feat: add Cerebras AI provider support and update release notes for v3.25.5 (#292)

Author: hannesrudolph

docs/providers/cerebras.md

@@ -0,0 +1,61 @@
+---
+sidebar_label: Cerebras
+description: Configure Cerebras AI's ultra-fast inference models in Roo Code. Access free and paid tiers with speeds up to 2600 tokens/second for coding and reasoning tasks.
+keywords:
+  - cerebras
+  - cerebras ai
+  - roo code
+  - api provider
+  - fast inference
+  - qwen coder
+  - llama models
+  - free tier
+  - high speed ai
+image: /img/social-share.jpg
+---
+
+# Using Cerebras With Roo Code
+
+Cerebras AI specializes in extremely fast inference speeds (up to 2600 tokens/second) with competitive pricing, including a free tier. Their models are optimized for coding, general intelligence, and reasoning tasks.
+
+**Website:** [https://cloud.cerebras.ai/](https://cloud.cerebras.ai/)
+
+---
+
+## Getting an API Key
+
+1. **Sign Up/Sign In:** Go to [Cerebras Cloud](https://cloud.cerebras.ai?utm_source=roocode). Create an account or sign in.
+2. **Navigate to API Keys:** Access the API keys section in your dashboard.
+3. **Create a Key:** Generate a new API key. Give it a descriptive name (e.g., "Roo Code").
+4. **Copy the Key:** **Important:** Copy the API key immediately. Store it securely.
+
+---
+
+## Supported Models
+
+Roo Code supports the following Cerebras models:
+
+* `qwen-3-coder-480b-free` (Default)
+* `qwen-3-coder-480b`
+* `qwen-3-235b-a22b-instruct-2507`
+* `llama-3.3-70b`
+* `qwen-3-32b`
+* `qwen-3-235b-a22b-thinking-2507`
+
+---
+
+## Configuration in Roo Code
+
+1. **Open Roo Code Settings:** Click the gear icon (<Codicon name="gear" />) in the Roo Code panel.
+2. **Select Provider:** Choose "Cerebras" from the "API Provider" dropdown.
+3. **Enter API Key:** Paste your Cerebras API key into the "Cerebras API Key" field.
+4. **Select Model:** Choose your desired model from the "Model" dropdown.
+
+---
+
+## Tips and Notes
+
+* **Performance:** Cerebras specializes in extremely fast inference speeds, making it ideal for real-time coding assistance.
+* **Free Tier:** The `qwen-3-coder-480b-free` model provides access to high-performance inference at no cost with rate limits.
+* **Context Windows:** Models support context windows ranging from 64K to 128K tokens.
+* **Pricing:** Refer to the [Cerebras Cloud](https://cloud.cerebras.ai?utm_source=roocode) dashboard for the latest pricing information.

docs/update-notes/index.md

@@ -19,6 +19,7 @@ This section contains notes about recent updates to Roo Code, listed by version
 
 ## Version 3.25
 
+*   [3.25.5](/update-notes/v3.25.5) (2025-08-01)
 *   [3.25.4](/update-notes/v3.25.4) (2025-07-31)
 *   [3.25.3](/update-notes/v3.25.3) (2025-07-30)
 *   [3.25.2](/update-notes/v3.25.2) (2025-07-29)

docs/update-notes/temp_pr_analysis.md

@@ -0,0 +1,11 @@
+## PR #5867: fix: improve Claude Code ENOENT error handling with installation guidance
+
+**Author**: app/roomote
+**Linked Issues**: #5866 (reporter: JamieJ1)
+**Category**: Bug Fix
+**User Impact**: Users encountering "spawn claude ENOENT" errors when trying to use Claude Code integration now receive clear, helpful error messages with installation instructions instead of cryptic system errors
+**Contributors**: app/roomote, JamieJ1
+**Documentation Needs**: docs-update
+**Documentation Notes**: The Claude Code integration documentation should be updated to include troubleshooting section for installation issues and PATH configuration
+
+---

