Replace @anthropic-ai/vertex-sdk with @ai-sdk/google-vertex/anthropic,
using streamText/generateText from the Vercel AI SDK for consistent
provider behavior.
Changes:
- Use createVertexAnthropic from @ai-sdk/google-vertex/anthropic
- Use streamText/generateText instead of direct Anthropic API calls
- Add AI SDK transform utilities for message/tool conversion
- Handle cache control via AI SDK providerOptions
- Handle thinking/reasoning via providerOptions.anthropic.thinking
- Add thought signature and redacted thinking block tracking
- Set isAiSdkProvider() to return true
- Remove unused deps: @anthropic-ai/vertex-sdk, google-auth-library
- Rewrite tests to mock AI SDK instead of @anthropic-ai/vertex-sdk
* Latest main branch snapshot from API
* feat: add dedicated Azure OpenAI provider using @ai-sdk/azure package
* feat: add Azure provider UI component and translations
* feat: add Azure provider translations for all locales
* chore: add missing Azure placeholder translations
* Delete .changeset/azure-ai-sdk-migration.md
* fix: add Azure provider validation for onboarding workflow
- Add azureApiKey to SECRET_STATE_KEYS for proper configuration detection
- Add Azure validation case in validateModelsAndKeysProvided
- Add validation translations for azureResourceName and azureDeploymentName across all 18 locales
This fixes the issue where the Finish button does nothing when setting up Azure provider in the onboarding workflow.
* feat(azure): add model metadata, model picker, rename to Azure AI Foundry
- Add static model metadata for 29 Azure models (from models.dev)
with Roo-specific flags (reasoning, tools, verbosity) matching
openAiNativeModels
- Add model picker dropdown to Azure provider settings for model
capability detection (context window, max tokens, pricing)
- Rename provider label from 'Azure OpenAI' to 'Azure AI Foundry'
across all 18 locales
- Make API key optional (supports Azure managed identity / Entra ID)
- Update default API version from 2024-08-01-preview to 2025-04-01-preview
- Fix maxOutputTokens validation (filter invalid values <= 0)
- Handler separates deployment name (API calls) from model ID
(capability lookup) with azureDefaultModelInfo (gpt-4o) fallback
- Remove unhelpful 'Get Azure AI Foundry Access' button
- Prevent stale model IDs from other providers carrying over
- Suppress validation errors on fresh provider selection
* fix(azure): add missing isAiSdkProvider() override for reasoning block preservation
* Azure Fixes for Hannes
* Quick Fix for Respones API Only (for Hannes)
* fix: use explicit azureOpenAiDefaultApiVersion fallback when apiVersion is empty
Addresses review feedback: the UI placeholder shows '2025-04-01-preview' via
azureOpenAiDefaultApiVersion, so the handler should use the same constant as
fallback instead of silently deferring to the SDK's internal default.
* fix: remove stale Cerebras references (retired provider)
* fix: add missing retiredProviderMessage translations for all locales
* fix: do not map promptCacheMissTokens to cacheWriteTokens for Azure
Azure uses OpenAI-compatible caching which does not report cache write
tokens separately. promptCacheMissTokens represents tokens NOT found in
cache (processed from scratch), not tokens written to cache. This aligns
the Azure handler with the OpenAI native handler behavior.
---------
Co-authored-by: Hannes Rudolph <hrudolph@gmail.com>
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: daniel-lxs <ricciodaniel98@gmail.com>
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
* feat: migrate OpenAI Native provider to @ai-sdk/openai
Replace the raw OpenAI SDK (openai) usage in OpenAiNativeHandler with
@ai-sdk/openai and AI SDK's streamText/generateText, following the same
pattern used by other migrated providers (Groq, xAI, Fireworks, etc.).
Key changes:
- Use createOpenAI from @ai-sdk/openai with provider.responses() for
the Responses API
- Use streamText/generateText from ai for streaming and completions
- Pass OpenAI-specific features via providerOptions.openai (store,
reasoningEffort, reasoningSummary, textVerbosity, serviceTier,
promptCacheRetention, parallelToolCalls, include)
- Capture responseId, serviceTier, and encrypted reasoning content
from providerMetadata after streaming
- Preserve getEncryptedContent() and getResponseId() for Task.ts
- Preserve service tier pricing adjustment in cost calculation
- Mark as isAiSdkProvider: true
- Eliminate ~1100 lines of manual SSE parsing, raw fetch fallback,
and event handling code
- Rewrite all 3 test files to use AI SDK mocking pattern
* fix: remove non-existent cacheWriteTokens from providerMetadata
The OpenAI Responses API does not report cache write tokens separately.
Remove the reference to providerMetadata?.openai?.cacheWriteTokens which
does not exist in the @ai-sdk/openai provider metadata schema.
