ai_sdk_provider 1.2.0

Provider interface specification for the AI SDK Dart port. Defines the LanguageModelV3, EmbeddingModelV2, and ImageModelV3 interfaces that all provider packages must implement.

1.2.0

Error hierarchy

• The full AiSdkError hierarchy now lives here (relocated from ai_sdk_dart) so that every provider package — which depends on ai_sdk_provider, not ai_sdk_dart — can throw the same typed errors. ai_sdk_dart re-exports the hierarchy, so its public surface is unchanged.
• AiApiCallError is enriched with statusCode, url, responseBody, responseHeaders, type, code, isRetryable, and cause, plus an AiApiCallError.fromResponse(...) factory that tolerantly parses each provider's error-body shape (OpenAI/Groq/Mistral, Anthropic, Google, Cohere, Ollama). The original positional-message constructor stays const and backward-compatible.

Removed

• VideoModelV1 and its VideoModelV1CallOptions / VideoModelV1GenerateResult types. No provider ever implemented video generation, so the interface was dead surface area. The matching generateVideo entry point was removed from ai_sdk_dart in the same release.

---

1.1.0

• LanguageModelV3CallOptions.outputSchema — new optional Map<String, dynamic>? field. When non-null, providers that support native structured output (e.g. OpenAI response_format: json_schema) should use it to request schema-validated responses directly from the model API.

---

1.0.0+1

• Improved pubspec descriptions for better pub.dev discoverability.
• Added example/example.md with usage examples and links to runnable apps.

1.0.0

First stable release. Defines the provider interface contract for all AI SDK Dart providers.

• LanguageModelV3 — doGenerate / doStream interface.
• EmbeddingModelV2<VALUE> — doEmbed interface.
• ImageModelV3 — doGenerate interface.
• SpeechModelV1 — doGenerate interface.
• TranscriptionModelV1 — doGenerate interface.
• RerankModelV1 — doRerank interface.
• Full LanguageModelV3StreamPart sealed class hierarchy (text, tool call, reasoning, source, file, finish, error parts).
• LanguageModelV3CallOptions with all v6 call-time options.
• LanguageModelV3GenerateResult and LanguageModelV3StreamResult.
• Shared types: JsonValue, ProviderMetadata, content parts, finish reasons, tool definitions, usage.

---

0.2.0

• Initial release.
• LanguageModelV3 interface with doGenerate / doStream.
• EmbeddingModelV2<VALUE> interface with doEmbed.
• ImageModelV3 interface with doGenerate.
• SpeechModelV1 interface with doGenerate.
• TranscriptionModelV1 interface with doGenerate.
• RerankModelV1 interface with doRerank.
• Full LanguageModelV3StreamPart sealed class hierarchy.
• LanguageModelV3CallOptions with all v6 call-time options.
• LanguageModelV3GenerateResult and LanguageModelV3StreamResult.
• Shared types: JsonValue, ProviderMetadata, content parts, finish reasons, tool definitions.