configure autocomplete provider based on cody LLM settings in site config

lib/shared/src/sourcegraph-api/graphql/client.ts

@@ -140,6 +140,7 @@ export interface CodyLLMSiteConfiguration {
     fastChatModelMaxTokens?: number
     completionModel?: string
     completionModelMaxTokens?: number
+    provider?: string
 }
 
 interface IsContextRequiredForChatQueryResponse {

lib/shared/src/sourcegraph-api/graphql/queries.ts

@@ -46,6 +46,7 @@ query CurrentSiteCodyLlmConfiguration {
             fastChatModelMaxTokens
             completionModel
             completionModelMaxTokens
+            provider
         }
     }
 }`

vscode/src/completions/createVSCodeInlineCompletionItemProvider.ts

@@ -43,7 +43,7 @@ export async function createInlineCompletionItemProvider({
     const disposables: vscode.Disposable[] = []
 
     const [providerConfig, graphContextFlag] = await Promise.all([
-        createProviderConfig(config, client, featureFlagProvider),
+        createProviderConfig(config, client, featureFlagProvider, authProvider.getAuthStatus().configOverwrites),
         featureFlagProvider?.evaluateFeatureFlag(FeatureFlag.CodyAutocompleteGraphContext),
     ])
     if (providerConfig) {

