opensource/kilocode
3
0
Fork 0
mirror of https://github.com/Kilo-Org/kilocode.git synced 2026-09-24 16:02:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update auto docs and add auto balanced and free

Browse Source
This commit is contained in:
Josh Lambert
2026-03-12 08:52:19 -04:00
parent 9a2b6e5cce
commit a115dc0e47
3 changed files with 83 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/kilo-docs/pages/code-with-ai/agents/auto-model.md
+47 -12
View File
@@ -5,19 +5,27 @@ description: "Smart model routing that automatically selects the optimal AI mode
# Auto Model
Auto Model (`kilo-auto/frontier`) is a smart model routing system that automatically selects the optimal AI model based on the Kilo Code mode you're using. It balances cost and capability so you get the best results without manual model switching.
Auto Model is a smart model routing system that automatically selects the optimal AI model based on the Kilo Code mode you're using. It comes in multiple tiers so you can balance cost and capability to fit your needs.
| Tier | Best For | Pricing |
| -------------------- | ------------------------------------------------- | ------- |
| `kilo-auto/frontier` | Maximum capability with the best available models | Paid |
| `kilo-auto/balanced` | Strong performance at a lower cost | Paid |
| `kilo-auto/free` | The best free models available | Free |
## How It Works
1. Select `kilo-auto/frontier` as your model in the model dropdown
1. Select an Auto Model tier (e.g. `kilo-auto/frontier`) in the model dropdown
2. Start working in any mode (Code, Architect, Debug, etc.)
3. The system automatically routes your requests to the best model for that task
That's it. No configuration needed.
## Mode-to-Model Mapping
## Auto Frontier
Auto Model routes to different models based on the task type:
`kilo-auto/frontier` routes to the latest and most capable paid models available, optimizing for performance, capability, and cost.
### Mode-to-Model Mapping
| Mode | Model Used | Best For |
| -------------- | ----------------- | ---------------------------- |
@@ -35,6 +43,32 @@ Auto Model routes to different models based on the task type:
**Implementation tasks** use Claude Sonnet 4.6, which is optimized for fast, accurate code generation and editing.
## Auto Balanced
`kilo-auto/balanced` follows the same mode-based routing structure as Frontier but uses more cost-effective models — Kimi K2.5 for reasoning-heavy modes and Minimax M2.5 (Free) for implementation modes.
### Mode-to-Model Mapping
| Mode | Model Used | Best For |
| -------------- | ------------------- | ---------------------------- |
| `architect` | Kimi K2.5 | System design, planning |
| `orchestrator` | Kimi K2.5 | Multi-step task coordination |
| `ask` | Kimi K2.5 | Questions, explanations |
| `plan` | Kimi K2.5 | Planning, reasoning |
| `general` | Kimi K2.5 | General assistance |
| `debug` | Kimi K2.5 | Debugging and fixing issues |
| `code` | Minimax M2.5 (Free) | Writing and editing code |
| `build` | Minimax M2.5 (Free) | Implementation tasks |
| `explore` | Minimax M2.5 (Free) | Codebase exploration |
**Planning and reasoning tasks** use Kimi K2.5, a strong open-weight reasoning model from Moonshot AI.
**Implementation tasks** use Minimax M2.5, which provides fast, capable code generation at a fraction of frontier model costs.
{% callout type="info" title="Image support" %}
Auto Balanced does not support image inputs, since Minimax M2.5 does not have vision capabilities.
{% /callout %}
## Benefits
### Simplified Setup
@@ -43,30 +77,31 @@ No need to manually switch models when changing modes. Auto Model handles the ro
### Cost Optimization
Uses the more economical Sonnet for implementation tasks where speed matters, while reserving Opus for planning tasks that benefit from deeper reasoning. You get optimal cost-to-capability ratio without thinking about it.
Uses the more economical models for implementation tasks where speed matters, while reserving stronger reasoning models for planning tasks. You get optimal cost-to-capability ratio without thinking about it.
### Best-in-Class Models
Auto Model always routes to Claude's latest and most capable models:
Auto Model routes to capable models matched to your task:
- **Claude Opus** for reasoning-intensive tasks
- **Claude Sonnet** for implementation-focused tasks
- **Auto Frontier** uses the latest and most effective models across all modes
- **Auto Balanced** uses more cost-effective models while still providing strong capabilities
- **Auto Free** uses the best available free models
## Requirements
{% callout type="warning" title="Version Requirements" %}
Auto Model requires **VS Code/JetBrains extension v5.2.3+** or **CLI v1.0.15+** for automatic mode-based switching. On older versions, `kilo-auto/frontier` will default to Claude Sonnet for all requests.
Auto Model requires **VS Code/JetBrains extension v5.2.3+** or **CLI v1.0.15+** for automatic mode-based switching. On older versions, Auto Model tiers will default to a single model for all requests.
