Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 15 additions & 1 deletion docs/basic-usage/context-mentions.md
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ Context mentions are a powerful way to provide Roo Code with specific informatio
| Mention Type | Format | Description | Example Usage |
|--------------|--------|-------------|--------------|
| **File** | `@/path/to/file.ts` | Includes file contents in request context | "Explain the function in @/src/utils.ts" |
| **Image** | `@/path/to/image.png` | Includes image as inline visual content | "What's wrong with this UI? @/screenshots/bug.png" |
| **Folder** | `@/path/to/folder` | Includes contents of all files directly in the folder (non-recursive) | "Analyze the code in @/src/components" |
| **Problems** | `@problems` | Includes VS Code Problems panel diagnostics | "@problems Fix all errors in my code" |
| **Terminal** | `@terminal` | Includes recent terminal command and output | "Fix the errors shown in @terminal" |
Expand All @@ -50,6 +51,18 @@ Context mentions are a powerful way to provide Roo Code with specific informatio
| **Works in** | Initial requests, feedback responses, and follow-up messages |
| **Limitations** | Very large files may be truncated; binary files not supported |

### Image Mentions

Image mentions let you include visual content directly in your conversation. When the model supports vision, the image is sent as inline visual content rather than text.

| Capability | Details |
|------------|---------|
| **Format** | `@/path/to/image.png` (same path format as file mentions) |
| **Provides** | Image sent as inline visual content to the model |
| **Supports** | PNG, JPG, JPEG, GIF, BMP, SVG, WEBP, ICO, AVIF |
| **Best for** | UI reviews, screenshot debugging, diagram analysis |
| **Requires** | A model with vision support (non-vision models can't interpret images) |

### Folder Mentions

<img src="/img/context-mentions/context-mentions-2.png" alt="Folder mention example showing directory contents being referenced in the chat" width="600" />
Expand Down Expand Up @@ -130,7 +143,7 @@ The dropdown automatically suggests:
- Special keywords (`problems`, `terminal`, `git-changes`)
- **All currently open files** (regardless of ignore settings or directory filters)

The dropdown automatically filters out common directories like `node_modules`, `.git`, `dist`, and `out` to reduce noise, even though their content could be included if manually typed.
The dropdown respects `.rooignore` by default, hiding ignored files from suggestions. Enable the `showRooIgnoredFiles` setting to include ignored files in the dropdown (they'll appear with a 🔒 indicator). Common directories like `node_modules`, `.git`, `dist`, and `out` are also filtered to reduce noise.

---

Expand All @@ -140,6 +153,7 @@ The dropdown automatically filters out common directories like `node_modules`, `

| Behavior | Description |
|----------|-------------|
| **Dropdown filtering** | The `@` dropdown hides `.rooignore`-matched files by default. Enable `showRooIgnoredFiles` to see them (marked with 🔒). |
| **`.rooignore` bypass** | File and folder `@mentions` bypass `.rooignore` checks when fetching content for context. Content from ignored files will be included if directly mentioned. |
| **`.gitignore` bypass** | Similarly, file and folder `@mentions` do not respect `.gitignore` rules when fetching content. |
| **Git command respect** | Git-related mentions (`@git-changes`, `@commit-hash`) do respect `.gitignore` since they rely on Git commands. |
Expand Down
12 changes: 12 additions & 0 deletions docs/features/api-configuration-profiles.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -120,6 +120,18 @@ In the <Codicon name="notebook" /> Prompts tab, you can explicitly associate a s

---

## Per-Task Profile Persistence

Each task remembers which profile it started with. This "sticky" behavior means:

- **Reopening from history**: When you resume a task from history, it uses the same profile it had originally—even if you've since changed the global selection.
- **Multi-workspace consistency**: If you switch profiles in another workspace window, existing tasks in the first window keep their original profile.
- **Orchestrator subtasks**: Child tasks created by the orchestrator inherit the parent's profile and retain it for their lifetime.

This prevents unexpected model switches mid-task and keeps your conversation context consistent with the model that generated it.

