feat(web): unify AI providers behind OpenAI-compatible config (Ollama, Whisper, etc.)

apps/web/__tests__/unit/generate-ai-title.test.ts

@@ -12,9 +12,9 @@ vi.mock("@cap/web-backend", () => ({
 	Storage: {},
 }));
 
-vi.mock("@/lib/groq-client", () => ({
-	GROQ_MODEL: "test-model",
-	getGroqClient: vi.fn(() => null),
+vi.mock("@/lib/ai-provider", () => ({
+	getAiClient: vi.fn(() => null),
+	getAiModel: vi.fn(() => "test-model"),
 }));
 
 vi.mock("@/lib/server", () => ({

apps/web/actions/videos/get-status.ts

@@ -60,7 +60,10 @@ export async function getVideoStatus(
 
 	const metadata: VideoMetadata = (video.metadata as VideoMetadata) || {};
 
-	if (!video.transcriptionStatus && serverEnv().DEEPGRAM_API_KEY) {
+	if (
+		!video.transcriptionStatus &&
+		(serverEnv().DEEPGRAM_API_KEY || serverEnv().STT_BASE_URL)
+	) {
 		const activeUpload = await db()
 			.select({
 				videoId: videoUploads.videoId,
@@ -151,7 +154,9 @@ export async function getVideoStatus(
 		video.transcriptionStatus === "COMPLETE" &&
 		!metadata.aiGenerationStatus &&
 		!metadata.summary &&
-		(serverEnv().GROQ_API_KEY || serverEnv().OPENAI_API_KEY);
+		(serverEnv().GROQ_API_KEY ||
+			serverEnv().OPENAI_API_KEY ||
+			serverEnv().AI_BASE_URL);
 
 	if (shouldTriggerAiGeneration) {
 		try {

apps/web/actions/videos/translate-transcript.ts

@@ -6,7 +6,7 @@ import { Storage } from "@cap/web-backend";
 import type { Video } from "@cap/web-domain";
 import { eq } from "drizzle-orm";
 import { Effect, Option } from "effect";
-import { GROQ_MODEL, getGroqClient } from "@/lib/groq-client";
+import { getAiClient, getAiModel } from "@/lib/ai-provider";
 import { runPromise } from "@/lib/server";
 import { decodeStorageVideo } from "@/lib/video-storage";
 import {
@@ -38,8 +38,8 @@ export async function translateTranscript(
 		};
 	}
 
-	const groq = getGroqClient();
-	if (!groq) {
+	const ai = getAiClient();
+	if (!ai) {
 		return {
 			success: false,
 			message: "Translation service not configured",
@@ -94,7 +94,7 @@ export async function translateTranscript(
 	const translatedVtt = await translateVttContent(
 		originalVtt.value,
 		targetLanguage,
-		groq,
+		ai,
 	);
 
 	if (!translatedVtt) {
@@ -124,7 +124,7 @@ export async function translateTranscript(
 async function translateVttContent(
 	vttContent: string,
 	targetLanguage: LanguageCode,
-	groq: NonNullable<ReturnType<typeof getGroqClient>>,
+	ai: NonNullable<ReturnType<typeof getAiClient>>,
 ): Promise<string | null> {
 	const targetLanguageName = SUPPORTED_LANGUAGES[targetLanguage];
 
@@ -144,8 +144,8 @@ VTT content to translate:
 ${vttContent}`;
 
 	try {
-		const response = await groq.chat.completions.create({
-			model: GROQ_MODEL,
+		const response = await ai.chat.completions.create({
+			model: getAiModel(),
 			messages: [{ role: "user", content: prompt }],
 			temperature: 0.3,
 			max_tokens: 8000,

