specification: API Commons Rate Limits specificationVersion: '0.1' schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits provider: UpTrain providerId: uptrain created: '2026-06-20' modified: '2026-06-20' reconciled: false tags: - AI - LLM - Evaluation - LLM Evaluation - Open Source - Rate Limiting - Quotas - Throttling description: >- UpTrain does not publish documented rate limits for its managed evaluation API. The open-source framework is self-hosted and effectively bounded only by the rate limits of whichever LLM provider you supply a key for (OpenAI, Anthropic, Azure OpenAI, Mistral, etc.) when grading. For the managed service (demo.uptrain.ai), any per-account throttling is set by UpTrain and is not reconciled in this artifact. notes: >- No public rate-limit documentation exists for the managed UpTrain API. When running the open-source framework, the practical ceiling is the throughput and rate limits of the configured upstream LLM grading provider. Confirm managed-tier limits with UpTrain on reconciliation. sources: - https://docs.uptrain.ai/ - https://github.com/uptrain-ai/uptrain responseCodes: throttled: 429 limits: - name: Managed Evaluation Requests scope: account metric: requests limit: see provider documentation notes: No documented per-account limit for the managed evaluate / log_and_evaluate endpoints. - name: Upstream LLM Grading Limits scope: account metric: requests limit: depends on configured LLM provider notes: >- In the open-source framework, grading calls go to your own LLM provider key and inherit that provider's rate limits. policies: - name: Provider-Bound Limits description: >- Open-source evaluation throughput is bound by the upstream LLM provider's rate limits, not by UpTrain itself. - name: Backoff Strategy description: Clients should implement exponential backoff with jitter and honor Retry-After on 429 responses. maintainers: - FN: Kin Lane email: kin@apievangelist.com