# Project export: Zue Research This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small. ## Project metadata - Hackathon: TreeHacks 2025 - Tagline: interactive multimodal AI experiences for neurodivergent learners run entirely on the edge - Devpost: https://devpost.com/software/zue-research - GitHub: https://github.com/zcsabbagh/zu-lm/ - Video: https://www.youtube.com/embed/ajYL27WPKcs?enablejsapi=1&hl=en_US&rel=0&start=&version=3&wmode=transparent - Result: winner (Orbstack: Best Use of Rust ($1k Cash)) - Team: 3 GitHub contributor(s) — zcsabbag (31 commits), uche-afk (28 commits), Emily Zhang (10 commits) ## Devpost submission (written by the team) ### Overview TL;DR: we created interactive multimodal AI journeys for neurodivergent learners which can be run entirely on the edge ### Inspiration 56% of students in the world do not have internet access at school [1]. Students learn best through multi-sensory, hands-on, structured experiences that are tailored to their interests. Yet, the primary method of learning for them continues to be textbooks. This is even more pertinent to the 129 million students globally who have ADHD [2]. We built in our product with neurodivergent and offline learners in mind. Drawing on bleeding edge research, we proved that multimodal applications can be deployed entirely at the edge using distributed inference using heterogeneous compute [3] [4]. What It Does After students describe what they’d like to learn more about, we create multiple adversarial agents which perform “deep research” (i.e. search the web, synthesize, reflect, and repeat). Think anything from “debate the entirety of modern Tunisian history” to “tell me how the organelles of a plant cell differ from those of an animal”. When these agents finish, they summarize their findings and provide related “branches” the students can look into. If the student is done researching, we generate a podcast with visual aids that shows multiple agents debating each other. At any point in the podcast, a student can double click on a segment to have a new research agent answer questions or clarify topics. How We Built It We needed to create a dynamic and engaging frontend that would be easy-to-use for teachers and students alike. We chose Typescript and spun up an application that could provide live insights into the actions that the agents were taking, and playback the podcast in a lively manner. For our backend, we faced the challenge of orchestrating researchers engaging in Multi-Agent Debates. Despite being first-time users, we chose Rust for its superior concurrent performance and memory safety features, which were crucial for managing shared state across multiple research agents and handling asynchronous web operations safely. To support our cloud implementations, we chose ElevenLabs to generate the voices of our podcast and to create an agent to converse with the student after the podcast to test their understanding using the Feynman technique. We used LumaLabs for the podcast image generation. We also made creative use of the Perplexity Sonar web search API and Mistral via groq. Accomplishments We're Proud Of We were able to get our backend to run locally by sharding full size large language models across multiple hardware devices (i.e. we ran Llama on between a MacBook Pro and 2x MacMinis). Despite never having written code in Rust, we wrote our entire research agent server in a Rust implementation of LangChain. Some cool things we did with Rust: Tokio for async runtime and concurrent processing for multiple agents let (track_one_result, track_two_result) = tokio::join!