adopt to new way of configuring BYOK models

[sandy081]

@lramos15 adopted to new way of configuring BYOK models - microsoft/vscode#286616

Vendors requiring API key

Adopted Anthropic, xAI, Google, Groq, OpenRouter and OpenAI vendors to declare the required apiKey secret as part of the configuration of respective languageModelChatProviders contributions.

cc @bhavyaus (Anthropic), @vijayupadya (Google)

@vijayupadya - Skipped gemini spec tests. Please update them as per the latest adoption.

Example:

			{
				"vendor": "anthropic",
				"displayName": "Anthropic",
				"configuration": {
					"properties": {
						"apiKey": {
							"type": "string",
							"secret": true,
							"description": "API key for Anthropic"
						}
					},
					"required": [
						"apiKey"
					]
				}
			}

Vendors requiring end point

Adopted Ollama vendor to declare the required Ollama endpoint as part of the configuration of respective languageModelChatProviders contributions.

			{
				"vendor": "ollama",
				"displayName": "Ollama",
				"configuration": {
					"type": "object",
					"properties": {
						"url": {
							"type": "string",
							"description": "The endpoint URL for the Ollama server",
							"default": "http://localhost:11434"
						}
					},
					"required": [
						"url"
					]
				}
			}

Open AI Compatible & Azure vendors

Adopted OpenAI Compatible and Azure vendors that requires users to provide the models configuration in the settings to use the new format. This configuration is now declared as part of the languageModelChantProviders declaration - package.json

cc @eleanorjboyd

[vijayupadya]

Does each model now require a vendor config to be explicitly defined (for e.g. google)? Also, any reason the test cannot be updated for gemini in this PR?

[Copilot]

Pull request overview

This PR adopts a new way of configuring BYOK (Bring Your Own Key) models by moving from custom API key management UI to declarative configuration in package.json. The changes include:

Changes:

• Deprecated old configuration settings (azureAuthType, azureModels, customOAIModels, ollamaEndpoint) and moved them to a Deprecated namespace
• Added configuration schemas to language model chat provider declarations in package.json for all BYOK vendors
• Refactored all provider implementations to use the new abstract base classes and configuration system
• Removed custom UI for API key management (commands, configurator, UI service)
• Added migration logic to automatically move old configurations to new format

Reviewed changes

Copilot reviewed 21 out of 22 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
package.json	Added configuration schemas for BYOK providers; removed management commands
package.nls.json	Removed localization strings for deprecated settings
configurationServiceImpl.ts	Enhanced setConfig to handle undefined values across all configuration targets
configurationService.ts	Added Deprecated namespace for old settings
vscode.proposed.chatProvider.d.ts	Added configuration support to PrepareLanguageModelChatModelOptions
abstractLanguageModelChatProvider.ts	New base classes for language model chat providers
All provider files	Refactored to extend new base classes and use configuration-based approach
byokContribution.ts	Removed command registrations for API key management
Deleted files	Removed custom UI components (configurator, UI service, base provider)
Test files	Skipped Gemini tests; removed Azure provider tests; updated configuration tests

Comments suppressed due to low confidence (1)

package.json:1964

• The 'gpu' property has an incorrect description that states "Display name of the custom OpenAI model", which appears to be a copy-paste error. This is a duplicate of the same issue in the customoai configuration. Either correct the description to explain what the gpu property is for, or remove this property if it's not used.

									"gpu": {
										"type": "string",
										"description": "Display name of the custom OpenAI model"
									},

[lramos15]

Please take a look at the copilot spelling mistakes and fix those too thanks!

[sandy081]

@vijayupadya

Does each model now require a vendor config to be explicitly defined (for e.g. google)?
Yes. It was similar before but was not as explicit as now. One had to configure API key.

Also, any reason the test cannot be updated for gemini in this PR?
Because they seem to need another big refactoring and changes are getting too big. May be one can try to update these tests using AI after.

@lramos15 Resolved all feedback

[lramos15]

CI is red