opensource/coder
3
0
Fork 0
mirror of https://github.com/coder/coder.git synced 2026-09-24 15:04:27 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: gate chat advisor and virtual desktop behind experiments, delete experiments page (#26809)

Browse Source
This commit is contained in:
Danielle Maywood
2026-06-30 21:57:40 +01:00
committed by GitHub
parent dcb120d6ab
commit 6b8c38b5a4
43 changed files with 907 additions and 2110 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/ai-coder/agents/architecture.md
+8 -8
View File
@@ -173,14 +173,14 @@ provider-native, and computer-use tools are not available.
These tools manage sub-agents — child chats that work on independent tasks in
parallel.
| Tool | What it does |
|---------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `spawn_agent` (`type=general` or `explore`) | Delegates a task to a sub-agent with its own context window. |
| `wait_agent` | Waits for a sub-agent to finish and collects its result. |
| `message_agent` | Sends a follow-up message to a running sub-agent. |
| `interrupt_agent` | Halts a sub-agent's current turn; it transitions to waiting or running if there are queued messages. |
| `spawn_agent` (`type=computer_use`) | Spawns a sub-agent with desktop interaction capabilities (screenshot, mouse, keyboard). Requires an administrator-configured computer-use provider (Anthropic or OpenAI) and the [virtual desktop experiment](./platform-controls/experiments.md#virtual-desktop) to be enabled. |
| `list_agents` | Lists spawned child agents, most recently active first. |
| Tool | What it does |
|---------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `spawn_agent` (`type=general` or `explore`) | Delegates a task to a sub-agent with its own context window. |
| `wait_agent` | Waits for a sub-agent to finish and collects its result. |
| `message_agent` | Sends a follow-up message to a running sub-agent. |
| `interrupt_agent` | Halts a sub-agent's current turn; it transitions to waiting or running if there are queued messages. |
| `spawn_agent` (`type=computer_use`) | Spawns a sub-agent with desktop interaction capabilities (screenshot, mouse, keyboard). Requires an administrator-configured computer-use provider (Anthropic or OpenAI) and the [virtual desktop experiment](./platform-controls/virtual-desktop.md) to be enabled. |
| `list_agents` | Lists spawned child agents, most recently active first. |
### Provider tools
docs/ai-coder/agents/platform-controls/advisor.md
+48
View File
@@ -0,0 +1,48 @@
# Advisor
> [!NOTE]
> This feature is experimental. Pin a release before broad rollout and review
> the release notes before upgrading.
## Enable the experiment
```shell
coder server --experiments=chat-advisor
```
Or set the environment variable:
```shell
CODER_EXPERIMENTS=chat-advisor
```
## What it does
Lets a root agent pause its current turn and request strategic guidance from
a separate, single-step model call. The advisor sees recent conversation
context, runs without any tools, and returns concise advice for the root
agent rather than the end user. While active, it is the only tool the root
agent can call for that turn.
Useful for planning ambiguity, architectural tradeoffs, debugging strategy
after repeated failures, or risk reduction before a destructive operation.
## Configuration
Once the experiment is enabled, configure the advisor under **AI Settings** >
**Coder Agents** > **Advisor**.
| Field | Default | Notes |
|-------------------|----------------------|-------------------------------------------------------------------------------------------------------------------------|
| Max uses per run | `0` (unlimited) | Caps how many times the root agent can call the advisor in a single chat run. Must be a non-negative integer. |
| Max output tokens | `0` (server default) | Caps the advisor model's response length. `0` uses the server default of 16,384 tokens. Must be a non-negative integer. |
| Advisor model | Use chat model | Optional dedicated chat model config for the advisor. When unset, the advisor reuses the root agent's model. |
The advisor is not available in plan mode or to subagents. Failed advisor
invocations refund the per-run budget, and advisor calls are not metered
against the root chat's usage limit.