( self.process_track(state.clone(), "one"), self.process_track(state.clone(), "two") ); Tokio for async runtime and concurrent processing for multiple agents Arc> for thread-safe shared state let state = Arc::new(Mutex::new( SummaryState::with_research_topic(input.research_topic.clone()) )); Arc> for thread-safe shared state What We Learned Building for education is both technologically challenging and highly rewarding. The members of our team were able to learn Rust from the ground up, taking advantage of its supreme efficiency, and learn how to build for the modern Ed-Tech consumer. Sources [1] https://ourworldindata.org/grapher/primary-schools-with-access-to-internet?tab=table [2] https://chadd.org/about-adhd/general-prevalence/ [3] https://arxiv.org/html/2405.14371v1 [4] https://github.com/exo-explore/exo ## README (from the GitHub repository) # ZU-LM A research-driven podcast generator that creates engaging conversations about any topic. ## Environment Setup This project requires several API keys to function properly. Follow these steps to set up your environment: 1. Copy the example environment files: ```bash # Root directory cp .env.example .env # zu-chat directory cd zu-chat cp .env.example .env.local # researcher directory cd ../researcher cp .env.example .env ``` 2. Obtain the required API keys: - [ElevenLabs](https://elevenlabs.io/) - For text-to-speech - [Groq](https://groq.com/) - For LLM inference - [OpenAI](https://openai.com/) - For AI capabilities - [Luma AI](https://lumalabs.ai/) - For image generation - [Perplexity](https://www.perplexity.ai/) - For search capabilities 3. Fill in your API keys in the respective .env files ## Required API Keys The following API keys are required for full functionality: - `ELEVENLABS_API_KEY` - For text-to-speech generation - `GROQ_API_KEY` - For LLM inference - `OPENAI_API_KEY` - For AI capabilities - `LUMAAI_API_KEY` - For image generation - `PERPLEXITY_API_KEY` - For search functionality ## Optional Configuration Some features can be configured through environment variables: - `LOCAL_LLM` - Specify which local LLM to use (default: "deepseek-r1:8b") - `MAX_WEB_RESEARCH_LOOPS` - Control research depth (default: 1) - `SEARCH_API` - Search provider to use (default: "perplexity") ## Run the project To run the frontend server: ``` cd zu-chat npm install npm run dev ``` To run the Rust backend: ``` cd researcher cargo run ``` Enjoy! ## Detected evidence (automated analysis) Indexed codebase: 63 recognized source files, 411 KB. - CSS (language) — detected in the code - JavaScript (language) — detected in the code - LangChain (technology) — detected in the code - Next.js (technology) — detected in the code - OpenAI (technology) — detected in the code - Python (language) — detected in the code - React (technology) — detected in the code - Rust (language) — detected in the code - Tailwind CSS (technology) — detected in the code - TypeScript (language) — detected in the code - Vercel AI SDK (technology) — detected in the code - Mistral AI (technology) — claimed on Devpost, not found in the code ## Codebase structure (from repository index) ### Files (80 of 80) ``` .env.example .gitattributes .gitignore .gitmodules agent_animation.py Cargo.toml package.json README.md researcher/.env.example researcher/.gitignore researcher/Cargo.toml researcher/langgraph.json researcher/src/assistant/configuration.rs researcher/src/assistant/debate.rs researcher/src/assistant/graph.rs researcher/src/assistant/groq.rs researcher/src/assistant/mod.rs researcher/src/assistant/prompts.rs researcher/src/assistant/state.rs researcher/src/assistant/utils.rs researcher/src/lib.rs researcher/src/main.rs researcher/src/server.rs src/demo.js src/index.js src/lib.rs src/luma.js src/main.rs src/schemas.js src/stream.js src/text.js src/voice.js tsconfig.json zu-chat/.env.example zu-chat/.gitignore zu-chat/components.json zu-chat/eslint.config.mjs zu-chat/log.txt zu-chat/next.config.ts zu-chat/package.json zu-chat/postcss.config.mjs zu-chat/README.md zu-chat/src/app/api/enrich/route.ts zu-chat/src/app/api/generate/images/route.ts zu-chat/src/app/api/generate/route.ts zu-chat/src/app/api/research/config/route.ts zu-chat/src/app/api/research/route.ts zu-chat/src/app/api/research/status/route.ts zu-chat/src/app/globals.css zu-chat/src/app/layout.tsx zu-chat/src/app/page.tsx zu-chat/src/app/research/App.css zu-chat/src/app/research/page.tsx zu-chat/src/app/test/page.tsx zu-chat/src/components/Chat.tsx zu-chat/src/components/podcast/EnrichmentPopup.tsx zu-chat/src/components/podcast/PodcastControls.tsx zu-chat/src/components/podcast/PodcastNavigation.tsx