apps/web/lib/ai-provider.ts

@@ -0,0 +1,59 @@
+import "server-only";
+
+import { serverEnv } from "@cap/env";
+import OpenAI from "openai";
+
+const GROQ_DEFAULT_BASE_URL = "https://api.groq.com/openai/v1";
+const GROQ_DEFAULT_MODEL = "openai/gpt-oss-120b";
+const OPENAI_DEFAULT_MODEL = "gpt-4o-mini";
+const STT_DEFAULT_MODEL = "whisper-1";
+
+let aiClient: OpenAI | null = null;
+let sttClient: OpenAI | null = null;
+
+export function getAiClient(): OpenAI | null {
+	if (aiClient) return aiClient;
+
+	const env = serverEnv();
+	if (env.AI_BASE_URL) {
+		aiClient = new OpenAI({
+			baseURL: env.AI_BASE_URL,
+			apiKey:
+				env.AI_API_KEY ?? env.GROQ_API_KEY ?? env.OPENAI_API_KEY ?? "none",
+		});
+	} else if (env.GROQ_API_KEY) {
+		aiClient = new OpenAI({
+			baseURL: GROQ_DEFAULT_BASE_URL,
+			apiKey: env.GROQ_API_KEY,
+		});
+	} else if (env.OPENAI_API_KEY) {
+		aiClient = new OpenAI({ apiKey: env.OPENAI_API_KEY });
+	}
+
+	return aiClient;
+}
+
+export function getAiModel(): string {
+	const env = serverEnv();
+	if (env.AI_MODEL) return env.AI_MODEL;
+	if (!env.AI_BASE_URL && env.GROQ_API_KEY) return GROQ_DEFAULT_MODEL;
+	return OPENAI_DEFAULT_MODEL;
+}
+
+export function getSttClient(): OpenAI | null {
+	if (sttClient) return sttClient;
+
+	const env = serverEnv();
+	if (!env.STT_BASE_URL) return null;
+
+	sttClient = new OpenAI({
+		baseURL: env.STT_BASE_URL,
+		apiKey: env.STT_API_KEY ?? "none",
+	});
+
+	return sttClient;
+}
+
+export function getSttModel(): string {
+	return serverEnv().STT_MODEL ?? STT_DEFAULT_MODEL;
+}

apps/web/lib/generate-ai.ts

@@ -16,10 +16,15 @@ export async function startAiGeneration(
 	videoId: Video.VideoId,
 	userId: string,
 ): Promise<GenerateAiResult> {
-	if (!serverEnv().GROQ_API_KEY && !serverEnv().OPENAI_API_KEY) {
+	if (
+		!serverEnv().GROQ_API_KEY &&
+		!serverEnv().OPENAI_API_KEY &&
+		!serverEnv().AI_BASE_URL
+	) {
 		return {
 			success: false,
-			message: "Missing AI API keys (Groq or OpenAI)",
+			message:
+				"Missing AI API keys (set GROQ_API_KEY, OPENAI_API_KEY, or AI_BASE_URL)",
 		};
 	}
 

apps/web/lib/messenger/agent.ts

@@ -2,7 +2,7 @@ import "server-only";
 
 import type { MessengerMessageRole } from "@cap/database/schema";
 import { serverEnv } from "@cap/env";
-import { GROQ_MODEL, getGroqClient } from "@/lib/groq-client";
+import { getAiClient, getAiModel } from "@/lib/ai-provider";
 import { CAP_REFERENCE_GUIDE, MESSENGER_AGENT_PROMPT } from "./constants";
 import { getKnowledgeTag, searchSupermemory } from "./supermemory";
 
@@ -165,18 +165,18 @@ const callOpenAi = async ({
 	return parseOpenAiContent(payload);
 };
 
-const callGroq = async ({
+const callAiProvider = async ({
 	systemPrompt,
 	history,
 }: {
 	systemPrompt: string;
 	history: ConversationMessage[];
 }) => {
-	const client = getGroqClient();
+	const client = getAiClient();
 	if (!client) return null;
 
 	const completion = await client.chat.completions.create({
-		model: GROQ_MODEL,
+		model: getAiModel(),
 		temperature: 0.65,
 		max_tokens: 500,
 		messages: [
@@ -227,8 +227,10 @@ export const generateMessengerAgentReply = async ({
 	);
 	if (fromOpenAi) return fromOpenAi;
 
-	const fromGroq = await callGroq({ systemPrompt, history }).catch(() => null);
-	if (fromGroq) return fromGroq;
+	const fromAi = await callAiProvider({ systemPrompt, history }).catch(
+		() => null,
+	);
+	if (fromAi) return fromAi;
 
 	return "Oh no, I'm so sorry about this! I'm having a little technical hiccup on my end. Someone from the team will jump in here shortly to help you out though!";
 };

apps/web/lib/transcribe.ts

@@ -17,7 +17,7 @@ export async function transcribeVideo(
 	aiGenerationEnabled = false,
 	_isRetry = false,
 ): Promise<TranscribeResult> {
-	if (!serverEnv().DEEPGRAM_API_KEY) {
+	if (!serverEnv().DEEPGRAM_API_KEY && !serverEnv().STT_BASE_URL) {
 		return {
 			success: false,
 			message: "Missing necessary environment variables",

apps/web/package.json

@@ -100,7 +100,6 @@
 		"framer-motion": "^11.13.1",
 		"geist": "^1.3.1",
 		"gif.js": "0.2.0",
-		"groq-sdk": "^0.29.0",
 		"hls.js": "^1.5.3",
 		"hono": "^4.7.1",
 		"js-cookie": "^3.0.5",
@@ -114,6 +113,7 @@
 		"next": "16.2.1",
 		"next-auth": "^4.24.5",
 		"next-mdx-remote": "^6.0.0",
+		"openai": "^4.86.0",
 		"posthog-js": "^1.215.3",
 		"posthog-node": "^4.11.2",
 		"prettier": "^3.7.4",