---

## Security Note

API keys are stored securely in VSCode's Secret Storage and are never exposed in plain text.
Expand Down
9 changes: 9 additions & 0 deletions docs/update-notes/index.md
Original file line number Diff line number Diff line change
Expand Up @@ -22,9 +22,18 @@ If you want to live on the edge and try things out before it's released, we have

---

### Version 3.39

* [3.39](/update-notes/v3.39) (Combined)
* [3.39.1](/update-notes/v3.39.1) (2026-01-08)
* [3.39.0](/update-notes/v3.39.0) (2026-01-08)

---

### Version 3.38

* [3.38](/update-notes/v3.38) (Combined)
* [3.38.3](/update-notes/v3.38.3) (2026-01-03)
* [3.38.2](/update-notes/v3.38.2) (2025-12-31)
* [3.38.1](/update-notes/v3.38.1) (2025-12-29)
* [3.38.0](/update-notes/v3.38.0) (2025-12-27)
Expand Down
22 changes: 22 additions & 0 deletions docs/update-notes/v3.38.3.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
---
description: Roo Code 3.38.3 adds optional recursive rules loading for multi-folder setups and fixes several reliability issues.
keywords:
- roo code 3.38.3
- rules
- agents.md
- bug fixes
---

# Roo Code 3.38.3 Release Notes (2026-01-03)

This release adds an option to recursively load project rules in subfolders and fixes several reliability issues.

## QOL Improvements

* **Recursive subfolder rules loading (optional)**: Adds a Context setting that can automatically load additional `.roo/rules` and `AGENTS.md` files from subdirectories, making it easier to work in multi-repo/monorepo folders without manually switching roots ([#10446](https://github.com/RooCodeInc/Roo-Code/pull/10446))

## Bug Fixes

* Fixes an issue where Claude Code users could be forced to sign in again roughly daily, improving session reliability and reducing interruptions ([#10410](https://github.com/RooCodeInc/Roo-Code/pull/10410))
* Fixes an issue where the in-chat text-to-speech (TTS) flow could crash if a message’s `text` value wasn’t a string, preventing the chat panel from breaking mid-task (thanks notglossy!) ([#10431](https://github.com/RooCodeInc/Roo-Code/pull/10431))
* Fixes an issue where Roo’s native `read_file` tool schema didn’t advertise a maximum number of files per call, reducing failed tool calls in environments that enforce low limits ([#10449](https://github.com/RooCodeInc/Roo-Code/pull/10449))
4 changes: 4 additions & 0 deletions docs/update-notes/v3.38.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ Roo now supports Agent Skills, which are portable skill folders containing instr
* Slash commands can declare a target mode in their front matter, so triggering a command can switch Roo to the right mode first ([#10344](https://github.com/RooCodeInc/Roo-Code/pull/10344)).
* Removes the legacy “simple read file” tool path so file reading consistently uses the standard `read_file` tool ([#10254](https://github.com/RooCodeInc/Roo-Code/pull/10254)).
* Agent Skills load more reliably across environments by aligning skill metadata validation and discovery with the Agent Skills specification ([#10409](https://github.com/RooCodeInc/Roo-Code/pull/10409)).
* Adds an optional Context setting that can automatically load additional `.roo/rules` and `AGENTS.md` files from subdirectories, making it easier to work in multi-repo/monorepo folders without manually switching roots ([#10446](https://github.com/RooCodeInc/Roo-Code/pull/10446)).
* Clarifies the navigation path for deleting a Roo Code Cloud account in the privacy policy, so you can find Security Settings faster ([#10367](https://github.com/RooCodeInc/Roo-Code/pull/10367)).
* Improves internal analytics consistency by recording custom tools the same way as MCP tools, which helps tool usage and error tracking stay accurate ([#10364](https://github.com/RooCodeInc/Roo-Code/pull/10364)).