zu-chat/src/components/podcast/PodcastPlayer.tsx zu-chat/src/components/podcast/PodcastSegment.tsx zu-chat/src/components/podcast/PodcastTest.tsx zu-chat/src/components/research/ResearchHeader.tsx zu-chat/src/components/research/ResearchPerspectives.tsx zu-chat/src/components/ResearchFlow.tsx zu-chat/src/components/ResearchInputNode.tsx zu-chat/src/components/ui/button.tsx zu-chat/src/components/ui/card.tsx zu-chat/src/components/ui/hover-card.tsx zu-chat/src/components/ui/input.tsx zu-chat/src/components/ui/label.tsx zu-chat/src/components/ui/switch.tsx zu-chat/src/components/ui/tabs.tsx zu-chat/src/lib/constants.ts zu-chat/src/lib/luma.ts zu-chat/src/lib/schemas.ts zu-chat/src/lib/text.ts zu-chat/src/lib/utils.ts zu-chat/src/lib/voice.ts zu-chat/tailwind.config.ts zu-chat/tsconfig.json ``` ### Dependencies - package.json: @ai-sdk/groq@latest, @ai-sdk/openai@^1.1.11, ai@^4.1.40, dotenv@latest, elevenlabs@^1.51.0, lumaai@^1.4.0, node-fetch@^3.3.2, tailwindcss-animate@^1.0.7 - researcher/Cargo.toml: anyhow@1.0, async-stream@0.3, async-trait@0.1, axum@0.7, dotenv@0.15, enum-as-inner@0.6, futures@0.3, http@1.0, langchain@0.2.2, ollama-rs@0.1, reqwest@0.11, serde@1.0, serde_json@1.0, thiserror@1.0, tokio@1.0, tokio-stream@0.1, tower-http@0.5, tracing@0.1, url@2.5 - zu-chat/package.json: @11labs/react@^0.0.7, @ai-sdk/groq@^1.1.9, @ai-sdk/openai@^1.1.11, @eslint/eslintrc@^3, @radix-ui/react-hover-card@^1.1.6, @radix-ui/react-label@^2.1.2, @radix-ui/react-slot@^1.1.2, @radix-ui/react-switch@^1.1.3, @radix-ui/react-tabs@^1.1.3, @types/d3@^7.4.3, @types/dagre@^0.7.52, @types/node@^20, @types/react@^19, @types/react-dom@^19, @xyflow/react@^12.4.3, ai@^4.1.40, class-variance-authority@^0.7.1, clsx@^2.1.1, crypto@^1.0.1, d3@^7.9.0, dagre@^0.8.5, elevenlabs@^1.51.0, eslint@^9, eslint-config-next@15.1.7, lucide-react@^0.475.0, lumaai@^1.4.0, next@15.1.7, postcss@^8, react@^19.0.0, react-d3-tree@^3.6.5, react-dom@^19.0.0, react-markdown@^9.0.3, tailwind-merge@^3.0.1, tailwindcss@^3.4.1, tailwindcss-animate@^1.0.7, typescript@^5, zod@^3.24.2 ### Recent commits (newest first) - Update README.md - Update README.md - chore: update gitignore and remove ignored files - minor - edit readme - update submodule name again - rename submodule - added exo submodule - final - fix - added feynman review - ALMOST DONE - merge - changes - finished integrating enrichment with main - Merge branch 'main' of https://github.com/zcsabbagh/zu-lm - enrich mode now working - new node - added voice agent to /test route - Merge branch 'main' of https://github.com/zcsabbagh/zu-lm ## Key source files (fetched from GitHub, selected and truncated for size) ### Cargo.toml ``` [workspace] members = ["researcher"] resolver = "2" ``` ### package.json ``` { "name": "zu-lm", "type": "module", "scripts": { "start": "ts-node src/index.ts", "build": "tsc", "dev": "ts-node src/index.ts" }, "dependencies": { "@ai-sdk/groq": "latest", "@ai-sdk/openai": "^1.1.11", "ai": "^4.1.40", "dotenv": "latest", "elevenlabs": "^1.51.0", "lumaai": "^1.4.0", "node-fetch": "^3.3.2", "tailwindcss-animate": "^1.0.7" } } ``` ### researcher/Cargo.toml ``` [package] name = "researcher" version = "0.1.0" edition = "2021" [[bin]] name = "researcher" path = "src/main.rs" [dependencies] serde = { version = "1.0", features = ["derive"] } serde_json = "1.0" langchain = "0.2.2" tokio = { version = "1.0", features = ["full", "net"] } async-trait = "0.1" anyhow = "1.0" dotenv = "0.15" reqwest = { version = "0.11", features = ["json"] } thiserror = "1.0" tracing = "0.1" url = "2.5" enum-as-inner = "0.6" ollama-rs = "0.1" # For Ollama integration axum = { version = "0.7", features = ["json", "macros"] } tower-http = { version = "0.5", features = ["cors", "fs"] } futures = "0.3" tokio-stream = "0.1" async-stream = "0.3" http = "1.0" ``` ### zu-chat/package.json ``` { "name": "zu-chat", "version": "0.1.0", "private": true, "scripts": { "dev": "next dev --turbopack", "build": "next build", "start": "next start", "lint": "next