* fix: filter standalone encrypted reasoning items from messages
Task.ts buildCleanConversationHistory injects standalone reasoning items
with { type: 'reasoning', encrypted_content: '...' } into the messages
array. These have no 'role' property and would be silently dropped by
convertToAiSdkMessages. Filter them explicitly to prevent confusion.
Note: Encrypted reasoning content round-tripping for stateless continuity
is a known limitation of the AI SDK migration. The @ai-sdk/openai
provider does not support injecting raw Responses API reasoning items.
Plain-text reasoning round-tripping works correctly via isAiSdkProvider().
* fix: restore reasoning round-trip for OpenAI Responses API via AI SDK
- Strip plain-text reasoning blocks from assistant messages before
convertToAiSdkMessages() to eliminate 'Non-OpenAI reasoning parts'
warnings from @ai-sdk/openai Responses provider
- Re-inject encrypted reasoning items as AI SDK reasoning parts with
providerOptions.openai.itemId and reasoningEncryptedContent, restoring
reasoning continuity that was silently broken after the migration
- Restructure createMessage() into a 5-step pipeline:
collect → filter → strip → convert → inject
- Add 21 new tests for both plain-text stripping and encrypted
reasoning injection
---------
Co-authored-by: Hannes Rudolph <hrudolph@gmail.com>
* refactor: migrate Anthropic provider to @ai-sdk/anthropic
Replace the raw @anthropic-ai/sdk implementation with @ai-sdk/anthropic
(Vercel AI SDK) for consistency with other providers (Bedrock, DeepSeek,
Mistral, etc.).
Changes:
- Replace Anthropic() client with createAnthropic() from @ai-sdk/anthropic
- Replace manual stream parsing with streamText() + processAiSdkStreamPart()
- Replace client.messages.create() with generateText() for completePrompt()
- Use convertToAiSdkMessages() and convertToolsForAiSdk() for format conversion
- Handle prompt caching via AI SDK providerOptions (cacheControl on messages)
- Handle extended thinking via providerOptions.anthropic.thinking
- Add getThoughtSignature() and getRedactedThinkingBlocks() for thinking
signature round-tripping (matching Bedrock pattern, improves on original
which had a TODO for this)
- Add isAiSdkProvider() returning true
- Update tests to mock @ai-sdk/anthropic and ai instead of raw SDK
* fix: address PR review - remove apiKey fallback and use system+systemProviderOptions pattern
* refactor: remove 9 low-usage providers (Phase 0)
Remove Cerebras, Chutes, DeepInfra, Doubao, Featherless, Groq,
Hugging Face, IO Intelligence, and Unbound providers from the codebase.
Each provider removal includes: handler, tests, model definitions,
type schemas, UI settings components, fetchers, i18n references,
and all wiring in shared registration/config files.
- Delete 42 provider-specific files (handlers, tests, fetchers, UI components)
- Remove @ai-sdk/cerebras and @ai-sdk/groq npm dependencies
- Clean provider references from 68 shared files across src/, packages/types/,
webview-ui/, and apps/cli/
- Remove ~490 dead i18n translation keys across 36 locale files
- Add docs/ai-sdk-migration-guide.md with updated migration status
- All TypeScript checks pass, 6505 tests pass with 0 failures
* feat: show retired-provider message for removed provider profiles
Preserve API profiles that reference removed providers instead of
silently stripping their apiProvider. When a user selects a profile
configured for a retired provider, the settings UI now shows an
empathetic message explaining the removal instead of the provider
configuration form.
- Add retiredProviderNames array and isRetiredProvider() helper to
packages/types/src/provider-settings.ts
- Update ProviderSettingsManager sanitization to preserve retired
providers (only strip truly unknown values)
- Update ContextProxy sanitization to preserve retired providers
- Render retired-provider message in ApiOptions.tsx when selected
provider is in the retired list
- Add tests for sanitization, ContextProxy, and UI behavior
* feat: add retired-provider warning banner in chat view
* Revert "feat: add retired-provider warning banner in chat view"
This reverts commit dd593e1056.
* feat: show retired-provider message as inline chat response
* fix: show retired provider warning on home screen
Move WarningRow outside {task && ...} conditional so it renders
regardless of task state. Preserve user input on retired provider
intercept so text isn't lost when switching providers.
- Move showRetiredProviderWarning WarningRow to unconditional render
area near ProfileViolationWarning
- Remove setInputValue/setSelectedImages clearing from retired
provider early return in handleSendMessage
- Delete unused RetiredProviderWarning.tsx (dead code)
* fix: address PR review — passthrough retired-provider fields and i18n strings
- Use passthrough() in saveConfig() and load() so legacy provider-specific
fields (e.g. groqApiKey, deepInfraModelId) are preserved instead of
silently stripped by strict Zod parse()
- Move hardcoded English strings in ApiOptions.tsx and ChatView.tsx to
i18n translation keys (settings:providers.retiredProviderMessage,
chat:retiredProvider.{title,message,openSettings})
- Update tests to assert legacy provider-specific fields survive
save and load round-trips
* i18n: add retired-provider translations for all 17 locales
Translate providers.retiredProviderMessage (settings) and
retiredProvider.{title,message,openSettings} (chat) into ca, de, es,
fr, hi, id, it, ja, ko, nl, pl, pt-BR, ru, tr, vi, zh-CN, zh-TW.