docs/update-notes/temp_pr_analysis_v3.25.5.md

@@ -0,0 +1,44 @@
+## PR #6552: Fix VB.NET indexing by implementing fallback chunking system
+
+**Author**: daniel-lxs
+**Linked Issues**: #6420 (reporter: JensvanZutphen)
+**Category**: Bug Fix
+**User Impact**: VB.NET files are now properly indexed in large monorepos, showing all code chunks instead of just 6
+**Contributors**: daniel-lxs, JensvanZutphen
+**Documentation Needs**: docs-update
+**Documentation Notes**: The indexing documentation should be updated to mention that VB.NET and Scala files now use length-based chunking as a fallback when tree-sitter parsers are not available. This is part of the new fallback extension system.
+
+---
+## PR #6487: fix: restore message sending when clicking save button
+
+**Author**: daniel-lxs
+**Linked Issues**: None (but references PR #6222)
+**Category**: Bug Fix
+**User Impact**: Fixes a regression where messages were not being sent when clicking the save button in the chat interface
+**Contributors**: daniel-lxs
+**Documentation Needs**: none
+**Documentation Notes**: This is a bug fix that restores previous functionality - no documentation changes needed
+
+---
+## PR #6296: fix: handle Qdrant deletion errors gracefully to prevent indexing interruption
+
+**Author**: daniel-lxs
+**Linked Issues**: None
+**Category**: Bug Fix
+**User Impact**: Code indexing now properly deletes outdated entries when files are modified or removed, maintaining an accurate code index without interruption
+**Contributors**: daniel-lxs
+**Documentation Needs**: none
+**Documentation Notes**: This is a bug fix for the internal vector store implementation. No user-facing documentation changes needed.
+
+---
+## PR #6538: Clean up the auto-approve UI
+
+**Author**: mrubens
+**Linked Issues**: None
+**Category**: QOL Improvements
+**User Impact**: Cleaner and more streamlined auto-approve interface with simplified layout and improved localization
+**Contributors**: mrubens
+**Documentation Needs**: none
+**Documentation Notes**: UI cleanup only - no functional changes that require documentation
+
+---

docs/update-notes/temp_pr_list_v3.25.5.md

@@ -0,0 +1,24 @@
+# PRs for Release v3.25.5
+**Date Range**: 2025-07-31T04:15:51Z to 2025-08-01T19:41:21Z
+**Previous Version**: v3.25.4
+**Total PRs**: 18
+
+## PR List
+- #6565: Changeset version bump (app/github-actions)
+- #6564: Release v3.25.5 (mrubens)
+- #6562: Add Qwen 3 Coder from Cerebras (kevint-cerebras)
+- #6552: fix: Fix VB.NET indexing by implementing fallback chunking system (daniel-lxs)
+- #6538: Clean up the auto-approve UI (mrubens)
+- #6537: More tolerant search/replace match (mrubens)
+- #6525: Revert "Migrate evals database when deploying roo-code-website" (mrubens)
+- #6519: Cloud service cleanup callbacks / move to events (jr)
+- #6487: fix: restore message sending when clicking save button (daniel-lxs)
+- #6486: Handle more variations of chaining and subshell command validation (mrubens)
+- #6484: Add auto-approved cost limits (hassoncs)
+- #6477: fix: linter not applied to locales/*/README.md (liwilliam2021)
+- #6475: Skip interpolation for non-existent slash commands (app/roomote)
+- #6392: Add Cerebras as a provider (kevint-cerebras)
+- #6296: fix: handle Qdrant deletion errors gracefully to prevent indexing interruption (daniel-lxs)
+- #6183: fix: LM Studio model context length (#5075) (pwilkin)
+- #6085: Phase 1 website updates (thill2323)
+- #5867: fix: improve Claude Code ENOENT error handling with installation guidance (app/roomote)