lint" }, "dependencies": { "@11labs/react": "^0.0.7", "@ai-sdk/groq": "^1.1.9", "@ai-sdk/openai": "^1.1.11", "@radix-ui/react-hover-card": "^1.1.6", "@radix-ui/react-label": "^2.1.2", "@radix-ui/react-slot": "^1.1.2", "@radix-ui/react-switch": "^1.1.3", "@radix-ui/react-tabs": "^1.1.3", "@types/d3": "^7.4.3", "@types/dagre": "^0.7.52", "@xyflow/react": "^12.4.3", "ai": "^4.1.40", "class-variance-authority": "^0.7.1", "clsx": "^2.1.1", "crypto": "^1.0.1", "d3": "^7.9.0", "dagre": "^0.8.5", "elevenlabs": "^1.51.0", "lucide-react": "^0.475.0", "lumaai": "^1.4.0", "next": "15.1.7", "react": "^19.0.0", "react-d3-tree": "^3.6.5", "react-dom": "^19.0.0", "react-markdown": "^9.0.3", "tailwind-merge": "^3.0.1", "tailwindcss-animate": "^1.0.7", "zod": "^3.24.2" }, "devDependencies": { "@eslint/eslintrc": "^3", "@types/node": "^20", "@types/react": "^19", "@types/react-dom": "^19", "eslint": "^9", "eslint-config-next": "15.1.7", "postcss": "^8", "tailwindcss": "^3.4.1", "typescript": "^5" } } ``` ### src/index.js ```javascript import { createText, MODELS as TEXT_MODELS } from './text.js'; import { createTextStream, MODELS as STREAM_MODELS } from './stream.js'; import { textToSpeech, playPodcastTranscript } from './voice.js'; export { // Text generation createText, createTextStream, TEXT_MODELS, STREAM_MODELS, // Voice synthesis textToSpeech, playPodcastTranscript }; ``` ### src/main.rs ```rust use researcher::{init, Configuration}; #[tokio::main] async fn main() { println!("Starting application..."); // Initialize environment variables from .env init(); println!("Initialized environment variables"); // Now try to create the configuration match Configuration::from_runnable_config(None) { Ok(config) => { println!("Successfully loaded configuration"); // Your research logic here } Err(e) => { eprintln!("Configuration error: {}", e); } } } ``` ### researcher/src/main.rs ```rust use researcher::{ init, assistant::configuration::Configuration, server::run_server, }; #[tokio::main] async fn main() { println!("Starting application..."); // Initialize environment variables from .env init(); println!("Initialized environment variables"); // Now try to create the configuration match Configuration::from_runnable_config(None) { Ok(config) => { println!("Successfully loaded configuration"); // Run the server run_server(config).await; } Err(e) => { eprintln!("Configuration error: {}", e); } } } ``` ### researcher/src/server.rs ```rust use axum::{ routing::{post, get, put}, Router, Json, extract::State, http::StatusCode, response::{IntoResponse, Response, sse::{Event, Sse}}, }; use std::{ sync::Arc, net::SocketAddr, convert::Infallible, time::{Duration, SystemTime, UNIX_EPOCH}, env, sync::OnceLock, }; use crate::assistant::{ configuration::Configuration, state::{SummaryStateInput, StatusUpdate}, graph::ResearchGraph, configuration::ResearchMode, }; use tower_http::cors::CorsLayer; use futures::stream::Stream; use tokio::sync::broadcast; use serde_json::json; use serde::Deserialize; use tokio::sync::Mutex; use http::{Method, header}; use http::header::HeaderValue; // Increase channel capacity const CHANNEL_CAPACITY: usize = 100; static STATUS_CHANNEL: OnceLock> = OnceLock::new(); fn get_status_channel() -> broadcast::Sender { STATUS_CHANNEL.get_or_init(|| { let (tx, _) = broadcast::channel(CHANNEL_CAPACITY); tx }).clone() } #[derive(Clone)] pub struct AppState { graph: Arc>, status_tx: broadcast::Sender, } #[derive(Deserialize)] struct ConfigUpdate { local_llm: Option, max_web_research_loops: Option, research_mode: Option, } #[derive(serde::Deserialize)] pub struct ResearchRequest { topic: String, } #[derive(serde::Serialize)] struct ResearchResponse { summary: String, status: String, } // Custom error type for our API struct ApiError(anyhow::Error); impl IntoResponse for ApiError { fn into_response(self) -> Response { ( StatusCode::INTERNAL_SERVER_ERROR, Json(ResearchResponse { summary: format!