feat(ai): implement experimental_prepareStep for streamText(v4) #7390
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ration - Add PrepareStepFunction and PrepareStepResult types for step-by-step customization - Enable dynamic model, toolChoice, and activeTools modification per step - Support system prompt and messages override during multi-step execution - Implement experimental_prepareStep parameter in generateText and streamText - Add comprehensive test coverage for experimental_prepareStep functionality - Fix stepNumber calculation bug in stream-text.ts (use currentStep instead of recordedSteps.length)
…e in stream-text.ts
… update related calls
…c maxSteps configuration
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Background
While AI SDK v5 includes native
prepareStepsupport for dynamic step-by-step configuration, it requiresLanguageModelV2interface which many AI providers haven't yet implemented. Production applications still rely on the stable v4 environment and cannot migrate to v5 due to:LanguageModelV1, including numerous third-party services that offer OpenAI-compatible APIs based on older interface standardsThis implementation backports the
prepareStepfunctionality from v5 to v4, enabling advanced multi-step AI workflows with existingLanguageModelV1providers.Summary
This PR implements
experimental_prepareStepforstreamTextin AI SDK v4, based on the v5prepareStepdesign but adapted to work withLanguageModelV1providers. The implementation includes:Core Changes
PrepareStepFunctiontype: Adapted from v5 implementation to work with LanguageModelV1PrepareStepResulttype: Defines step-specific configuration options (model, toolChoice, activeTools, system, messages)generateText: Addedmessagesas a new experimental_prepareStep parameterstreamText: Added experimental_prepareStep support for streamTextprepare-step.ts: Core types and interfaces following v5 patternsKey Features (based on v5 design)
experimental_activeToolsFiles Modified
packages/ai/core/generate-text/generate-text.ts: Added experimental_prepareStep integrationpackages/ai/core/generate-text/stream-text.ts: Added experimental_prepareStep support + stepNumber fixpackages/ai/core/generate-text/prepare-step.ts: New file with types (based on v5 implementation)packages/ai/core/generate-text/stream-text.test.ts: Comprehensive test coveragepackages/ai/core/generate-text/__snapshots__/stream-text.test.ts.snap: Updated snapshotsVerification
New test case added