docs/update-notes/v3.25.5.mdx

@@ -0,0 +1,61 @@
+---
+description: Roo Code 3.25.5 adds Cerebras AI provider support with Qwen 3 Coder models, introduces auto-approved cost limits for budget control, and fixes VB.NET indexing along with several other improvements.
+keywords:
+  - roo code 3.25.5
+  - cerebras provider
+  - qwen 3 coder
+  - cost limits
+  - vb.net indexing
+  - bug fixes
+image: /img/social-share.jpg
+---
+
+# Roo Code 3.25.5 Release Notes (2025-08-01)
+
+This release adds Cerebras AI provider support with powerful Qwen 3 Coder models, introduces auto-approved cost limits for better budget control, and includes important bug fixes for VB.NET indexing and message handling.
+
+## Cerebras Provider Support
+
+We've added support for Cerebras as a new AI provider (thanks kevint-cerebras!) ([#6392](https://github.com/RooCodeInc/Roo-Code/pull/6392), [#6562](https://github.com/RooCodeInc/Roo-Code/pull/6562)):
+
+- **Qwen 3 Coder Models**: Access to Qwen 3 Coder with both free and paid tier options
+- **Multiple Model Variants**: Includes Llama 3.3 70B and various Qwen 3 configurations (32B, 235B)
+- **Automatic Thinking Token Filtering**: Cerebras reasoning models automatically filter thinking tokens for cleaner output
+
+The Cerebras provider offers competitive performance with flexible pricing tiers, making it an excellent choice for both experimentation and production use.
+
+> **📚 Documentation**: See [Cerebras Provider Guide](/providers/cerebras) for setup instructions and available models.
+
+## Auto-approved Cost Limits
+
+We've introduced a new cost control feature in the auto-approve settings to help manage API spending (thanks hassoncs!) ([#6484](https://github.com/RooCodeInc/Roo-Code/pull/6484)):
+
+- **Budget Protection**: Set a maximum cost limit in your auto-approve settings to control total API spending
+- **Automatic Prompting**: Roo Code will prompt for approval when approaching your cost limit
+- **Complements Request Limits**: Works alongside existing request count limits for comprehensive control
+
+<img src="/img/v3.25.5/v3.25.5.png" alt="Auto-approve settings showing the new Max Cost field" width="400" />
+
+This feature provides peace of mind when using expensive models or running long sessions, ensuring you stay within budget. You can find this new "Max Cost" setting alongside the existing "Max Count" in the auto-approve configuration panel.
+
+## QOL Improvements
+
+* **Auto-approve UI**: Cleaner and more streamlined interface with improved localization ([#6538](https://github.com/RooCodeInc/Roo-Code/pull/6538))
+![alt text](../../static/img/v3.25.5/v3.25.5-1.png)
+* **Command Validation**: Better handling of background execution (`&`) and subshell patterns like `(cmd1; cmd2)` ([#6486](https://github.com/RooCodeInc/Roo-Code/pull/6486))
+
+## Bug Fixes
+
+* **VB.NET Indexing**: Files are now properly indexed in large monorepos using fallback chunking (thanks JensvanZutphen!) ([#6552](https://github.com/RooCodeInc/Roo-Code/pull/6552))
+* **Message Sending**: Restored functionality when clicking the save button ([#6487](https://github.com/RooCodeInc/Roo-Code/pull/6487))
+* **Search/Replace Tolerance**: More forgiving of AI-generated diffs with extra `>` characters ([#6537](https://github.com/RooCodeInc/Roo-Code/pull/6537))
+* **Qdrant Deletion**: Gracefully handles deletion errors to prevent indexing interruption ([#6296](https://github.com/RooCodeInc/Roo-Code/pull/6296))
+* **LM Studio Context Length**: Models now correctly display their actual context length instead of "1" (thanks pwilkin, Angular-Angel!) ([#6183](https://github.com/RooCodeInc/Roo-Code/pull/6183))
+* **Claude Code Errors**: ENOENT errors now show helpful installation guidance (thanks JamieJ1!) ([#5867](https://github.com/RooCodeInc/Roo-Code/pull/5867))
+
+## Misc Improvements
+
+* **Slash Command Interpolation**: Skip interpolation for non-existent commands ([#6475](https://github.com/RooCodeInc/Roo-Code/pull/6475))
+* **Linter Coverage**: Applied to locale README files (thanks liwilliam2021!) ([#6477](https://github.com/RooCodeInc/Roo-Code/pull/6477))
+* **Cloud Service Events**: Migrated from callbacks to event-based architecture ([#6519](https://github.com/RooCodeInc/Roo-Code/pull/6519))
+* **Website Updates**: Phase 1 improvements (thanks thill2323!) ([#6085](https://github.com/RooCodeInc/Roo-Code/pull/6085))

