You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: docs/features/custom-modes.mdx
+4-2Lines changed: 4 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -37,6 +37,10 @@ Roo Code allows you to create **custom modes** to tailor Roo's behavior to speci
37
37
Each mode—including custom ones—features **Sticky Models**. This means Roo Code automatically remembers and selects the last model you used with a particular mode. This lets you assign different preferred models to different tasks without constant reconfiguration, as Roo switches between models when you change modes.
38
38
:::
39
39
40
+
:::info Discover Community Modes
41
+
Looking for ready-to-use custom modes? Visit the [Roo Code Marketplace](/features/marketplace) to browse and install community-contributed modes with a single click. The marketplace offers specialized modes for various tasks like React development, documentation writing, testing, and more—all created and shared by the Roo Code community.
42
+
:::
43
+
40
44
---
41
45
42
46
## Why Use Custom Modes?
@@ -661,5 +665,3 @@ In addition to mode-specific rules directories, there's also a generic `.roo/rul
661
665
-**Invalid regex patterns:** If your `fileRegex` pattern is invalid, you'll receive an error message. Test your patterns using online regex testers before applying them.
662
666
-**Precedence confusion:** Remember that project modes completely override global modes with the same slug - no properties are merged.
663
667
664
-
## Community Resources
665
-
Ready to explore more? Check out the [Community Projects](/community) page to discover tools and projects that extend Roo Code's capabilities, including sophisticated custom mode collections like [Roo Commander](/community/roo-commander) and [Maestro](/community/maestro).
Copy file name to clipboardExpand all lines: docs/features/enhance-prompt.md
+90-13Lines changed: 90 additions & 13 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -13,7 +13,7 @@ image: /img/social-share.jpg
13
13
14
14
# Enhance Prompt
15
15
16
-
The "Enhance Prompt" feature in Roo Code helps you improve the quality and effectiveness of your prompts before sending them to the AI model. By clicking the <Codiconname="wand" /> icon in the chat input, you can automatically refine your initial request, making it clearer, more specific, and more likely to produce the desired results.
16
+
The "Enhance Prompt" feature in Roo Code helps you improve the quality and effectiveness of your prompts before sending them to the AI model. By clicking the wand icon in the chat input, you can automatically refine your initial request, making it clearer, more specific, and more likely to produce the desired results.
17
17
18
18
---
19
19
@@ -24,38 +24,115 @@ The "Enhance Prompt" feature in Roo Code helps you improve the quality and effec
24
24
***Better Instructions:** Roo Code can add instructions to guide the AI towards a more helpful response (e.g., requesting specific formatting or a particular level of detail).
25
25
***Reduced Ambiguity:** Enhance Prompt helps to eliminate ambiguity and ensure that Roo Code understands your intent.
26
26
***Consistency**: Roo will consistently format prompts the same way to the AI.
27
+
***Context-Aware Suggestions:** When enabled, uses your recent conversation history to generate more relevant and accurate enhancements.
27
28
28
29
---
29
30
30
31
## How to Use Enhance Prompt
31
32
32
33
1.**Type your initial prompt:** Enter your request in the Roo Code chat input box as you normally would. This can be a simple question, a complex task description, or anything in between.
33
-
2.**Click the <Codiconname="wand" /> Icon:** Instead of pressing Enter, click the <Codiconname="wand" /> icon located in the bottom right of the chat input box.
34
-
3.**Review the Enhanced Prompt:** Roo Code will replace your original prompt with an enhanced version. Review the enhanced prompt to make sure it accurately reflects your intent. You can further refine the enhanced prompt before sending.
34
+
2.**Click the Wand Icon:** Instead of pressing Enter, click the wandicon located in the top right corner of the chat input box. While Roo processes your enhancement request, the wand icon will spin to indicate it's working.
35
+
3.**Review the Enhanced Prompt:** Roo Code will replace your original prompt with an enhanced version. Review the enhanced prompt to make sure it accurately reflects your intent. You can further refine the enhanced prompt before sending. Changed your mind? You can undo the enhancement using Ctrl+Z (Cmd+Z on Mac) to restore your original prompt.
35
36
4.**Send the Enhanced Prompt:** Press Enter or click the Send icon (<Codiconname="send" />) to send the enhanced prompt to Roo Code.
36
37
37
38
---
38
39
40
+
## Special Behaviors
41
+
42
+
### Empty Prompt Enhancement
43
+
44
+
If you click the enhance button with an empty prompt, Roo will show you a helpful message explaining how the feature works. This is a great way to learn about the enhancement feature if you're new to it.
45
+
46
+
### Message Queueing Support
47
+
48
+
The enhance button remains enabled even when message sending is disabled. This allows you to enhance prompts that will be queued for later sending.