Expand All @@ -33,8 +34,11 @@ Roo now supports Agent Skills, which are portable skill folders containing instr
* Fixes an issue where some Claude Sonnet 4.5 requests could fail with HTTP 400 errors after context condensing ([#10359](https://github.com/RooCodeInc/Roo-Code/pull/10359)).
* Fixes an issue where manually condensing context while a tool is still running could trigger provider protocol errors, so tool runs and manual condenses can safely overlap ([#10379](https://github.com/RooCodeInc/Roo-Code/pull/10379)).
* Reverts a change for OpenAI-compatible providers to avoid message-format regressions related to text appearing after tool results ([#10381](https://github.com/RooCodeInc/Roo-Code/pull/10381)).
* Fixes an issue where Claude Code users could be forced to sign in again roughly daily, improving session reliability and reducing interruptions ([#10410](https://github.com/RooCodeInc/Roo-Code/pull/10410)).
* Enforces the `maxConcurrentFileReads` limit for `read_file`, returning a clear error instead of silently ignoring the setting ([#10363](https://github.com/RooCodeInc/Roo-Code/pull/10363)).
* Fixes an issue where the in-chat text-to-speech (TTS) flow could crash if a message’s `text` value wasn’t a string, preventing the chat panel from breaking mid-task (thanks notglossy!) ([#10431](https://github.com/RooCodeInc/Roo-Code/pull/10431)).
* Improves the error shown when `read_file` is run on a directory by explicitly explaining the issue and suggesting `list_files` instead ([#10371](https://github.com/RooCodeInc/Roo-Code/pull/10371)).
* Fixes an issue where Roo’s native `read_file` tool schema didn’t advertise a maximum number of files per call, reducing failed tool calls in environments that enforce low limits ([#10449](https://github.com/RooCodeInc/Roo-Code/pull/10449)).
* Fixes an issue where the "open source" link on roocode.com could point to a non-existent GitHub organization, so it reliably opens the correct repository (thanks jishnuteegala!) ([#10377](https://github.com/RooCodeInc/Roo-Code/pull/10377)).
* Fixes an issue where rate limiting looked like a provider error so a client-side rate limit now shows as a normal “rate limit wait” status row and completes cleanly ([#10389](https://github.com/RooCodeInc/Roo-Code/pull/10389)).
* Fixes an issue where `write_to_file` could create files at an unintended truncated path, reducing accidental file placement and cleanup ([#10415](https://github.com/RooCodeInc/Roo-Code/pull/10415)).
Expand Down
64 changes: 64 additions & 0 deletions docs/update-notes/v3.39.0.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
---
description: Roo Code 3.39.0 adds image file @mentions, makes provider profiles stick to tasks, and updates key UI labels.
keywords:
- roo code 3.39.0
- image @mentions
- providers
- qol improvements
- bug fixes
image: /img/v3.39.0/v3.39.0.png
---

# Roo Code 3.39.0 Release Notes (2026-01-08)

This release adds image file @mentions, makes provider profiles stick to tasks, and updates key UI labels.

<img src="/img/v3.39.0/v3.39.0.png" alt="Roo Code v3.39.0 Release" width="600" />

## Image file @mentions

You can now @mention image files to include them as inline images in your message, making it easier to share screenshots and UI mockups without manually attaching files (thanks hannesrudolph!) ([#10189](https://github.com/RooCodeInc/Roo-Code/pull/10189)).

> **📚 Documentation**: See [Context Mentions](/basic-usage/context-mentions) for usage details.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Repo docs avoid emojis in prose; can you remove the "📚" from this "Documentation" callout (and the other similar ones in this file) to keep style consistent?

Fix it with Roo Code or mention @roomote and request a fix.


## Sticky provider profiles

Tasks now remember the provider profile (API configuration) they started with, so switching profiles elsewhere doesn’t affect running tasks or resumed tasks (thanks hannesrudolph!) ([#10018](https://github.com/RooCodeInc/Roo-Code/pull/10018)).

> **📚 Documentation**: See [API Configuration Profiles](/features/api-configuration-profiles) for details.

## YOLO → BRRRRRRRRRR

The auto-approve mode label has been renamed from “YOLO” to “BRRRRRRRRRR” across the UI (thanks app/roomote!) ([#10507](https://github.com/RooCodeInc/Roo-Code/pull/10507)).

> **📚 Documentation**: See [Auto-Approving Actions](/features/auto-approving-actions) for details.