docs/update-notes/v3.25.mdx

@@ -1,19 +1,19 @@
 ---
-description: Roo Code 3.25 introduces custom slash commands, Gemini web tools, image support, message queueing, and many quality-of-life improvements.
+description: Roo Code 3.25 introduces custom slash commands, Gemini web tools, image support, message queueing, Cerebras AI provider, cost limits, and many quality-of-life improvements.
 keywords:
   - roo code 3.25
   - custom slash commands
   - gemini tools
   - image support
   - message queueing
+  - cerebras provider
+  - cost limits
 image: /img/social-share.jpg
 ---
 
 # Roo Code 3.25 Release Notes
 
-## 3.25.0 (2025-07-29)
-
-This release brings powerful new capabilities to Roo Code, including custom slash commands for workflow automation, enhanced Gemini models with web access, comprehensive image support, and seamless message queueing for uninterrupted conversations.
+This release brings powerful new capabilities to Roo Code, including custom slash commands for workflow automation, enhanced Gemini models with web access, comprehensive image support, seamless message queueing for uninterrupted conversations, the new Cerebras AI provider, and auto-approved cost limits for budget control.
 
 ### Custom Slash Commands
 
@@ -80,6 +80,18 @@ To enable this feature: Settings → Prompts tab → Select "ENHANCE" → Check
 
 > **📚 Documentation**: See [Prompt Enhancement Guide](/features/enhance-prompt) for configuration options.
 
+### Auto-approved Cost Limits
+
+We've introduced a new cost control feature in the auto-approve settings to help manage API spending:
+
+- **Budget Protection**: Set a maximum cost limit in your auto-approve settings to control total API spending
+- **Automatic Prompting**: Roo Code will prompt for approval when approaching your cost limit
+- **Complements Request Limits**: Works alongside existing request count limits for comprehensive control
+
+<img src="/img/v3.25.5/v3.25.5.png" alt="Auto-approve settings showing the new Max Cost field" width="600" />
+
+This feature provides peace of mind when using expensive models or running long sessions, ensuring you stay within budget. You can find this new "Max Cost" setting alongside the existing "Max Count" in the auto-approve configuration panel.
+
 ### Quality of Life Improvements
 
 Small changes that make a big difference in your daily workflow:
@@ -90,9 +102,11 @@ Small changes that make a big difference in your daily workflow:
   <img src="/img/v3.25.0/v3.25.0.png" alt="Mode Selector Popover with Search" width="600" />
 * **API Selector Popover Redesign**: Updated to match the new mode selector design with improved layout ([#6148](https://github.com/RooCodeInc/Roo-Code/pull/6148))
   <img src="/img/v3.25.0/v3.25.0-1.png" alt="API Configuration Selector Redesign" width="600" />
+* **Auto-approve UI**: Cleaner and more streamlined interface with improved localization
 * **Sticky Task Modes**: Tasks remember their last-used mode and restore it automatically ([#6177](https://github.com/RooCodeInc/Roo-Code/pull/6177))
 * **ESC Key Support**: Close popovers with ESC for better keyboard navigation ([#6175](https://github.com/RooCodeInc/Roo-Code/pull/6175))
 * **Improved Command Highlighting**: Only valid commands are highlighted in the input field ([#6336](https://github.com/RooCodeInc/Roo-Code/pull/6336))
+* **Command Validation**: Better handling of background execution (`&`) and subshell patterns like `(cmd1; cmd2)`
 * **Subshell Validation**: Improved handling and validation of complex shell commands with subshells, preventing potential errors when using command substitution patterns ([#6379](https://github.com/RooCodeInc/Roo-Code/pull/6379))
 * **Slash Command Icon Hover State**: Fixed the hover state for the slash command icon to provide better visual feedback during interactions ([#6388](https://github.com/RooCodeInc/Roo-Code/pull/6388))
 * **Cleaner Welcome View**: Gemini checkboxes are now hidden on the welcome screen for a cleaner interface ([#6415](https://github.com/RooCodeInc/Roo-Code/pull/6415))
@@ -106,6 +120,12 @@ Small changes that make a big difference in your daily workflow:
 
