode-ai/anythingmcp-website#516.

Expected: about 70k backend spans a day plus about 40k from the website, back under the quota.

Type

• Other: Cloud configuration

Testing

• No tests needed: a compose default.
• The droplet .env has no SENTRY_TRACES_SAMPLE_RATE (checked).
• After the deploy: Sentry spans accepted > 0 and span_usage_exceeded near 0 (find_dropped_events).

Checklist

• My code follows the existing code style
• I have tested my changes locally
• I have updated documentation (if applicable): comment in the compose file
• All existing tests pass
• This PR has a descriptive title

Related Issues

Daily check 2026-11-05 (follow-up 4: Sentry spans 100 % dropped).

Deployment: Cloud only. Self-hosted keeps its own default (0, tracing off).