- Add webSearch method to FirecrawlService with WebSearchOptions, WebSearchResult, and WebSearchResponse interfaces
- Export new types from web-tools package
- Create web-search-tool in AI package with proper input/output schemas
- Add comprehensive unit tests for webSearch method (16 tests total)
- Add tool tests with mocking (4 tests)
- Add integration tests (skipped when no API key)
- Update package exports and dependencies
- Default search limit: 5, default format: markdown, content scraping enabled
- Proper error handling using CompanyResearchError
- AI package has no direct Firecrawl dependency
Co-Authored-By: Dallin Bentley <dallinbentley98@gmail.com>
- Added new steps for the documentation agent, including context initialization, todo creation, and main processing logic.
- Introduced schemas for input and output validation using Zod.
- Created a new workflow that orchestrates the steps for generating documentation based on user input.
- Enhanced the existing file structure to support the new documentation agent functionality.
- Updated the export name in index.ts and adjusted imports in the corresponding test file.
- Modified the tool definition in bash-execute-tool.ts to reflect the new naming convention.
The Snowflake adapter implementation transforms column names to lowercase
for consistency, but the tests were expecting uppercase column names.
This commit updates the tests to match the implementation:
- Update test expectations to use lowercase column names (id, name)
- Fix hasMoreRows assertions to match implementation logic (only true when rowCount > limit)
- Ensure all Snowflake-related tests pass with the current adapter behavior
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Increased allowed variance in cached query time checks to accommodate network fluctuations.
- Corrected property name in test assertions to match expected lowercase format.
- Enhanced SnowflakeAdapter to transform column names to lowercase and adjusted logic for determining if more rows are available from the stream.
- Implemented a custom fetch function in both anthropic.ts and vertex.ts to modify the request body.
- If tool_choice is present, added disable_parallel_tool_use to the request body.
- Included error handling for JSON parsing to ensure fallback to original request on failure.
- Added tests for handling AbortError scenarios in both memory-safe and standard streaming models.
- Updated the FallbackModel to gracefully handle intentional aborts without retrying or causing errors.
- Ensured that the controller closes properly on AbortError and that no fallback to subsequent models occurs.
- Enhanced mock models to simulate AbortError conditions for comprehensive testing.
- Replace SELECT * with explicit column names in permission validation tests
- Update CTE test to use explicit columns in final SELECT
- Maintain test intent while aligning with new wildcard security validation
- Ensure all tests pass with new wildcard blocking behavior
Co-Authored-By: Dallin Bentley <dallinbentley98@gmail.com>
- Updated references across multiple files to reflect the new tool name.
- Adjusted unit tests and schemas to ensure compatibility with the renamed tool.
- Removed the old respondWithoutAnalysis tool implementation and its associated tests.
- Ensured backward compatibility for existing functionality using the old tool name.
- Import Select from node-sql-parser for type safety
- Replace function parameter types with Record<string, unknown> for dynamic AST objects
- Use proper type conversions through 'unknown' for incompatible types
- Maintain existing wildcard validation functionality
- Resolve all 8 noExplicitAny linting violations
Co-Authored-By: Dallin Bentley <dallinbentley98@gmail.com>
- Update query() method to use streamResult: true and stmt.streamRows()
- Add network-level row limiting with default 5000 row cap
- Process stream events (data, error, end) to build result set
- Maintain backward compatibility with existing adapter interface
- Update unit tests to mock streaming behavior
- Fix integration test imports and property names
- Preserve query caching by using original SQL unchanged
Co-Authored-By: Dallin Bentley <dallinbentley98@gmail.com>
- Add validateWildcardUsage function to sql-parser-helpers.ts
- Block SELECT * and qualified wildcards on physical database tables
- Allow wildcards on CTEs and derived tables
- Integrate wildcard validation into permission-validator.ts
- Add comprehensive tests for all wildcard validation scenarios
- Revert Rust SQL analyzer changes to focus on TypeScript implementation
- Fix CTE alias handling for qualified wildcards (e.g., SELECT cte_alias.*)
This prevents bypassing column-level permissions through wildcard queries
while maintaining backward compatibility with legitimate query patterns.
Co-Authored-By: Dallin Bentley <dallinbentley98@gmail.com>
- Add curly braces around default case content to satisfy noSwitchDeclarations rule
- Prevents variable declarations from being accessible to other switch cases
- Apply Biome formatting fixes for import statement and spacing
Co-Authored-By: nate@buster.so <nate@buster.so>
- Add validateWildcardUsage() function to block SELECT * on physical tables
- Allow wildcards on CTEs but block on physical database tables
- Add comprehensive tests for wildcard validation scenarios
- Integrate wildcard validation into permission validator
- Supports all SQL dialects via node-sql-parser
- Prevents permission bypass through wildcard queries
Co-Authored-By: Dallin Bentley <dallinbentley98@gmail.com>
- Import database queries from main @buster/database package
- Use getUserOrganizationId() instead of user.organizationId
- Add proper error handling for users without organizations
- Follow existing patterns from other API handlers
Co-Authored-By: nate@buster.so <nate@buster.so>
- Added support for Google Vertex AI model version 2.2.27 in pnpm-lock.yaml and package.json.
- Introduced new model wrappers for Sonnet4 and Haiku35 with fallback logic.
- Updated various agents to utilize the new Sonnet4 model instead of the previous anthropicCachedModel.
- Enhanced error handling and retry logic in the fallback model implementation.
- Increased processing row limit in Snowflake query to 5000 for better performance.
- Added tests for new AI fallback models and their streaming capabilities.
Nate Kelley
Devin AI
dal