{% /callout %}
## Getting Started
{% callout type="tip" title="Quick Setup" %}
Select `kilo-auto/frontier` from the model dropdown in the Kilo Code chat interface. That's all you need to do.
Select an Auto Model tier from the model dropdown in the Kilo Code chat interface. That's all you need to do.
{% /callout %}
1. Open Kilo Code in VS Code or JetBrains
2. Click the model selector dropdown
3. Choose `kilo-auto/frontier`
3. Choose an Auto Model such as `kilo-auto/frontier` or `kilo-auto/balanced`
4. Start chatting - the right model is selected automatically based on your current mode
## When to Use Auto Model
@@ -90,7 +125,7 @@ You may want to select a specific model instead when:
## Feedback
{% callout type="note" title="Help Us Improve" %}
Auto Model is a new feature and we're actively improving it. We'd love to hear how it's working for you! Share feedback in our [Discord](https://kilo.ai/discord) or [open an issue on GitHub](https://github.com/Kilo-Org/kilocode/issues).
Auto Model is actively being improved. We'd love to hear how it's working for you! Share feedback in our [Discord](https://kilo.ai/discord) or [open an issue on GitHub](https://github.com/Kilo-Org/kilocode/issues).
{% /callout %}
## Related
packages/kilo-docs/pages/gateway/authentication.md
+8 -8
View File
@@ -111,11 +111,11 @@ This provides the benefit of Vercel's reliability infrastructure while using you
The gateway accepts the following headers:
| Header | Required | Description |
| --------------------------- | ----------------------- | ------------------------------------------------ |
| `Authorization` | Yes (unless free model) | `Bearer <api_key>` |
| `Content-Type` | Yes | `application/json` |
| `X-KiloCode-OrganizationId` | No | Organization context for org-scoped requests |
| `X-KiloCode-TaskId` | No | Task identifier for prompt cache keying |
| `X-KiloCode-Version` | No | Client version string |
| `x-kilocode-mode` | No | Mode hint for `kilo-auto/frontier` model routing |
| Header | Required | Description |
| --------------------------- | ----------------------- | -------------------------------------------- |
| `Authorization` | Yes (unless free model) | `Bearer <api_key>` |
| `Content-Type` | Yes | `application/json` |
| `X-KiloCode-OrganizationId` | No | Organization context for org-scoped requests |
| `X-KiloCode-TaskId` | No | Task identifier for prompt cache keying |
| `X-KiloCode-Version` | No | Client version string |
| `x-kilocode-mode` | No | Mode hint for `kilo-auto` model routing |
packages/kilo-docs/pages/gateway/models-and-providers.md
+28 -3
View File
@@ -64,9 +64,13 @@ Several models are available at no cost, subject to rate limits:
Free models are available to both authenticated and anonymous users. Anonymous users are rate-limited to 200 requests per hour per IP address.
## The `kilo-auto/frontier` model
## Auto models
The `kilo-auto/frontier` virtual model automatically selects the best model based on the task type. The selection is controlled by the `x-kilocode-mode` request header:
Kilo Auto virtual models automatically select the best underlying model based on the task type. The selection is controlled by the `x-kilocode-mode` request header.
### `kilo-auto/frontier`
Routes to the most capable paid models optimizing for cost, performance, and capabilities.
| Mode | Resolved Model |
| -------------------------------------------------------------- | ----------------------------- |
@@ -74,6 +78,27 @@ The `kilo-auto/frontier` virtual model automatically selects the best model base
| `build`, `explore`, `code` | `anthropic/claude-sonnet-4.6` |
| Default (no mode specified) | `anthropic/claude-sonnet-4.6` |
### `kilo-auto/balanced`
Follows the same mode-based routing as Frontier but uses more cost-effective models.
| Mode | Resolved Model |
| -------------------------------------------------------------- | --------------------------- |
| `plan`, `general`, `architect`, `orchestrator`, `ask`, `debug` | `moonshotai/kimi-k2.5` |
| `build`, `explore`, `code` | `minimax/minimax-m2.5:free` |
| Default (no mode specified) | `minimax/minimax-m2.5:free` |
### `kilo-auto/free`
The best available free model for each mode.
| Mode | Resolved Model |
| --------------------------- | --------------------------- |
| All modes | `minimax/minimax-m2.5:free` |
| Default (no mode specified) | `minimax/minimax-m2.5:free` |
### Example usage
```json
{
"model": "kilo-auto/frontier",
@@ -88,7 +113,7 @@ curl -X POST "https://api.kilo.ai/api/gateway/chat/completions" \
-H "Authorization: Bearer $KILO_API_KEY" \
-H "x-kilocode-mode: plan" \
-H "Content-Type: application/json" \
-d '{"model": "kilo-auto/frontier", "messages": [{"role": "user", "content": "Design a database schema"}]}'
-d '{"model": "kilo-auto/balanced", "messages": [{"role": "user", "content": "Design a database schema"}]}'
```
## Providers