* fix: add cache reporting support for OpenAI-Native provider
- Add normalizeUsage method to properly extract cache tokens from Responses API
- Support both detailed token shapes (input_tokens_details) and legacy fields
- Calculate cache read/write tokens with proper fallbacks
- Include reasoning tokens when available in output_tokens_details
- Ensure accurate cost calculation using uncached input tokens
This fixes the issue where caching information was not being reported
when using the OpenAI-Native provider with the Responses API.
* fix: improve cache token normalization and add comprehensive tests
- Add fallback to derive total input tokens from details when totals are missing
- Remove unused convertToOpenAiMessages import
- Add comment explaining cost calculation alignment with Gemini provider
- Add comprehensive test coverage for normalizeUsage method covering:
- Detailed token shapes with cached/miss tokens
- Legacy field names and SSE-only events
- Edge cases including missing totals with details-only
- Cost calculation with uncached input tokens
* fix: address PR review comments
- Remove incorrect fallback to missFromDetails for cache write tokens
- Fix cost calculation to pass total input tokens (calculateApiCostOpenAI handles subtraction)
- Improve readability by extracting cache detail checks to intermediate variables
- Remove redundant ?? undefined
- Update tests to reflect correct behavior (miss tokens are not cache writes)
- Add clarifying comments about cache miss vs cache write tokens
* feat: add Ollama API key support for Turbo mode
- Add ollamaApiKey field to ProviderSettings schema
- Add ollamaApiKey to SECRET_STATE_KEYS for secure storage
- Update Ollama and NativeOllama providers to use API key for authentication
- Add UI field for Ollama API key (shown when custom base URL is provided)
- Add test coverage for API key functionality
This enables users to use Ollama Turbo with datacenter-grade hardware by providing an API key for authenticated Ollama instances or cloud services.
* fix: use VSCodeTextField for Ollama API key field
Remove non-existent ApiKeyField import and use standard VSCodeTextField with password type, matching other provider implementations
* Add missing translation keys for Ollama API key support
- Add providers.ollama.apiKey and providers.ollama.apiKeyHelp to all 18 language files
- Support for authenticated Ollama instances and cloud services
- Relates to PR #7425
* refactor: improve type safety for Ollama client configuration
- Replace 'any' type with proper OllamaOptions (Config) type
- Import Config type from ollama package for better type checking
---------
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: Daniel Riccio <ricciodaniel98@gmail.com>
* fix: handle Mistral thinking content as reasoning chunks
- Add TypeScript interfaces for Mistral content types (text and thinking)
- Update createMessage to yield reasoning chunks for thinking content
- Update completePrompt to filter out thinking content in non-streaming mode
- Add comprehensive tests for reasoning content handling
- Follow the pattern used by other providers (Anthropic, OpenAI, Gemini, etc.)
Fixes#6842
* fix: resolve TypeScript type issue in completePrompt method
* fix: handle Mistral thinking content chunks in streaming responses
- Added ContentChunkWithThinking type helper to handle thinking chunks
- Properly converts thinking content to reasoning chunks in streaming
- Filters out thinking content in non-streaming completePrompt responses
- Confirmed that Mistral API does send thinking chunks with type 'thinking'
- Works with Mistral SDK v1.9.18
---------
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: daniel-lxs <ricciodaniel98@gmail.com>
* fix: update DeepSeek models context window to 128k
- Updated deepseek-chat and deepseek-reasoner models from 64k to 128k context window
- Updated corresponding test expectations
- Aligns with DeepSeek API documentation at https://api-docs.deepseek.com/quick_start/pricing/Fixes#7268
* feat: update deepseek-reasoner maxTokens to 64K based on official documentation
* fix: use default maxTokens values instead of maximum for DeepSeek models
- deepseek-chat: 4096 (4K default) instead of 8192 (8K max)
- deepseek-reasoner: 32768 (32K default) instead of 65536 (64K max)
- Updated tests to match new default values
- Updated description to clarify default vs max output tokens
* fix: use maximum output tokens for both DeepSeek models
- deepseek-chat: 8192 (8K max)
- deepseek-reasoner: 65536 (64K max)
- Updated tests to match maximum values
- Updated description to reflect 64K max output
---------
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: daniel-lxs <ricciodaniel98@gmail.com>
* fix: omit temperature parameter when not explicitly set for OpenAI Compatible providers
- Modified OpenAiHandler to only include temperature when modelTemperature is defined
- Modified BaseOpenAiCompatibleProvider to only include temperature when modelTemperature is defined
- Added tests to verify temperature is omitted when undefined
- Updated existing tests to explicitly set temperature where needed
This allows backend services (LiteLLM, vLLM) to use their configured default temperatures
instead of being forced to use temperature=0 when "Use custom temperature" is unchecked.