The same configuration is available at:
- `GET /api/experimental/chats/config/advisor`
- `PUT /api/experimental/chats/config/advisor`
docs/ai-coder/agents/platform-controls/experiments.md → docs/ai-coder/agents/platform-controls/chat-debug-logging.md
+3 -59
View File
@@ -1,60 +1,4 @@
# Experiments
The **Experiments** tab under **Agents** > **Settings** > **Manage Agents**
is where administrators opt in to features that are still iterating. The
behavior, configuration surface, and APIs documented here may change between
releases without notice.
> [!NOTE]
> Everything in this page is experimental. Pin a release before broad rollout
> and review the release notes before upgrading.
## Virtual desktop
Lets agents drive a graphical desktop inside the workspace through
`spawn_agent` with `type=computer_use` (screenshots, mouse, keyboard).
To enable, toggle **Virtual Desktop** on, then choose a **Computer use
provider** (Anthropic or OpenAI). It also requires:
- The [portabledesktop](https://registry.coder.com/modules/coder/portabledesktop)
module installed in the workspace template.
- An API key for the selected provider configured under the **Providers**
tab.
The Anthropic and OpenAI computer-use models are fixed by Coder per provider
and are not selectable from this UI. Anthropic is the default when no
provider is set.
## Advisor
Lets a root agent pause its current turn and request strategic guidance from
a separate, single-step model call. The advisor sees recent conversation
context, runs without any tools, and returns concise advice for the parent
agent rather than the end user. While active, it is the only tool the parent
can call for that turn.
Useful for planning ambiguity, architectural tradeoffs, debugging strategy
after repeated failures, or risk reduction before a destructive operation.
| Field | Default | Notes |
|-------------------|----------------------|-------------------------------------------------------------------------------------------------------------------------|
| Advisor (toggle) | Off | Master switch. When off, the advisor tool is not attached to new chats. |
| Max uses per run | `0` (unlimited) | Caps how many times an agent can call the advisor in a single chat run. Must be a non-negative integer. |
| Max output tokens | `0` (server default) | Caps the advisor model's response length. `0` uses the server default of 16,384 tokens. Must be a non-negative integer. |
| Reasoning effort | Use chat model | One of unset, `low`, `medium`, or `high`. Unset delegates to the underlying model's default. |
| Advisor model | Use chat model | Optional dedicated chat model config for the advisor. When unset, the advisor reuses the parent chat's model. |
The advisor is not available in plan mode or to subagents. Failed advisor
invocations refund the per-run budget, and advisor calls are not metered
against the parent chat's usage limit.
The same configuration is available at:
- `GET /api/experimental/chats/config/advisor`
- `PUT /api/experimental/chats/config/advisor`
## Chat debug logging
# Chat debug logging
Records a detailed trace of each chat turn for troubleshooting: the
normalized request sent to the LLM provider, the full response, token usage,
@@ -67,7 +11,7 @@ Off by default. Three layers control whether it runs for a given chat:
on for every chat. The runtime admin and user toggles become read-only.
1. **Runtime admin gate.** With the deployment override unset, the
*Let users record chat debug logs* toggle decides whether users can opt
in. Configure it at
in. Configure it under **AI Settings** > **Lifecycle**, or at
`GET/PUT /api/experimental/chats/config/debug-logging`.
1. **Per-user toggle.** Users with the admin gate enabled can turn debug
logging on for their own chats from **Agents** > **Settings** > **General**
@@ -89,7 +33,7 @@ that chat's owner. The tab lists recent debug runs and lets you expand a run
into its per-step request, response, token usage, retry attempts, errors,
and policy metadata.
### Export debug logs
## Export debug logs
You can export the same captured debug data from the UI:
docs/ai-coder/agents/platform-controls/index.md
+10 -5
View File
@@ -144,12 +144,17 @@ for details.
### Experiments
Administrators can opt in to experimental features under **Agents** >
**Settings** > **Manage Agents** > **Experiments**. Behavior, configuration
surface, and APIs may change between releases.