## QOL Improvements

* The `@` file picker now respects `.rooignore`, reducing noise in large workspaces and helping you avoid accidentally attaching ignored/generated files (thanks app/roomote, jerrill-johnson-bitwerx!) ([#10174](https://github.com/RooCodeInc/Roo-Code/pull/10174))
* Adds debug-only proxy routing settings so you can inspect extension network traffic while running under the VS Code debugger (F5) (thanks hannesrudolph, SleeperSmith!) ([#10467](https://github.com/RooCodeInc/Roo-Code/pull/10467))
* Improves the follow-up suggestion mode badge styling for better readability (thanks mrubens!) ([#9260](https://github.com/RooCodeInc/Roo-Code/pull/9260))
* Clarifies in the native `read_file` tool description that image formats are supported when the model supports vision (thanks app/roomote, nabilfreeman!) ([#10442](https://github.com/RooCodeInc/Roo-Code/pull/10442))

## Bug Fixes

* Fixes an issue where conversations could fail after condensation due to missing/mismatched tool call IDs, improving reliability in longer chats (thanks daniel-lxs!) ([#10471](https://github.com/RooCodeInc/Roo-Code/pull/10471))
* Fixes an issue where duplicate `tool_result` blocks could cause provider API errors (including Anthropic “duplicate toolResult” failures), improving reliability in tool-heavy workflows (thanks daniel-lxs!) ([#10497](https://github.com/RooCodeInc/Roo-Code/pull/10497))
* Fixes an edge case where switching terminals mid-run could produce duplicate tool results and trigger protocol errors, reducing unattended-mode soft-locks (thanks app/roomote, nabilfreeman!) ([#10466](https://github.com/RooCodeInc/Roo-Code/pull/10466))
* Fixes an issue where Roo could generate the wrong command chaining syntax on Windows, making suggested terminal commands more likely to work without edits (thanks app/roomote, AlexNek!) ([#10434](https://github.com/RooCodeInc/Roo-Code/pull/10434))
* Fixes an issue where chat requests could fail on Windows systems without PowerShell in PATH (“spawnSync powershell ENOENT”) (thanks app/roomote, Yang-strive!) ([#9897](https://github.com/RooCodeInc/Roo-Code/pull/9897))
* Fixes a rare edge case where an API rate limit setting could be ignored when provider state is temporarily unavailable (thanks app/roomote!) ([#10266](https://github.com/RooCodeInc/Roo-Code/pull/10266))
* Fixes validation failures in nightly builds by adding missing setting descriptions for debug proxy configuration (thanks app/roomote!) ([#10505](https://github.com/RooCodeInc/Roo-Code/pull/10505))

## Provider Updates

* Adds the `kimi-k2-thinking` model to the Fireworks AI provider (thanks app/roomote, kavehsfv!) ([#9202](https://github.com/RooCodeInc/Roo-Code/pull/9202))
* Adds an “Extra High (xhigh)” reasoning effort option for OpenAI-compatible providers when supported by the endpoint/model (thanks app/roomote, Soorma718!) ([#10061](https://github.com/RooCodeInc/Roo-Code/pull/10061))
* Adds `zai-glm-4.7` to the Cerebras model list (thanks sebastiand-cerebras!) ([#10500](https://github.com/RooCodeInc/Roo-Code/pull/10500))
* Removes legacy Claude 2 models from AWS Bedrock to prevent selecting unsupported options (thanks app/roomote, KevinZhao!) ([#10501](https://github.com/RooCodeInc/Roo-Code/pull/10501))
* Improves compatibility with OpenAI’s Responses API for MCP tool schemas, reducing tool-calling failures when schemas omit `additionalProperties: false` (thanks daniel-lxs!) ([#10472](https://github.com/RooCodeInc/Roo-Code/pull/10472))

## Misc Improvements

* Adds a one-line CLI installer workflow to simplify installing/upgrading the Roo Code CLI (thanks cte!) ([#10474](https://github.com/RooCodeInc/Roo-Code/pull/10474))
* Adds early-stage CLI support used by eval tooling, including a new “CLI” execution option for eval runs (thanks cte!) ([#10452](https://github.com/RooCodeInc/Roo-Code/pull/10452), [#10456](https://github.com/RooCodeInc/Roo-Code/pull/10456))
18 changes: 18 additions & 0 deletions docs/update-notes/v3.39.1.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
---
description: This patch release improves native tool-call reliability and fixes task resume issues for Gemini.
keywords:
- roo code 3.39.1
- bug fixes
image: /img/social-share.jpg
---

# Roo Code 3.39.1 Release Notes (2026-01-08)

This patch release improves native tool-call reliability and fixes task resume issues for Gemini.

