refactor: clean up LLM generate methods and improve type safety

[wardpeet]

Summary

• Refactored Agent class to consolidate LLM generate and stream methods
• Improved type safety across LLM interfaces and agent methods
• Fixed type annotations in tests and removed console.log statements
• Moved LLM-specific types to proper modules for better organization

Changes Made

• Agent class refactoring: Consolidated duplicate code between generate() and stream() methods by extracting common logic into prepareLLMOptions()
• Type improvements: Enhanced type safety for LLM operations with proper generic constraints and cleaner type definitions
• Code organization: Moved LLM callback types to base.types.ts for better modularity
• Test fixes: Added proper type annotations to test variables
• Cleanup: Removed debugging console.log statements

Test plan

• All existing tests continue to pass
• Type checking passes without errors
• Agent generate and stream methods work correctly
• LLM interfaces maintain backward compatibility

🤖 Generated with Claude Code

[codesandbox]

Review or Edit in CodeSandbox

Open the branch in Web Editor • VS Code • Insiders

Open Preview

[changeset-bot]

🦋 Changeset detected

Latest commit: 2af1f3a

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 11 packages

Name	Type
@mastra/server	Patch
@mastra/core	Patch
@mastra/deployer	Patch
@mastra/client-js	Patch
@mastra/dane	Patch
@mastra/longmemeval	Patch
mastra	Patch
@mastra/deployer-cloudflare	Patch
@mastra/deployer-netlify	Patch
@mastra/deployer-vercel	Patch
@mastra/playground-ui	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
assistant-ui	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 17, 2025 9:29am
mastra-docs	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 17, 2025 9:29am
openapi-spec-writer	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 17, 2025 9:29am

[greptile-apps]

PR Summary

Major refactoring of LLM-related code to improve type safety and reduce duplication, focusing on code organization and type definitions.

• Created new packages/core/src/llm/model/base.types.ts to centralize LLM-related type definitions and improve type organization
• Consolidated duplicate code in Agent class by extracting prepareLLMOptions() from generate() and stream() methods
• Enhanced type safety by adding proper generic constraints and improving callback type definitions in packages/core/src/llm/model/base.ts
• Warning: Found potentially unintended console.log statement in packages/core/src/agent/index.ts that should be removed
• Fixed type annotations in test files to properly handle MockLanguageModelV1 instances

_{13 files reviewed, 6 comments
Edit PR Review Bot Settings | Greptile}

[wardpeet]

@greptile-apps

[greptile-apps]

PR Summary

Comprehensive implementation of agent working memory and UI improvements across the platform.

• Added working memory management UI with new agent-working-memory-context.tsx context provider and hook-based state management
• Implemented working memory API endpoints in server and deployer packages with proper validation and error handling
• Replaced legacy Cohere client with direct API calls in CohereRelevanceScorer, improving error handling and reducing dependencies
• Enhanced model settings UI with new provider-specific configuration options and JSON editor in agent-advanced-settings.tsx
• Added comprehensive icon set for different LLM providers including Anthropic, Azure, OpenAI, Groq and others with consistent styling

_{145 files reviewed, 6 comments
Edit PR Review Bot Settings | Greptile}

[github-actions]

Braintrust eval report

No experiments to report