You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Introduced the Responses.Provider registry with Responses.Provider.OpenAI and Responses.Provider.XAI implementations for credential lookup, model routing, and provider-specific warnings.
Added first-party support for xAI models including pricing data, telemetry metadata, and request/streaming tests that validate multi-provider flows.
Exposed a provider_warnings option (and :provider_warning_mode config) to control whether unsupported option warnings are emitted per provider.
Documented multi-provider configuration patterns and added Livebook/tutorial coverage for running responses against xAI.
Changed
Responses.list_models/1 now requires a provider atom and list_models/2 applies provider-scoped filtering while returning the raw upstream payload.
Cost calculation automatically scopes to the resolved provider and surfaces descriptive errors when pricing metadata is missing.
Telemetry events include provider identifiers so downstream instrumentation can differentiate usage per provider.
Migration Guide
Update your dependency entry to {:responses, "~> 0.1.0"}.
Configure provider-specific credentials with config :responses, :openai_api_key and/or :xai_api_key (or the matching environment variables).
When requesting models, supply explicit provider hints (model: "provider:model") or rely on the documented prefixes (gpt-*, o1*, o3*, o4-mini* map to OpenAI; grok-* maps to xAI).
Review telemetry consumers for the new provider metadata and adjust alerting or dashboards as needed.
[0.0.1] - 2024-09-21
Changed
Renamed the library namespace from OpenAI.Responses to Responses.
Updated the OTP application name to :responses and refreshed package metadata.
Adjusted configuration keys and documentation snippets to use the new namespace.
Migration Guide
Update your dependency entry to {:responses, "~> 0.0.1"}.
Replace module references from OpenAI.Responses to Responses.
Update application configuration to use config :responses, :openai_api_key, "your-key" (legacy keys continue to work with warnings).
Review CHANGELOG_OLD.md for historical notes from earlier releases.