Implements a more robust bundling strategy for Mermaid diagrams to prevent
dynamic chunk loading at runtime, which was causing CSP violations.
Key changes:
- Added manualChunks function in vite.config.ts to consolidate all Mermaid
code and dependencies into a single bundle
- Added mermaid and dagre to optimizeDeps.include for pre-bundling
- Customized chunkFileNames to ensure predictable output
- Removed incorrect static imports in MermaidBlock.tsx
This resolves the CSP errors where Mermaid was attempting to dynamically
load chunks like chunk-RZ5BOZE2.js and channel.js from the webview origin.
Fixes: #3680
Signed-off-by: Eric Wheeler <roo-code@z.ewheeler.org>
Co-authored-by: Eric Wheeler <roo-code@z.ewheeler.org>
Co-authored-by: Daniel <57051444+daniel-lxs@users.noreply.github.com>
* feat(prompts): enforce codebase_search as primary code understanding tool
- Add conditional codebase_search enforcement in tool use guidelines
- Modify objective section to prioritize codebase_search when available
- Update rules section with critical codebase_search-first rule
- Pass CodeIndexManager to prompt sections for availability checks
- Ensure graceful degradation when codebase_search is unavailable
* chore(docs): remove codebase search enforcement documentation
* fix: update snapshot and reorder capabilities section
- Update system.test.ts snapshot to reflect architect mode without codebase_search enforcement
- Reorder capabilities section to place search_files description after codebase_search
- Ensures logical flow: codebase_search (semantic) → search_files (regex) → other tools
* refactor: improve tool-use-guidelines numbering logic
- Replace subsequentNumbers object with array-based approach
- Use automatic incrementing with itemNumber++ for sequential numbering
- Build guidelines as an array and join at the end
- Fix potential numbering issues with conditional logic
- Update tests and snapshots to match new format
As suggested by daniel-lxs in PR #4340
* feat: Add OpenAI Compatible embedder for codebase indexing
- Implement OpenAiCompatibleEmbedder with batching and retry logic
- Add configuration support for base URL and API key
- Update UI with provider selection and input fields
- Add comprehensive test coverage
- Support for all OpenAI-compatible endpoints (LiteLLM, LMStudio, Ollama, etc.)
- Add internationalization for 17 languages
* fix: Update CodeIndexSettings tests for OpenAI Compatible provider
- Fix field count expectations (4 fields including Qdrant)
- Use specific test IDs for button selection
- Fix input handling with clear() before type()
- Use toHaveBeenLastCalledWith for better assertions
- Fix status text matching with regex pattern
* fix: resolve UI test failures and ESLint errors
- Remove unused waitFor import to fix ESLint error
- Fix test expectations to match actual component behavior for input fields
- Simplify provider selection test by removing complex mock interactions
- All CodeIndexSettings tests now pass (20/20)
* feat: add custom model infrastructure for OpenAI-compatible embedder
- Add manual model ID and embedding dimension configuration
- Enable custom model input via text field in settings UI
- Add modelDimension parameter to OpenAiCompatibleEmbedder
- Update configuration management to persist dimension setting
- Prioritize manual dimension over hardcoded model profiles
- Add comprehensive test coverage for new functionality
This allows users to specify any custom embedding model and its
dimension for OpenAI-compatible providers, removing dependency
on hardcoded model profiles.
* Add missing translations for OpenAI-compatible model dimension settings in all locales
* refactor: remove unused modelDimension parameter from OpenAiCompatibleEmbedder
- Remove modelDimension property and constructor parameter from OpenAiCompatibleEmbedder class
- Update ServiceFactory to not pass dimension to embedder constructor
- Update tests to match new constructor signature
- The dimension is still used for QdrantVectorStore configuration
* chore: bot suggestion
Co-authored-by: ellipsis-dev[bot] <65095814+ellipsis-dev[bot]@users.noreply.github.com>
* chore: bot suggestion
Co-authored-by: ellipsis-dev[bot] <65095814+ellipsis-dev[bot]@users.noreply.github.com>
* refactor: rename OpenAiCompatibleEmbedder to OpenAICompatibleEmbedder for consistency
* feat: add model dimension validation for OpenAI-compatible settings
* refactor: improve default model ID retrieval logic for embedding providers
* feat: add default model ID retrieval for openai-compatible provider
* refactor: update default model ID retrieval to use shared utility function
* fix: Remove unnecessary type assertion in OpenAICompatibleEmbedder
* feat: add model dimension input for openai-compatible provider
---------
Co-authored-by: Daniel Riccio <ricciodaniel98@gmail.com>
Co-authored-by: Daniel <57051444+daniel-lxs@users.noreply.github.com>
Co-authored-by: ellipsis-dev[bot] <65095814+ellipsis-dev[bot]@users.noreply.github.com>
- Replace manual PATH and HOME env var handling with getDefaultEnvironment()
- Improves consistency and reliability of MCP client environment setup
- Leverages SDK's built-in environment configuration
- Allow specific binary formats (.pdf, .docx, .ipynb) to be processed by extractTextFromFile
- Block unsupported binary files with existing "Binary file" notice
- Update tests to cover both supported and unsupported binary file scenarios
- Refactor test mocks for better maintainability and coverage
* Improves model info detection for custom Bedrock ARNs
Adds heuristics to better estimate model capabilities when using unknown or custom model ARNs, including context window and max tokens. Allows user overrides for key model parameters via provider settings, improving flexibility and reliability for non-standard model integrations.
Fixes#3712
* Improves JSON syntax error handling in import flow
Provides more informative error messages for JSON syntax
errors by extracting the error position and formatting it
for clarity during import. Enhances user feedback when
invalid JSON is encountered.
* Fixed failing tests
* Delete pnpm-lock.yaml
* Added Rory's cache fix from PR #3099
PR #3009 has an important fix, alerted to me by @JBBrown
It was a one liner so I pulled it in.
This brings up a question can we merge PR's in the GH UI?
* Add Claude 4 and Opus 4 to modelID's
Kept previous parameters, did not see any changes in those.
* Fixed types being moved and me breaking the merge.
* Fix merge
---------
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
