Skip to content

performance.md

Latest commit

 

History

History
35 lines (29 loc) · 1.12 KB

performance.md

File metadata and controls

35 lines (29 loc) · 1.12 KB
Raw

Performance

Good defaults for most apps:

MODEL_MIND_CONTEXT_CACHE_SECONDS=600
MODEL_MIND_RETRIEVAL_ENABLED=true
MODEL_MIND_RETRIEVAL_LIMIT=8
MODEL_MIND_RECENT_MESSAGES=8
MODEL_MIND_BROWSER_MESSAGES=60
MODEL_MIND_MESSAGE_CHARACTERS=800
MODEL_MIND_SUMMARY_CHARACTERS=2000
MODEL_MIND_MAX_OUTPUT_TOKENS=450
MODEL_MIND_TIMEOUT=12
MODEL_MIND_CONNECT_TIMEOUT=3
MODEL_MIND_STREAMING=true
MODEL_MIND_BACKGROUND_MODE=after_response

Large Applications

  • Keep per-model limit small.
  • Keep question-aware retrieval enabled for large searchable models.
  • Enable streaming so visitors see partial answer text while the provider is still generating.
  • Use MODEL_MIND_BACKGROUND_MODE=queue with a worker for high-traffic apps.
  • Prefer scoped public records with modelMindContextQuery().
  • Use custom context providers for summarized analytics instead of sending many raw rows.
  • Run php artisan model-mind:inspect-context and check the resulting character size.
  • Keep MODEL_MIND_MAX_OUTPUT_TOKENS low for faster chat responses.

Clear cached context after imports or config changes:

php artisan model-mind:clear-context