← Back to Documentation
Operations
Performance Tips
Improve perceived latency, provider reliability, and private model throughput.
Key Steps
- Place servers close to users and provider endpoints where possible.
- Track latency by provider and model before changing routing.
- Capacity-test private model hosts with representative prompts.
Article Scope
This guide focuses on the operational steps needed to run Orchestris with team access, provider control, and predictable usage.
Visual Reference
Provider latency
- Use usage reports to identify slow providers or models.
- Prefer lower-latency models for interactive workflows.
- Use fallback routing for providers that are occasionally unavailable.
Server tuning
- Right-size database, connection pool, and server resources for expected concurrency.
- Keep logs useful but avoid excessive debug logging in production.
- Monitor memory, CPU, queue depth, and request duration.
Private model hosts
- Measure throughput with real prompt sizes, not tiny smoke tests.
- Watch GPU or CPU saturation under concurrent use.
- Document limits so admins know when to add capacity or restrict access.
Related Topics
Operations
Downloads
Find the Orchestris Chat and Community Edition options currently available to customers.
Operations
Usage Analytics
Track requests, tokens, latency, and estimated cost across users, providers, and models.
Operations
Conversation Management
Help users organize work while keeping shared policies and retention expectations clear.