Autocomplete: add request-params tests for the openaicompatible pro…

…vider (#5616)

- Adding request params tests for all possible configurations of the
`openaicompatible` provider.
- Follow up for #5604

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

@@ -0,0 +1,86 @@
+import { Observable } from 'observable-fns'
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
+
+import { featureFlagProvider, modelsService } from '@sourcegraph/cody-shared'
+
+import { mockLocalStorage } from '../../services/LocalStorageProvider'
+
+import {
+    type AutocompleteProviderValuesToAssert,
+    getAutocompleteProviderFromLocalSettings,
+    getAutocompleteProviderFromServerSideModelConfig,
+    getAutocompleteProviderFromSiteConfigCodyLLMConfiguration,
+    getRequestParamsWithoutMessages,
+} from './shared/helpers'
+
+describe('openaicompatible autocomplete provider', () => {
+    beforeEach(async () => {
+        mockLocalStorage()
+        vi.spyOn(featureFlagProvider, 'evaluatedFeatureFlag').mockReturnValue(Observable.of(false))
+    })
+
+    afterEach(() => {
+        modelsService.reset()
+    })
+
+    const valuesToAssert = {
+        providerId: 'openaicompatible',
+        legacyModel: 'llama-3.1-70b-versatile',
+        requestParams: {
+            maxTokensToSample: 256,
+            temperature: 0.2,
+            timeoutMs: 7000,
+            topK: 0,
+        },
+    } satisfies AutocompleteProviderValuesToAssert
+
+    it('throws if used with local-editor-settings', async () => {
+        const createCall = getAutocompleteProviderFromLocalSettings({
+            providerId: 'openaicompatible',
+            legacyModel: 'gpt-4o',
+            isDotCom: true,
+        })
+
+        await expect(createCall).rejects.toThrowErrorMatchingInlineSnapshot(
+            // biome-ignore lint/style/noUnusedTemplateLiteral: snapshot value
+            `[Error: Model definition is missing for \`openaicompatible\` provider.]`
+        )
+    })
+
+    it('[dotcom] server-side-model-config', async () => {
+        const provider = await getAutocompleteProviderFromServerSideModelConfig({
+            modelRef: 'groq::v1::llama-3.1-70b-versatile',
+            isDotCom: true,
+        })
+
+        // Switches to the first available model, because `llama-3.1-70b-versatile` is
+        // the enterprise tier model and cannot be used on DotCom.
+        expect(provider.id).toBe('anthropic')
+        expect(provider.legacyModel).toBe('anthropic/claude-instant-1.2')
+    })
+
+    it('[enterprise] server-side-model-config', async () => {
+        const provider = await getAutocompleteProviderFromServerSideModelConfig({
+            modelRef: 'groq::v1::llama-3.1-70b-versatile',
+            isDotCom: false,
+        })
+        const { providerId, legacyModel, requestParams } = valuesToAssert
+
+        expect(provider.id).toBe(providerId)
+        expect(provider.legacyModel).toBe(legacyModel)
+        expect(getRequestParamsWithoutMessages(provider)).toMatchObject(requestParams)
+    })
+
+    it('throws if used with site-config-cody-llm-configuration', async () => {
+        const createCall = getAutocompleteProviderFromSiteConfigCodyLLMConfiguration({
+            providerId: 'openaicompatible',
+            legacyModel: 'gpt-4o',
+            isDotCom: true,
+        })
+
+        await expect(createCall).rejects.toThrowErrorMatchingInlineSnapshot(
+            // biome-ignore lint/style/noUnusedTemplateLiteral: snapshot value
+            `[Error: Model definition is missing for \`openaicompatible\` provider.]`
+        )
+    })
+})

vscode/src/completions/providers/unstable-openai.test.ts

@@ -13,7 +13,7 @@ import {
     testAutocompleteProvider,
 } from './shared/helpers'
 
-describe('unstable-openai', () => {
+describe('unstable-openai autocomplete provider', () => {
     beforeEach(async () => {
         mockLocalStorage()
         vi.spyOn(featureFlagProvider, 'evaluatedFeatureFlag').mockReturnValue(Observable.of(false))