49
+
50
+
---
51
+
39
52
## Customizing the Enhancement Process
40
53
41
-
The "Enhance Prompt" feature uses a customizable prompt template. You can modify this template to tailor the enhancement process to your specific needs.
54
+
The "Enhance Prompt" feature uses a customizable prompt template. You can modify this template to tailor the enhancement process to your specific needs.
55
+
56
+
### Accessing Prompts Settings
42
57
43
-
1.**Open the Prompts Tab:** Click the <Codiconname="notebook" /> icon in the Roo Code top menu bar.
44
-
2.**Select "ENHANCE" Tab:**You should see listed out support prompts, including "ENHANCE". Click on this tab.
45
-
3.**Edit the Prompt Template:**Modify the text in the "Prompt" field.
58
+
1.**Open Settings:** Click the gear icon (<Codiconname="gear" />) in the Roo Code panel or use the settings command.
59
+
2.**Navigate to Prompts:**Go to the "Prompts" tab in the settings.
60
+
3.**Select "ENHANCE":**From the dropdown menu, select "ENHANCE" to view and edit the enhancement prompt.
46
61
47
-
The default prompt template includes the placeholder `${userInput}`, which will be replaced with your original prompt. You can modify this to fit the model's prompt format, and instruct it how to enhance your request.
62
+
### Editing the Enhancement Prompt
63
+
64
+
The default enhancement prompt template is:
65
+
66
+
```
67
+
Generate an enhanced version of this prompt (reply with only the enhanced prompt - no conversation, explanations, lead-in, bullet points, placeholders, or surrounding quotes):
68
+
69
+
${userInput}
70
+
```
71
+
72
+
The `${userInput}` placeholder will be replaced with your original prompt. You can modify this template to fit your needs and the model's prompt format.
73
+
74
+
### Testing Your Custom Prompt
75
+
76
+
The Prompts settings include a test area where you can preview how your custom enhancement prompt works:
77
+
78
+
1. After editing your enhancement prompt, look for the "Test Enhancement" section
79
+
2. Enter a sample prompt to test
80
+
3. Click "Test" to see how your custom prompt would enhance it
81
+
4. Adjust your enhancement prompt as needed based on the results
48
82
49
83
---
50
84
51
85
## API Configuration
52
86
53
-
The API configuration used for Enhance Prompt is, by default, the same one that is selected for Roo Code tasks,
54
-
but it can be changed:
87
+
The API configuration used for Enhance Prompt is, by default, the same one that is selected for Roo Code tasks, but it can be changed:
88
+
89
+
1.**Open Settings:** Navigate to Roo Code settings
90
+
2.**Go to Prompts Tab:** Select the "Prompts" tab
91
+
3.**Select "ENHANCE":** Choose "ENHANCE" from the dropdown
92
+
4.**Configure API:** You'll see an "API Configuration" dropdown where you can choose an existing configuration. Future Enhance Prompt requests will be sent to that configured provider/model.
93
+
94
+
---
95
+
96
+
## Context-Aware Enhancement
97
+
98
+
The Enhance Prompt feature can now use your conversation history to generate more relevant suggestions. This helps reduce hallucinations and provides more accurate enhancements based on what you've been working on.
99
+
100
+
### How It Works
101
+
102
+
When enabled, the enhancement process includes your last 10 messages from the current conversation as context. This allows the AI to:
103
+
- Understand what you've been working on
104
+
- Maintain consistency with previous discussions
105
+
- Avoid suggesting unrelated or incorrect enhancements
106
+
- Provide more targeted and useful prompt improvements
107
+
108
+
### Enabling Task History Context
109
+
110
+
To enable or disable the use of conversation history in prompt enhancement:
111
+
112
+
1.**Open Settings:** Navigate to Roo Code settings
113
+
2.**Go to Prompts Tab:** Select the "Prompts" tab
114
+
3.**Select "ENHANCE":** Choose "ENHANCE" from the dropdown
115
+
4.**Toggle History Option:** Check or uncheck "Include task history in enhancement" for better context
116
+
117
+
When disabled, the enhancement will only consider your current prompt without any conversation context.
118
+
119
+
---
120
+
121
+
## Visual Feedback and UI Elements
122
+
123
+
### Button Appearance
124
+
- The wand icon appears semi-transparent (60% opacity) by default
125
+
- Becomes fully opaque (100%) when you hover over it
126
+
- Located in the top-right corner of the chat input box
127
+
- Has a focus ring for keyboard accessibility
128
+
129
+
### Loading State
130
+
- While processing your enhancement request, the wand icon spins
131
+
- This provides clear visual feedback that Roo is working on your prompt
55
132
56
-
1.**Open the Prompts Tab:** Click the <Codiconname="notebook" /> icon in the Roo Code top menu bar.