Fixes#7187
* test: update tests to match new temperature handling behavior
- Remove temperature parameter expectations from provider tests
- Tests now expect temperature to be omitted when not explicitly set
- Aligns with PR #7188 changes to fix OpenAI Compatible provider behavior
---------
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: daniel-lxs <ricciodaniel98@gmail.com>
* fix: prevent duplicate LM Studio models with case-insensitive deduplication
- Keep both listDownloadedModels and listLoaded APIs to support JIT loading
- Implement case-insensitive deduplication to prevent duplicates
- When duplicates are found, prefer loaded model data for accurate runtime info
- Add test coverage for deduplication logic
- Addresses feedback about LM Studio's JIT Model Loading feature (v0.3.5+)
Fixes#6954
* fix: correct deduplication logic to prefer loaded models
- When a loaded model ID is found in any downloaded model key (case-insensitive)
- Remove the downloaded model and replace with the loaded model
- This ensures loaded models with runtime info take precedence
- Updated tests to verify the correct deduplication behavior
* fix: improve deduplication logic and add comprehensive test coverage
- Enhanced deduplication to use path segment matching instead of simple substring
- Prevents false positives like 'llama' matching 'codellama'
- Added comprehensive test cases for edge cases and multiple scenarios
- Maintains support for JIT Model Loading feature
* fix: add explicit max_output_tokens for GPT-5 Responses API
- Added max_output_tokens parameter to GPT-5 request body using model.maxTokens
- This prevents GPT-5 from defaulting to very large token limits (e.g., 120k)
- Updated tests to expect max_output_tokens in GPT-5 request bodies
- Fixed test for handling unhandled stream events by properly mocking SDK fallback
* fix: add missing translations for reasoningEffort.minimal in Indonesian and Dutch locales
* fix: correct GPT-5 response ID persistence and usage
- Renamed metadata field from 'previous_response_id' to 'response_id' for clarity
- Fixed logic to correctly use the response_id from the previous message as previous_response_id for the next request
- This resolves the 'Previous response with id not found' errors that occurred after multiple turns in the same session
* feat: add robust error handling for GPT-5 previous_response_id failures
- Automatically retry without previous_response_id when it's not found (400 error)
- Clear stored lastResponseId to prevent reusing stale IDs
- Handle errors in both SDK and SSE fallback paths
- Log warnings when retrying to help with debugging
* fix: handle GPT-5 response ID race condition with nano model
- Add promise-based synchronization for response ID persistence
- Wait for pending response ID from previous request before using it
- Resolve promise when response ID is received or cleared
- Add 100ms timeout to avoid blocking too long on ID resolution
- Properly clean up resolver on errors to prevent memory leaks
This fixes the race condition where fast nano model responses could cause
the next request to be initiated before the response ID was fully persisted.
* fix: address PR review comments for GPT-5 implementation
- Extract usage normalization helper to reduce duplication
- Suppress conversation continuity for first message (but respect explicit metadata)
- Deduplicate response ID resolver logic
- Remove dead enableGpt5ReasoningSummary option references
- DRY up GPT-5 event/usage handling with normalizeGpt5Usage helper
- Centralize default GPT-5 reasoning effort using model info
- Fix Indonesian locale minimal string misplacement
- Add clarifying comments for Developer prefix usage
- Add TODO for future verbosity UI capability gating
- Fix failing test in reasoning.spec.ts
* fix(openai-native): address Roomote inline feedback\n\n- Delegate standard GPT-5 SSE event types to shared processor to reduce duplication\n- Add JSDoc for response ID accessors\n- Standardize key error messages for GPT-5 Responses API fallback\n- Extract persistGpt5Metadata() in Task to simplify metadata writes\n- Add malformed JSON SSE parsing test\n
* fix(openai-native,gpt5): correct usage cost calc (use calculateApiCostOpenAI incl. cache); enforce 'skip once' continuity via suppressPreviousResponseId; dedupe responseId resolver on SSE 400; feat: gate reasoning.summary by enableGpt5ReasoningSummary; centralize default reasoning effort; types/ui: add ModelInfo.supportsVerbosity and gate Verbosity UI by capability; refactor: avoid duplicate usage emission in SSE done/completed
* fix(gpt5): default enableGpt5ReasoningSummary=true to preserve tests and expected behavior
* fix(gpt5): canonicalize GPT-5 metadata key to previous_response_id and align enableGpt5ReasoningSummary default docs
* fix(openai-native): remove review artifact comments and guard GPT-5 in completePrompt
* feat: add GPT-5 model support
- Added GPT-5 models (gpt-5-2025-08-07, gpt-5-mini-2025-08-07, gpt-5-nano-2025-08-07)
- Added nectarine-alpha-new-reasoning-effort-2025-07-25 experimental model
- Set gpt-5-2025-08-07 as default OpenAI Native model
- Implemented GPT-5 specific handling with streaming and reasoning effort support