vscode/src/completions/providers/createProvider.test.ts

@@ -0,0 +1,258 @@
+import { describe, expect, it } from 'vitest'
+
+import { Configuration } from '@sourcegraph/cody-shared/src/configuration'
+import { DOTCOM_URL } from '@sourcegraph/cody-shared/src/sourcegraph-api/environments'
+import { CodyLLMSiteConfiguration } from '@sourcegraph/cody-shared/src/sourcegraph-api/graphql/client'
+
+import { CodeCompletionsClient } from '../client'
+
+import { createProviderConfig } from './createProvider'
+
+const DEFAULT_VSCODE_SETTINGS: Configuration = {
+    serverEndpoint: DOTCOM_URL.href,
+    proxy: null,
+    codebase: '',
+    customHeaders: {},
+    chatPreInstruction: 'My name is John Doe.',
+    useContext: 'embeddings',
+    autocomplete: true,
+    experimentalCommandLenses: false,
+    experimentalEditorTitleCommandIcon: false,
+    experimentalChatPredictions: false,
+    experimentalGuardrails: false,
+    experimentalLocalSymbols: false,
+    inlineChat: true,
+    isRunningInsideAgent: false,
+    experimentalNonStop: false,
+    experimentalSymfAnthropicKey: '',
+    experimentalSymfPath: 'symf',
+    debugEnable: false,
+    debugVerbose: false,
+    debugFilter: null,
+    telemetryLevel: 'all',
+    autocompleteAdvancedProvider: null,
+    autocompleteAdvancedServerEndpoint: null,
+    autocompleteAdvancedModel: null,
+    autocompleteAdvancedAccessToken: null,
+    autocompleteAdvancedEmbeddings: true,
+    autocompleteExperimentalCompleteSuggestWidgetSelection: false,
+    autocompleteExperimentalSyntacticPostProcessing: false,
+    autocompleteExperimentalGraphContext: false,
+}
+
+const getVSCodeSettings = (config: Partial<Configuration> = {}): Configuration => ({
+    ...DEFAULT_VSCODE_SETTINGS,
+    ...config,
+})
+
+const dummyCodeCompletionsClient: CodeCompletionsClient = {
+    complete: () => Promise.resolve({ completion: '', stopReason: '' }),
+    onConfigurationChange: () => undefined,
+}
+
+describe('createProviderConfig', () => {
+    describe('if completions provider fields are defined in VSCode settings', () => {
+        it('returns null if completions provider is not supported', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({
+                    autocompleteAdvancedProvider: 'nasa-ai' as Configuration['autocompleteAdvancedProvider'],
+                }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider).toBeNull()
+        })
+
+        it('returns "codegen" provider config if the corresponding provider name and endpoint are specified', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({
+                    autocompleteAdvancedProvider: 'unstable-codegen',
+                    autocompleteAdvancedServerEndpoint: 'https://unstable-codegen.com',
+                }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider?.identifier).toBe('codegen')
+            expect(provider?.model).toBe('codegen')
+        })
+
+        it('returns null if provider is "unstable-codegen", but the server endpoint is not set', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({ autocompleteAdvancedProvider: 'unstable-codegen' }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider).toBeNull()
+        })
+
+        it('returns "fireworks" provider config and corresponding model if specified', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({
+                    autocompleteAdvancedProvider: 'unstable-fireworks',
+                    autocompleteAdvancedModel: 'starcoder-3b',
+                }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider?.identifier).toBe('fireworks')
+            expect(provider?.model).toBe('starcoder-3b')
+        })
+
+        it('returns "fireworks" provider config if specified in settings and default model', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({ autocompleteAdvancedProvider: 'unstable-fireworks' }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider?.identifier).toBe('fireworks')
+            expect(provider?.model).toBe('starcoder-7b')
+        })
+
+        it('returns "openai" provider config if specified in VSCode settings; model is ignored', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({
+                    autocompleteAdvancedProvider: 'unstable-openai',
+                    autocompleteAdvancedModel: 'hello-world',
+                }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider?.identifier).toBe('unstable-openai')
+            expect(provider?.model).toBe('gpt-35-turbo')
+        })
+
+        it('returns "anthropic" provider config if specified in VSCode settings; model is ignored', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({
+                    autocompleteAdvancedProvider: 'anthropic',
+                    autocompleteAdvancedModel: 'hello-world',
+                }),
+                dummyCodeCompletionsClient,
+                undefined,
+                {}
+            )
+            expect(provider?.identifier).toBe('anthropic')
+            expect(provider?.model).toBe('claude-instant-1')
+        })
+
+        it('provider specified in VSCode settings takes precedence over the one defined in the site config', async () => {
+            const provider = await createProviderConfig(
+                getVSCodeSettings({
+                    autocompleteAdvancedProvider: 'unstable-codegen',
+                    autocompleteAdvancedServerEndpoint: 'https://unstable-codegen.com',
+                }),
+                dummyCodeCompletionsClient,
+                undefined,
+                { provider: 'azure-open-ai', completionModel: 'gpt-35-turbo-test' }
+            )
+            expect(provider?.identifier).toBe('codegen')
+            expect(provider?.model).toBe('codegen')
+        })
+    })
+
+    describe('completions provider and model are defined in the site config and not set in VSCode settings', () => {
+        describe('if provider is "sourcegraph"', () => {
+            const testCases: {
+                codyLLMConfig: CodyLLMSiteConfiguration
+                expected: { provider: string; model?: string } | null
+            }[] = [
+                // sourcegraph
+                { codyLLMConfig: { provider: 'sourcegraph', completionModel: 'hello-world' }, expected: null },
+                {
+                    codyLLMConfig: { provider: 'sourcegraph', completionModel: 'anthropic/claude-instant-1' },
+                    expected: { provider: 'anthropic', model: 'claude-instant-1' },
+                },
+                {
+                    codyLLMConfig: { provider: 'sourcegraph', completionModel: 'anthropic/' },
+                    expected: null,
+                },
+                {
+                    codyLLMConfig: { provider: 'sourcegraph', completionModel: '/claude-instant-1' },
+                    expected: null,
+                },
+
+                // aws-bedrock
+                { codyLLMConfig: { provider: 'aws-bedrock', completionModel: 'hello-world' }, expected: null },
+                {
+                    codyLLMConfig: { provider: 'aws-bedrock', completionModel: 'anthropic.claude-instant-1' },
+                    expected: { provider: 'anthropic', model: 'claude-instant-1' },
+                },
+                {
+                    codyLLMConfig: { provider: 'aws-bedrock', completionModel: 'anthropic.' },
+                    expected: null,
+                },
+                {
+                    codyLLMConfig: { provider: 'aws-bedrock', completionModel: 'anthropic/claude-instant-1' },
+                    expected: null,
+                },
+
+                // open-ai
+                {
+                    codyLLMConfig: { provider: 'openai', completionModel: 'gpt-35-turbo-test' },
+                    expected: { provider: 'unstable-openai', model: 'gpt-35-turbo-test' },
+                },
+                {
+                    codyLLMConfig: { provider: 'openai' },
+                    expected: { provider: 'unstable-openai', model: 'gpt-35-turbo' },
+                },
+
+                // azure-openai
+                {
+                    codyLLMConfig: { provider: 'azure-openai', completionModel: 'gpt-35-turbo-test' },
+                    expected: { provider: 'unstable-openai', model: 'gpt-35-turbo-test' },
+                },
+                {
+                    codyLLMConfig: { provider: 'azure-openai' },
+                    expected: { provider: 'unstable-openai', model: 'gpt-35-turbo' },
+                },
+
+                // fireworks
+                {
+                    codyLLMConfig: { provider: 'fireworks', completionModel: 'llama-code-7b' },
+                    expected: { provider: 'fireworks', model: 'llama-code-7b' },
+                },
+                {
+                    codyLLMConfig: { provider: 'fireworks' },
+                    expected: { provider: 'fireworks', model: 'starcoder-7b' },
+                },
+
+                // unknown-provider
+                {
+                    codyLLMConfig: { provider: 'unknown-provider', completionModel: 'llama-code-7b' },
+                    expected: null,
+                },
+            ]
+
+            for (const { codyLLMConfig, expected } of testCases) {
+                it(`returns ${JSON.stringify(expected)} when cody LLM config is ${JSON.stringify(
+                    codyLLMConfig
+                )}`, async () => {
+                    const provider = await createProviderConfig(
+                        getVSCodeSettings(),
+                        dummyCodeCompletionsClient,
+                        undefined,
+                        codyLLMConfig
+                    )
+                    if (expected === null) {
+                        expect(provider).toBeNull()
+                    } else {
+                        expect(provider?.identifier).toBe(expected.provider)
+                        expect(provider?.model).toBe(expected.model)
+                    }
+                })
+            }
+        })
+    })
+
+    it('returns anthropic provider config if no completions provider specified in VSCode settings or site config', async () => {
+        const provider = await createProviderConfig(getVSCodeSettings(), dummyCodeCompletionsClient, undefined, {})
+        expect(provider?.identifier).toBe('anthropic')
+        expect(provider?.model).toBe('claude-instant-1')
+    })
+})