[Error] chatagenterror-Auto mode failed: no available model found in known endpoints.

[vs-code-engineering]

Error Message

chatagenterror-Auto mode failed: no available model found in known endpoints.

Stack Trace

    at Cte._selectDefaultModel (../src/platform/endpoint/node/automodeService.ts:374:9)
    at Cte.resolveAutoModeEndpoint (../src/platform/endpoint/node/automodeService.ts:217:24)
    at async $4e.switchToBaseModel (../src/extension/conversation/vscode-node/chatParticipants.ts:302:19)
    at async nw._requestHandler (../src/extension/conversation/vscode-node/chatParticipants.ts:210:14)
    at async i.$invokeAgent (./src/vs/workbench/api/common/extHostChatAgents2.ts:990:10)

Raw Stack Trace (minified)

Error: Auto mode failed: no available model found in known endpoints. 	at Cte._selectDefaultModel (\extensions\copilot\dist\extension.js:1661:6998) 	at Cte.resolveAutoModeEndpoint (\extensions\copilot\dist\extension.js:1661:3280) 	at async $4e.switchToBaseModel (\extensions\copilot\dist\extension.js:6778:3160) 	at async nw._requestHandler (\extensions\copilot\dist\extension.js:6778:1099) 	at async i.$invokeAgent (out/vs/workbench/api/node/extensionHostProcess.js:529:20483)

Details

Property	Value
Bucket ID	3449f980-e74a-85fc-72a6-586f52ed8f7a
Version	1.117.0
Commit	10c8e557
Last Seen	Wed Apr 22 2026 19:31:10 GMT+0000 (Coordinated Universal Time)
Platforms	Mac, Windows, Linux
Product	VSCode

---

This issue was automatically created from the VS Code Errors Dashboard

[vs-code-engineering]

Stable Release Anomaly Details

• Anomaly type: new
• Previous stable release: 1.116.0 — not present
• Current stable release: 1.117.0 — 4,869 users (16,098 hits)

Analysis

This error originates from the Copilot extension (extensions/copilot/dist/extension.js). The stack trace shows:

1. _selectDefaultModel fails in auto mode — no available model found
2. Called from resolveAutoModeEndpoint → switchToBaseModel → _requestHandler
3. Surfaces as a chatagenterror during chat agent invocation

This is a new error in v1.117.0 suggesting that the Copilot extension's auto-mode model resolution has endpoint availability issues. With 4,869 users affected across all platforms, this is a high-impact Copilot functionality regression. The error stopped appearing after Apr 22, suggesting a possible server-side fix.

Generated by errors-stable-scan · ● 9.6M · ◷

[vs-code-engineering]

🔍 Error Investigation

Summary

The error Auto mode failed: no available model found in known endpoints originates entirely within the GitHub Copilot extension (extensions/copilot/dist/extension.js), not VS Code core. When a user sends a chat message with "Auto mode" enabled, the extension's automodeService attempts to resolve an endpoint and select a default model. If no model is found across all known endpoints, it throws this error. The error propagates back through the VS Code extension host, which logs it as chatAgentError telemetry. With 17,110 hits across 5,127 users (predominantly on stable 1.117.0 commit 10c8e557), this has significant user impact.

Culprit Commit

No culprit commit in VS Code core was identified. The error originates in the Copilot extension's internal auto mode service (automodeService.ts), not in VS Code source code. No commit range was provided by the regression scan workflow, and no VS Code core code changes are involved in the error path. The regression likely corresponds to a Copilot extension update or a backend service change affecting model endpoint availability.

Code Flow

sequenceDiagram
    participant User as User Chat Input
    participant Main as mainThreadChatAgents2.ts
    participant ExtHost as extHostChatAgents2.ts
    participant CopilotHandler as Copilot: chatParticipants.ts
    participant AutoMode as Copilot: automodeService.ts

    User->>Main: Send chat message
    Main->>ExtHost: $invokeAgent(handle, request)
    ExtHost->>ExtHost: getModelForRequest() → OK
    ExtHost->>CopilotHandler: agent.invoke(request, context)
    CopilotHandler->>CopilotHandler: _requestHandler() [line 210]
    CopilotHandler->>CopilotHandler: switchToBaseModel() [line 302]
    CopilotHandler->>AutoMode: resolveAutoModeEndpoint() [line 217]
    AutoMode->>AutoMode: _selectDefaultModel() [line 374]
    Note over AutoMode: ⚠️ Root cause:<br/>No models found in<br/>known endpoints list
    AutoMode-->>ExtHost: 💥 Error: "Auto mode failed:<br/>no available model found<br/>in known endpoints."
    ExtHost->>ExtHost: catch(e) → packErrorForTelemetry
    ExtHost-->>Main: { errorDetails, errorCallstack }
    Main->>Main: publicLogError2('chatAgentError', ...)

Loading

Affected Files

File	Role
extensions/copilot/.../automodeService.ts:374	💥 Crash site — _selectDefaultModel() throws when no model available
extensions/copilot/.../automodeService.ts:217	Data flow — resolveAutoModeEndpoint() calls model selection
extensions/copilot/.../chatParticipants.ts:302	Data flow — switchToBaseModel() triggers auto mode resolution
extensions/copilot/.../chatParticipants.ts:210	Entry point — _requestHandler() for chat requests
src/vs/workbench/api/common/extHostChatAgents2.ts:991	VS Code dispatch — agent.invoke() calls extension handler
src/vs/workbench/api/browser/mainThreadChatAgents2.ts:373	Telemetry — logs as chatAgentError via publicLogError2

Repro Steps

1. Open VS Code 1.117.0 with GitHub Copilot extension installed
2. Ensure "Auto mode" is selected as the model in the Copilot Chat model picker (not a specific model)
3. Send a chat message to Copilot
4. The error occurs when the extension's auto mode service cannot find any available model in its endpoint registry
5. This may be intermittent depending on backend model availability, network conditions, or account entitlements

Fix Proposal

The fix must be applied in the Copilot extension (not VS Code core). The error originates in automodeService.ts:_selectDefaultModel() when it finds no models across all known endpoints.

Fix location: Copilot extension → src/platform/endpoint/node/automodeService.ts → _selectDefaultModel() and resolveAutoModeEndpoint()

Approach:

1. Add resilient fallback behavior in _selectDefaultModel() — instead of throwing immediately when no model is found, retry with refreshed endpoint data or fall back to a default/known-good model
2. In resolveAutoModeEndpoint(), catch the "no model found" error and gracefully degrade (e.g., prompt the user to manually select a model instead of failing with an unrecoverable error)
3. Investigate why "known endpoints" returns empty — this may indicate a service-side regression in model availability or a stale endpoint cache

Recommended Owner

@lramos15 — owns "Auto Mode for Copilot Chat", "Model Picker", and "BYOK API for Copilot Chat" per working areas. Already correctly assigned to this issue.

Generated by errors-investigate · ● 5.4M · ◷

[bryanchen-d]

@lramos15 , is this an user error that we should suppress it from firing as an unhandled error into telemetry?