## Bug Fixes

* Fixes an issue where file paths shown during native tool-call streaming could appear incorrect or truncated, making it harder to confirm which file Roo is reading or editing ([#10555](https://github.com/RooCodeInc/Roo-Code/pull/10555)).
* Fixes an issue where resuming a task with Gemini models that use extended thinking could fail with a “Corrupted thought signature” / INVALID_ARGUMENT error ([#10554](https://github.com/RooCodeInc/Roo-Code/pull/10554)).
* Fixes an issue where `ask_followup_question` could fail with some Anthropic-backed setups due to strict tool schema validation ([#10551](https://github.com/RooCodeInc/Roo-Code/pull/10551)).

48 changes: 48 additions & 0 deletions docs/update-notes/v3.39.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
---
description: Roo Code 3.39 includes task-level provider profile persistence, improved multimodal workflows, new provider/model options, and reliability fixes across tool calling.
keywords:
- roo code 3.39
- new features
- bug fixes
image: /img/social-share.jpg
---

# Roo Code 3.39 Release Notes (Combined)

Roo Code 3.39 includes task-level provider profile persistence, improved multimodal workflows, new provider/model options, and reliability fixes across tool calling.

## Features

* **Task-level API configuration profiles**: Roo can persist the selected provider profile per task so follow-up sessions keep using the same provider/model settings.
* **Image file @mentions**: You can mention images in your workspace to give the model visual context directly from chat.
* **Headless Roo Code (CLI + VS Code shim)**: Run Roo Code in a headless environment with a basic CLI, plus an installer and an option to use the CLI for evals.
* **Debug proxy routing**: Adds debug-mode proxy routing to help troubleshoot API calls.

## QOL Improvements

* **BRRR naming update**: Renames YOLO to BRRR.
* **Smarter file @mention search**: Filters file @mention results using `.rooignore`.
* **Better tooling docs**: Improves the native `read_file` tool description and examples.
* **Web evals usability**: Remembers your last model selection in web evals and adds an evals skill.

## Bug Fixes

* Fixes PowerShell ENOENT issues on Windows.
* Fixes shell-aware command chaining examples for Windows.
* Fixes tool-result preservation/deduplication issues that could lead to API errors.
* Fixes issues around tool protocol UI and provider configuration details.

## Provider Updates

* Adds new/updated models across multiple providers (e.g., Fireworks AI, OpenAI-compatible endpoints, Cerebras).

---

## Patch Releases

### 3.39.1 (2026-01-08)

* Fixes an issue where file paths shown during native tool-call streaming could appear incorrect or truncated.
* Fixes an issue where resuming a task with Gemini extended thinking could fail with a “Corrupted thought signature” / INVALID_ARGUMENT error.
* Fixes an Anthropic tool schema compatibility issue affecting `ask_followup_question`.

10 changes: 10 additions & 0 deletions sidebars.ts
Original file line number Diff line number Diff line change
Expand Up @@ -172,11 +172,21 @@ const sidebars: SidebarsConfig = {
label: 'Extension Release Notes',
items: [
'update-notes/index',
{
type: 'category',
label: '3.39',
items: [
{ type: 'doc', id: 'update-notes/v3.39', label: '3.39 Combined' },
{ type: 'doc', id: 'update-notes/v3.39.1', label: '3.39.1' },
{ type: 'doc', id: 'update-notes/v3.39.0', label: '3.39.0' },
],
},
{
type: 'category',
label: '3.38',
items: [
{ type: 'doc', id: 'update-notes/v3.38', label: '3.38 Combined' },
{ type: 'doc', id: 'update-notes/v3.38.3', label: '3.38.3' },
{ type: 'doc', id: 'update-notes/v3.38.2', label: '3.38.2' },
{ type: 'doc', id: 'update-notes/v3.38.1', label: '3.38.1' },
{ type: 'doc', id: 'update-notes/v3.38.0', label: '3.38.0' },
Expand Down
Binary file added static/img/v3.39.0/v3.39.0.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.