Self-hosting
Platform LLM
Configure the LLM 143 uses for platform reasoning, classification, and analysis.
The platform LLM powers product reasoning inside 143: prioritization, summaries, failure explanation, and classification. Keep it separate from coding-agent credentials, which are used to run code-editing sessions.
Configuration model
Self-hosted deployments should define:
- Provider.
- Model.
- API key or credential reference.
- Rate limits appropriate for expected usage.
- Fallback behavior when the provider is unavailable.
Operational guidance
- Keep platform LLM credentials separate from coding-agent credentials.
- Monitor provider errors and latency.
- Start with conservative limits and raise them after observing real workload.
- Document which models are approved for production use.
For agents
When debugging behavior, distinguish platform LLM decisions from coding-agent execution. They use different credentials and serve different parts of the workflow.