57
-
2.**Select "ENHANCE" Tab:** You should see an "API Configuration" dropdown
58
-
3.**Select an API Configuration:** Choose an existing configuration, and future Enhance Prompt requests will be sent to that configured provider/model.
133
+
### Tooltip
134
+
- Hovering over the button shows: "Enhance prompt with additional context"
-**Description**: Controls the maximum number of tokens Claude can generate in a single response.
88
+
-**When to change**: If you need longer responses or want to limit output length for cost/performance reasons.
89
+
90
+
**Example configuration:**
91
+
```bash
92
+
export CLAUDE_CODE_MAX_OUTPUT_TOKENS=32768 # Set to 32k tokens
93
+
```
94
+
86
95
---
87
96
88
97
## Supported Models
@@ -100,6 +109,17 @@ The specific models available depend on your Claude CLI subscription and plan.
100
109
101
110
---
102
111
112
+
## Output Token Limits
113
+
114
+
The Claude Code provider now defaults to 16,384 (16k) max output tokens, allowing for longer and more complete responses. This is particularly useful for:
115
+
- Generating large code files
116
+
- Detailed explanations and documentation
117
+
- Complex refactoring operations
118
+
- Multi-file changes
119
+
120
+
You can customize this limit using the `CLAUDE_CODE_MAX_OUTPUT_TOKENS` environment variable if you need different limits for your use case.
description: Configure ByteDance's Doubao AI models in Roo Code. Access competitive language models with full integration and internationalized support.
4
+
keywords:
5
+
- doubao
6
+
- bytedance
7
+
- bytedance ai
8
+
- roo code
9
+
- api provider
10
+
- doubao models
11
+
- chinese ai
12
+
- language models
13
+
image: /img/social-share.jpg
14
+
---
15
+
16
+
# Using Doubao With Roo Code
17
+
18
+
Doubao is ByteDance's Chinese AI service, offering competitive language models for various development tasks. The provider includes full API integration with embedding support and internationalized prompts.
description: Configure SambaNova's high-speed AI models in Roo Code. Experience enterprise-grade inference with competitive performance and diverse model selection.
4
+
keywords:
5
+
- sambanova
6
+
- sambanova ai
7
+
- roo code
8
+
- api provider
9
+
- high-speed inference
10
+
- enterprise ai
11
+
- llm provider
12
+
- fast inference
13
+
image: /img/social-share.jpg
14
+
---
15
+
16
+
# Using SambaNova With Roo Code
17
+
18
+
SambaNova specializes in providing high-speed inference for large language models, utilizing their Reconfigurable Dataflow Units (RDUs) through their SambaCloud portal. This delivers fast response times for supported models.
To use SambaNova with Roo Code, you'll need an API key from the [SambaCloud](https://cloud.sambanova.ai?utm_source=roocode&utm_medium=external&utm_campaign=cloud_signup). After signing up, navigate to the API Keys section in the left panel to create and copy your SambaCloud API key.
27
+
28
+
---
29
+
30
+
## Supported Models
31
+
32
+
Roo Code will attempt to fetch the list of available models from the SambaNova API. Common models available via SambaCloud include:
33
+
34
+
*`DeepSeek-R1`
35
+
*`DeepSeek-V3-0324`
36
+
*`DeepSeek-R1-Distill-Llama-70B`
37
+
*`Meta-Llama-3.3-70B-Instruct`
38
+
*`Meta-Llama-3.1-8B-Instruct`
39
+
*`Llama-4-Maverick-17B-128E-Instruct`
40
+
*`Qwen3-32B`
41
+
*`Llama-3.3-Swallow-70B-Instruct-v0.4`
42
+
43
+
Refer to the [SambaCloud Documentation](https://docs.sambanova.ai/cloud/docs/get-started/supported-models) for the most up-to-date list of supported models and their capabilities.
44
+
45
+
---
46
+
47
+
## Configuration in Roo Code
48
+
49
+
1.**Open Roo Code Settings:** Click the gear icon (<Codiconname="gear" />) in the Roo Code panel.
50
+
2.**Select Provider:** Choose "SambaNova" from the "API Provider" dropdown.
51
+
3.**Enter API Key:** Paste your SambaNova API key into the "SambaNova API Key" field.
52
+
4.**Select Model:** Choose your desired model from the "Model" dropdown.
53
+
5.**(Optional) Custom Base URL:** If using a private deployment, check "Use custom base URL" and enter your endpoint URL.
0 commit comments