* test: update ApiOptions retired-provider test to expect i18n key
* feat: add lock toggle to pin API config across all modes in workspace
Add a lock/unlock toggle inside the API config selector popover (next to
the settings gear) that, when enabled, applies the selected API
configuration to all modes in the current workspace.
- Add lockApiConfigAcrossModes to ExtensionState and WebviewMessage types
- Store setting in workspaceState (per-workspace, not global)
- When locked, activateProviderProfile sets config for all modes
- Lock icon in ApiConfigSelector popover bottom bar next to gear
- Full i18n: English + 17 locale translations (all mention workspace scope)
- 9 new tests: 2 ClineProvider, 2 handler, 5 UI (77 total pass)
* refactor: replace write-fan-out with read-time override for lock API config
The original lock implementation used setModeConfig() fan-out to write the
locked config to ALL modes globally. Since the lock flag lives in workspace-
scoped workspaceState but modeApiConfigs are in global secrets, this caused
cross-workspace data destruction.
Replaced with read-time guards:
- handleModeSwitch: early return when lock is on (skip per-mode config load)
- createTaskWithHistoryItem: skip mode-based config restoration under lock
- activateProviderProfile: removed fan-out block
- lockApiConfigAcrossModes handler: simplified to flag + state post only
- Fixed pre-existing workspaceState mock gap in ClineProvider.spec.ts and
ClineProvider.sticky-profile.spec.ts
* refactor: migrate baseten provider to AI SDK
* refactor(baseten): migrate to native @ai-sdk/baseten package
Replace OpenAICompatibleHandler with dedicated @ai-sdk/baseten package,
following the same pattern used by other native AI SDK providers (groq,
deepseek, etc.). This uses createBaseten() for provider initialization
and extends BaseProvider directly instead of the generic OpenAI-compatible
handler.
* feat: migrate Bedrock provider to AI SDK
Replace the raw AWS SDK (@aws-sdk/client-bedrock-runtime) Bedrock handler
with the Vercel AI SDK (@ai-sdk/amazon-bedrock). Reduces provider from
1,633 lines to 575 lines (65% reduction).
Key changes:
- Use streamText()/generateText() instead of ConverseStreamCommand/ConverseCommand
- Use createAmazonBedrock() with native auth (access key, secret, session,
profile via credentialProvider, API key, VPC endpoint as baseURL)
- Reasoning config via providerOptions.bedrock.reasoningConfig
- Anthropic beta headers via providerOptions.bedrock.anthropicBeta
- Thinking signature captured from providerMetadata.bedrock.signature
on reasoning-delta stream events
- Thinking signature round-tripped via providerOptions.bedrock.signature
on reasoning parts in convertToAiSdkMessages()
- Redacted thinking captured from providerMetadata.bedrock.redactedData
- isAiSdkProvider() returns true for reasoning block preservation
- Keep: getModel, ARN parsing, cross-region inference, cost calculation,
service tier pricing, 1M context beta
Tests: 83 tests skipped (mock old AWS SDK internals, need rewrite for
AI SDK mocking). 106 tests pass. 0 tests fail.
* fix: address review feedback for Bedrock AI SDK migration
- Wire usePromptCache into AI SDK via providerOptions.bedrock.cachePoint
on system prompt and last two user messages
- Remove debug logger.info that fires on every stream event with
providerMetadata
- Tighten isThrottlingError to match 'rate limit' instead of broad
'rate'/'limit' substrings that false-positive on context length errors
- Use shared handleAiSdkError utility for consistent error handling
with status code preservation for retry logic
* fix: bedrock AI SDK migration - fix usage metrics, rewrite tests, remove dead code
- Fix reasoningTokens always 0 (usage.details?.reasoningTokens → usage.reasoningTokens)
- Fix cacheReadInputTokens always 0 (read from usage.inputTokenDetails instead of providerMetadata)
- Fix invokedModelId not extracted for prompt router cost calculation
- Rewrite all 6 skipped bedrock test suites for AI SDK mocking pattern (140 tests pass)
- Remove dead code: bedrock-converse-format.ts, cache-strategy/ (6 files, ~2700 lines)
* chore: remove dead @anthropic-ai/bedrock-sdk dep and stale AWS SDK mocks
* chore: update pnpm-lock.yaml after removing @anthropic-ai/bedrock-sdk
* fix: compute cache point indices from original Anthropic messages before AI SDK conversion
The previous approach naively targeted the last 2 user messages in the
post-conversion AI SDK array, but convertToAiSdkMessages() splits user
messages containing tool_results into separate tool + user messages,
causing cache points to land on the wrong messages (tiny text fragments
instead of the intended meaty user turns).