 Critical fixes that improve stability and compatibility:
 
+* **VB.NET Indexing**: Files are now properly indexed in large monorepos using fallback chunking
+* **Message Sending**: Restored functionality when clicking the save button
+* **Search/Replace Tolerance**: More forgiving of AI-generated diffs with extra `>` characters
+* **Qdrant Deletion**: Gracefully handles deletion errors to prevent indexing interruption
+* **LM Studio Context Length**: Models now correctly display their actual context length instead of "1"
+* **Claude Code Errors**: ENOENT errors now show helpful installation guidance
 * **Multi-file Edit Fix**: Fixed issue where Git diff views interfered with file operations (thanks hassoncs, szermatt!) ([#6350](https://github.com/RooCodeInc/Roo-Code/pull/6350))
 * **Non-QWERTY Keyboard Support**: Fixed keyboard shortcuts for Dvorak, AZERTY, and other layouts (thanks shlgug!) ([#6162](https://github.com/RooCodeInc/Roo-Code/pull/6162))
 * **Mode Export/Import**: Fixed custom mode export to handle slug changes correctly ([#6186](https://github.com/RooCodeInc/Roo-Code/pull/6186))
@@ -126,6 +146,7 @@ Critical fixes that improve stability and compatibility:
 
 ### Provider Updates
 
+* **Cerebras Provider**: Added support for Cerebras as a new AI provider with Qwen 3 Coder models, offering both free and paid tier options with automatic thinking token filtering
 * **Prompt Caching for LiteLLM**: Reduce API costs and improve response times with caching support for Claude 3.5 Sonnet and compatible models (thanks MuriloFP, steve-gore-snapdocs!) ([#6074](https://github.com/RooCodeInc/Roo-Code/pull/6074))
 * **Chutes AI - GLM-4.5-Air Model**: Added support for the GLM-4.5-Air model to the Chutes AI provider with 151K token context window for complex reasoning tasks and large codebase analysis - completely free to use (thanks matbgn!) ([#6377](https://github.com/RooCodeInc/Roo-Code/pull/6377))
 * **Doubao Provider**: Added support for ByteDance's AI model provider with full integration including API handling (thanks AntiMoron!) ([#6345](https://github.com/RooCodeInc/Roo-Code/pull/6345))
@@ -136,6 +157,10 @@ Critical fixes that improve stability and compatibility:
 
 ### Misc. Improvements
 
+* **Slash Command Interpolation**: Skip interpolation for non-existent commands
+* **Linter Coverage**: Applied to locale README files
+* **Cloud Service Events**: Migrated from callbacks to event-based architecture
+* **Website Updates**: Phase 1 improvements
 * **Background Editing (Experimental)**: Work uninterrupted while Roo edits files in the background—no more losing focus from automatic diff views ([#6214](https://github.com/RooCodeInc/Roo-Code/pull/6214)). Files change silently while you keep coding, with diagnostics and error checking still active. See [Background Editing](/features/experimental/background-editing) for details.
 * **Security Update**: Updated form-data dependency to address security vulnerability ([#6332](https://github.com/RooCodeInc/Roo-Code/pull/6332))
 * **Contributor Updates**: Refreshed contributor acknowledgments across all localizations ([#6302](https://github.com/RooCodeInc/Roo-Code/pull/6302))

sidebars.ts

@@ -158,6 +158,7 @@ const sidebars: SidebarsConfig = {
         'providers/anthropic',
         'providers/claude-code',
         'providers/bedrock',
+        'providers/cerebras',
         'providers/deepseek',
         'providers/chutes',
         'providers/gemini',
@@ -206,6 +207,7 @@ const sidebars: SidebarsConfig = {
           label: '3.25',
           items: [
             { type: 'doc', id: 'update-notes/v3.25', label: '3.25 Combined' },
+            { type: 'doc', id: 'update-notes/v3.25.5', label: '3.25.5' },
             { type: 'doc', id: 'update-notes/v3.25.4', label: '3.25.4' },
             { type: 'doc', id: 'update-notes/v3.25.3', label: '3.25.3' },
             { type: 'doc', id: 'update-notes/v3.25.2', label: '3.25.2' },