Administrators enable experimental features using the `--experiments` flag on
`coder server` (or the `CODER_EXPERIMENTS` environment variable). Once enabled,
runtime configuration for those features is available under **AI Settings** >
**Coder Agents**.
See [Experiments](./experiments.md) for the current list of experiments, how
to enable them, and the relevant API endpoints.
See the following pages for experiment-gated features:
- [Advisor](./advisor.md) (`--experiments=chat-advisor`)
- [Virtual desktop](./virtual-desktop.md) (`--experiments=chat-virtual-desktop`)
For chat debug logging (not experiment-gated), see [Chat debug logging](./chat-debug-logging.md).
## Where we are headed
docs/ai-coder/agents/platform-controls/virtual-desktop.md
+43
View File
@@ -0,0 +1,43 @@
# Virtual desktop
> [!NOTE]
> This feature is experimental. Pin a release before broad rollout and review
> the release notes before upgrading.
## Enable the experiment
```shell
coder server --experiments=chat-virtual-desktop
```
Or set the environment variable:
```shell
CODER_EXPERIMENTS=chat-virtual-desktop
```
## What it does
Lets agents drive a graphical desktop inside the workspace through
`spawn_agent` with `type=computer_use` (screenshots, mouse, keyboard).
## Configuration
Once the experiment is enabled, configure the computer-use provider under
**AI Settings** > **Coder Agents** > **Virtual desktop**.
Choose a **Computer use provider** (Anthropic or OpenAI). Virtual desktop also requires:
- The [portabledesktop](https://registry.coder.com/modules/coder/portabledesktop)
module installed in the workspace template.
- An API key for the selected provider configured under the **Providers**
tab.
The Anthropic and OpenAI computer-use models are fixed by Coder per provider
and are not selectable from this UI. Anthropic is the default when no
provider is set.
The same configuration is available at:
- `GET /api/experimental/chats/config/computer-use-provider`
- `PUT /api/experimental/chats/config/computer-use-provider`
docs/manifest.json
+16 -4
View File
@@ -1080,10 +1080,22 @@
"state": ["beta"]
},
{
"title": "Experiments",
"description": "Experimental Coder Agents features admins can opt in to: virtual desktop, advisor, and chat debug logging",
"path": "./ai-coder/agents/platform-controls/experiments.md",
"state": ["beta"]
"title": "Advisor",
"description": "Strategic guidance for root agent chats",
"path": "./ai-coder/agents/platform-controls/advisor.md",
"state": ["early access"]
},
{
"title": "Virtual desktop",
"description": "Allow agents to use a graphical desktop within workspaces",
"path": "./ai-coder/agents/platform-controls/virtual-desktop.md",
"state": ["early access"]
},
{
"title": "Chat debug logging",
"description": "Record detailed traces of chat turns for troubleshooting",
"path": "./ai-coder/agents/platform-controls/chat-debug-logging.md",
"state": ["early access"]
}
]
},
docs/reference/api/schemas.md
Generated
+3 -3
View File
@@ -7087,9 +7087,9 @@ CreateWorkspaceRequest provides options for creating a new workspace. Only one o
#### Enumerated Values
| Value(s) |
|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `agent-app-tabs`, `ai-gateway-cost-control`, `auto-fill-parameters`, `example`, `mcp-server-http`, `minimum-implicit-member`, `nats_pubsub`, `notifications`, `oauth2`, `workspace-build-updates`, `workspace-usage` |
| Value(s) |
|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `agent-app-tabs`, `ai-gateway-cost-control`, `auto-fill-parameters`, `chat-advisor`, `chat-virtual-desktop`, `example`, `mcp-server-http`, `minimum-implicit-member`, `nats_pubsub`, `notifications`, `oauth2`, `workspace-build-updates`, `workspace-usage` |
## codersdk.ExternalAPIKeyScopes