* fix: remove hardcoded temperature from GPT-5 handler
- Updated handleGPT5Message to use configurable temperature
- Now uses this.options.modelTemperature ?? OPENAI_NATIVE_DEFAULT_TEMPERATURE
- Maintains consistency with other model handlers
* feat: add reasoning effort support for all OpenAI models
* fix: update test to expect new default model gpt-5-2025-08-07
* feat: increase GPT-5 models context window to 400,000
- Updated context window from 256,000 to 400,000 for gpt-5-2025-08-07
- Updated context window from 256,000 to 400,000 for gpt-5-mini-2025-08-07
- Updated context window from 256,000 to 400,000 for gpt-5-nano-2025-08-07
- Updated context window from 256,000 to 400,000 for nectarine-alpha-new-reasoning-effort-2025-07-25
As requested by @daniel-lxs in PR #6819
* revert: remove GPT-5 models, keep only nectarine experimental model
- Removed gpt-5-2025-08-07, gpt-5-mini-2025-08-07, gpt-5-nano-2025-08-07
- Kept nectarine-alpha-new-reasoning-effort-2025-07-25 experimental model
- Reverted default model back to gpt-4o
- Updated tests and changeset accordingly
* feat: add GPT-5 models with updated context windows
- Added gpt-5-2025-08-07, gpt-5-mini-2025-08-07, gpt-5-nano-2025-08-07 models
- All GPT-5 models configured with 400,000 context window
- Updated nectarine model context window to 256,000
- All models configured with reasoning effort support
- Set gpt-5-2025-08-07 as default OpenAI Native model
- Added GPT-5 model handling in openai-native.ts
- Updated tests to reflect new default model
* fix: restore reasoning effort support for o1 series models
- Added supportsReasoningEffort: true to o1, o1-preview, and o1-mini models
- This restores the ability to use reasoning effort parameters with these models
- The existing code in openai-native.ts already handles reasoning effort correctly
* Revert "fix: restore reasoning effort support for o1 series models"
This reverts commit 7251237ae8.
* fix: restore reasoning effort support for o3 and o4 models
- Added supportsReasoningEffort: true to o3, o3-high, o3-low models
- Added supportsReasoningEffort: true to o4-mini, o4-mini-high, o4-mini-low models
- Added supportsReasoningEffort: true to o3-mini, o3-mini-high, o3-mini-low models
- These models have both supportsReasoningEffort and reasoningEffort properties
* Revert "fix: restore reasoning effort support for o3 and o4 models"
This reverts commit a75a2b8a69.
* fix: restore reasoning effort support for o3 and o4 models
- Added supportsReasoningEffort: true to o3, o3-high, o3-low models
- Added supportsReasoningEffort: true to o4-mini, o4-mini-high, o4-mini-low models
- Added supportsReasoningEffort: true to o3-mini, o3-mini-high, o3-mini-low models
* fix: adjust reasoning effort support for o3/o4 models
- Keep supportsReasoningEffort only for base o3, o4-mini, and o3-mini models
- Remove supportsReasoningEffort from -high and -low variants
- Position supportsReasoningEffort right before reasoningEffort property
* fix: remove nectarine experimental model
- Removed nectarine-alpha-new-reasoning-effort-2025-07-25 from openai.ts
- Removed nectarine handling from openai-native.ts (renamed to handleGpt5Message)
- Removed associated changeset file
- Keep GPT-5 models with developer role handling
* feat: implement full GPT-5 support with verbosity and minimal reasoning
- Add all three GPT-5 models with accurate pricing (.25/0 for gpt-5, /bin/sh.25/ for mini, /bin/sh.05//bin/sh.40 for nano)
- Implement verbosity control (low/medium/high) that passes through to API
- Add minimal reasoning effort support for fastest response times
- GPT-5 models use developer role instead of system role
- Set gpt-5-2025-08-07 as default OpenAI Native model
- Add Responses API infrastructure for future migration
- Update tests to verify all GPT-5 features
- All 27 tests passing
Note: UI controls for verbosity still need to be added in a follow-up PR
* feat: add verbosity setting for GPT-5 models
- Add VerbosityLevel type definition to model types
- Add verbosity field to ProviderSettings schema
- Create Verbosity UI component for settings
- Add verbosity labels to all localization files
- Integrate verbosity handling in model parameters transformation
- Update OpenAI native handler to support verbosity for GPT-5
- Add comprehensive tests for verbosity setting
- Update existing GPT-5 tests to use verbosity from settings
* Delete .roorules
---------
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: hannesrudolph <hrudolph@gmail.com>
Co-authored-by: Daniel Riccio <ricciodaniel98@gmail.com>
Co-authored-by: Daniel <57051444+daniel-lxs@users.noreply.github.com>
* feat: add GLM-4.5 and OpenAI gpt-oss models to Fireworks provider
- Added GLM-4.5 (355B/32B active) and GLM-4.5-Air (106B/12B active) models from Z.ai
- Added gpt-oss-20b and gpt-oss-120b models from OpenAI
- All models configured with 128K context window
- Added comprehensive test coverage for all new models
Fixes#6753
* fix: update GLM-4.5 model IDs to use p instead of hyphen
- Changed glm-4-5 to glm-4p5
- Changed glm-4-5-air to glm-4p5-air
- Updated corresponding test cases
---------
Co-authored-by: Roo Code <roomote@roocode.com>