Now we identify the last 2 user messages in the original Anthropic
message array (matching the Anthropic provider's caching strategy) and
build a parallel-walk mapping to apply cachePoint to the correct
corresponding AI SDK message.
* perf: optimize prompt caching with 3-point message strategy + anchor for 20-block window
Previous approach only cached the last 2 user messages (using 2 of 4
available cache checkpoints for messages). This left significant cache
savings on the table for longer conversations.
New strategy uses up to 3 message cache points (+ 1 system = 4 total):
- Last user message: write to cache for next request
- Second-to-last user message: read from cache for current request
- Anchor message at ~1/3 position: ensures the 20-block lookback window
from the second-to-last breakpoint hits a stable cache entry, covering
all assistant/tool messages in the middle of the conversation
Also extracted the parallel-walk mapping logic into a reusable
applyCachePointsToAiSdkMessages() helper method.
Industry benchmarks show 70-95% token cache rates are achievable;
this change should significantly improve our 39% baseline for longer
multi-turn conversations.
* chore: remove stale bedrock-sdk external, fix arnInfo property name, remove unused exports
---------
Co-authored-by: daniel-lxs <ricciodaniel98@gmail.com>
* feat: migrate Gemini and Vertex providers to AI SDK
- Migrate GeminiHandler from @google/genai to @ai-sdk/google
- Create standalone VertexHandler using @ai-sdk/google-vertex
- Use shared AI SDK utilities (streamText, generateText, convertToAiSdkMessages)
- Support thinkingConfig via providerOptions.google.thinkingConfig
- Support Google Search and URL Context grounding tools
- Preserve cost calculation with tiered pricing
- Remove gemini-format.ts (AI SDK handles message conversion)
EXT-643
* fix: remove unused import and implement allowedFunctionNames tool filtering
- Remove unused handleAiSdkError import from gemini.ts
- Implement tool filtering based on allowedFunctionNames in both
GeminiHandler and VertexHandler createMessage methods
- Filter tools before converting to AI SDK format to restrict
model access to only allowed functions
* feat: add AWS Bedrock support for codebase indexing
- Add bedrock as a new EmbedderProvider type
- Add AWS Bedrock embedding model profiles (titan-embed-text models)
- Create BedrockEmbedder class with support for Titan and Cohere models
- Add Bedrock configuration support to config manager and interfaces
- Update service factory to create BedrockEmbedder instances
- Add comprehensive tests for BedrockEmbedder
- Add localization strings for Bedrock support
Closes#8658
* fix: add missing bedrockOptions to loadConfiguration return type
* Fix various issues that the original PR missed.
* Remove debug logs
* Rename AWS Bedrock -> Amazon Bedrock
* Remove some 'as any's
* Revert README changes
* Add translations
* More translations
* Remove leftover code from a debugging session.
* fix: add bedrock to codebaseIndexModelsSchema and update brace-expansion override
- Add bedrock provider to codebaseIndexModelsSchema type definition to fix empty model dropdown in UI
- Update pnpm override for brace-expansion from '>=2.0.2' to '^2.0.2' to resolve ESM/CommonJS compatibility issues
* Improvements to AWS Bedrock embeddings support
- Enhanced bedrock.ts embedder implementation
- Added comprehensive test coverage in bedrock.spec.ts
- Updated config-manager.ts for better Bedrock configuration handling
- Improved service-factory.ts integration
- Updated embeddingModels.ts with Bedrock models
- Enhanced CodeIndexPopover.tsx UI for Bedrock options
- Added auto-populate test for CodeIndexPopover
- Updated pnpm-lock.yaml dependencies
* Restore openrouter config
* Remove debug log
* Fix config-manager.spec.ts unit test.
* Add translations for "optional"
* Revert unnecessary change related to open ia embedder
---------
Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
Co-authored-by: Smartsheet-JB-Brown <jb.brown@smartsheet.com>
* fix: Include nativeArgs in tool repetition detection
Fixes false positive 'stuck in a loop' error for native protocol tools
like read_file that store parameters in nativeArgs instead of params.
Previously, the ToolRepetitionDetector only compared the params object,
which was empty for native protocol tools. This caused all read_file
calls to appear identical, triggering false loop detection even when
reading different files.
Changes:
- Updated serializeToolUse() to include nativeArgs in comparison
- Added comprehensive tests for native protocol scenarios
- Maintains backward compatibility with XML protocol tools
Closes: Issue reported in Discord about read_file loop detection
* Try to use safe-stable-stringify in the tool repetition detector
---------
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