("Error: {}", self.0), status: "Error occurred".to_string(), }) ).into_response() } } // Convert anyhow::Error into our ApiError impl From for ApiError { fn from(err: anyhow::Error) -> Self { ApiError(err) } } // Add this new struct for the config response #[derive(serde::Serialize)] struct ConfigResponse { local_llm: String, max_web_research_loops: i32, research_mode: ResearchMode, groq_model: String, } pub async fn run_server(config: Configuration) { let status_tx = get_status_channel(); let mut graph = ResearchGraph::new(config); graph.set_status_sender(status_tx.clone()); let state = Arc::new(AppState { graph: Arc::new(Mutex::new(graph)), status_tx, }); let frontend_origin = env::var("FRONTEND_URL") .unwrap_or_else(|_| "http://localhost:3000".to_string()); let cors = CorsLayer::new() .allow_methods([Method::GET, Method::POST, Method::PUT]) .allow_headers([ header::CONTENT_TYPE, header::ACCEPT, header::CACHE_CONTROL, header::CONNECTION, ]) .allow_credentials(true) .allow_origin(frontend_origin.parse::().unwrap()); let app = Router::new() .route("/research", post(handle_research)) .route("/config", put(update_config)) .route("/config", get(get_config)) .route("/status", get(status_stream)) .layer(cors) .with_state(state); let port = env::var("PORT").unwrap_or_else(|_| "4000".to_string()).parse::().unwrap_or(4000); let addr = SocketAddr::from(([0, 0, 0, 0], port)); println!("Starting server on http://localhost:{}", port); println!("Allowing CORS for origin: {}", frontend_origin); axum::serve( tokio::net::TcpListener::bind(&addr).await.unwrap(), app.into_make_service(), ) .await .unwrap(); } async fn handle_research( State(state): State>, Json(request): Json, ) -> Response { let input = SummaryStateInput { research_topic: request.topic, }; // Send initial status update let mut status = StatusUpdate::default(); status.phase = "init".to_string(); status.message = format!("Starting research on topic: {}", input.research_topic); status.timestamp = SystemTime::now().duration_since(UNIX_EPOCH).unwrap().as_secs(); if let Err(e) = state.status_tx.send(status) { eprintln!("Failed to send initial status update: {}", e); } let mut graph = state.graph.lock().await; // Update the graph's status sender graph.set_status_sender(state.status_tx.clone()); match graph.process_research(input).await { Ok(output) => { // Send final status update let mut status = StatusUpdate::default(); status.phase = "complete".to_string(); status.message = output.running_summary.clone(); status.timestamp = SystemTime::now().duration_since(UNIX_EPOCH).unwrap().as_secs(); let _ = state.status_tx.send(status); ( StatusCode::OK, Json(ResearchResponse { summary: output.running_summary, status: "Research completed".to_string(), }) ).into_response() }, Err(e) => { eprintln!("Research error: {:?}", e); let error_message = e.to_string(); let mut status = StatusUpdate::default(); status.phase = "error".to_string(); status.message = format!("Error: {}", error_message); status.timestamp = SystemTime::now().duration_since(UNIX_EPOCH).unwrap().as_secs(); let _ = state.status_tx.send(status); ( StatusCode::INTERNAL_SERVER_ERROR, Json(ResearchResponse { summary: format!("Error: {}", error_message), status: "Error occurred".to_string(), }) ).into_response() }, } } async fn status_stream( [truncated — 3501 more characters] ``` ### zu-chat/src/app/layout.tsx ```typescript import type { Metadata } from "next"; import { Geist, Geist_Mono } from "next/font/google"; import "./globals.css"; const geistSans = Geist({ variable: "--font-geist-sans", subsets: ["latin"], }); const geistMono = Geist_Mono({ variable: "--font-geist-mono", subsets: ["latin"], }); export const metadata: Metadata = { title: "Create Next App", description: "Generated by create next app", }; export default function RootLayout({ children, }: Readonly<{ children: React.ReactNode; }>) { return ( {children} ); } ``` ### zu-chat/src/app/page.tsx ```typescript "use client"; import { useState, useEffect, useRef } from "react"; import { useChat } from "ai/react"; import Link from "next/link"; import { Button } from "@/components/ui/button"; import { Card } from "@/components/ui/card"; import { PodcastControls } from "@/components/podcast/PodcastControls"; import { PodcastPlayer } from "@/components/podcast/PodcastPlayer"; import { ResearchHeader } from "@/components/research/ResearchHeader"; import { ResearchPerspectives } from "@/components/research/ResearchPerspectives"; import { DEFAULT_RESEARCH_SUMMARY } from "@/lib/constants"; import { useConversation } from "@11labs/react"; const LANGUAGES = [ { value: "English", label: "English" }, { value: "Spanish", label: "Spanish" }, { value: "Chinese", label: "Chinese" }, { value: "Russian", label: "Russian" }, { value: "Arabic", label: "Arabic" }, { value: "Japanese", label: "Japanese" }, { value: "Korean", label: "Korean" }, { value: "French", label: "French" }, { value: "German", label: "German" }, { value: "Italian", label: "Italian" }, { value: "Portuguese", label: "Portuguese" }, ] as const; const DURATIONS = [ { value: "1", label: "1 minute" }, { value: "2", label: "2 minutes" }, { value: "3", label: "3 minutes" }, { value: "5", label: "5 minutes" }, ] as const; interface SpeakerImage { speaker: string; imageUrl: string | null; error?: string; } export default function Home() { const [isGenerating, setIsGenerating] = useState(false); const [isGeneratingImages, setIsGeneratingImages] = useState(false); const [isReady, setIsReady] = useState(false); const [language, setLanguage] = useState("English"); const [duration, setDuration] = useState("3"); const [audioSegments, setAudioSegments] = useState([]); const [transcript, setTranscript] = useState([]); const [currentSegment, setCurrentSegment] = useState(0); const [displayedMessages, setDisplayedMessages] = useState([]); const [speakerImages, setSpeakerImages] = useState([]); const [trackOne, setTrackOne] = useState(null); const [trackTwo, setTrackTwo] = useState(null); const [selectedTrack, setSelectedTrack] = useState<"one" | "two">("one"); const audioRef = useRef(null); const { messages, setMessages } = useChat(); const [isTestMode, setIsTestMode] = useState(false); useEffect(() => { // Load research summaries from localStorage const trackOne = localStorage.getItem("researchSummaryTrackOne"); const trackTwo = localStorage.getItem("researchSummaryTrackTwo"); if (trackOne) setTrackOne(trackOne); if (trackTwo) setTrackTwo(trackTwo); }, []); useEffect(() => { // Handle audio playback and transcript synchronization if (isReady && audioSegments.length > 0 && currentSegment < audioSegments.length) { const audio = audioRef.current; if (audio) { // Stop any current playback audio.pause(); // Create and play the new segment const blob = new Blob([audioSegments[currentSegment]], { type: "audio/mpeg" }); const url = URL.createObjectURL(blob); audio.src = url; // Start playing automatically audio.play().catch((error) => { console.error("Error playing audio:", error); }); // Clean up previous URL if it exists if (audio.dataset.previousUrl) { URL.revokeObjectURL(audio.dataset.previousUrl); } audio.dataset.previousUrl = url; // Update displayed message if not already shown if (!displayedMessages[currentSegment]) { setDisplayedMessages((prev) => { const newMessages = [...prev]; newMessages[currentSegment] = { role: "assistant", content: `${transcript[currentSegment].speaker}: ${transcript[currentSegment].text}`, }; return newMessages; }); } audio.onended = () => { if (currentSegment < audioSegments.length - 1) { setCurrentSegment((prev) => prev + 1); } }; // Cleanup function return () => { audio.pause(); if (audio.dataset.previousUrl) { URL.revokeObjectURL(audio.dataset.previousUrl); } }; } } }, [audioSegments, currentSegment, transcript, isReady, displayedMessages]); const handleGenerate = async () => { setIsGenerating(true); setIsReady(false); setDisplayedMessages([]); setCurrentSegment(0); setSpeakerImages([]); try { const researchSummary = selectedTrack === "one" ? trackOne : trackTwo; // Generate podcast audio const response = await fetch("/api/generate", { method: "POST", headers: { "Content-Type": "application/json", }, body: JSON.stringify({ language, minutes: duration, researchSummary: researchSummary || DEFAULT_RESEARCH_SUMMARY, }), }); if (!response.ok) { throw new Error("Failed to generate podcast"); } const data = await response.json(); if (data.error) { throw new Error(data.error); } // Convert base64 audio segments to ArrayBuffer const segments = data.audioSegments.map((base64Audio: string) => { try { const binaryString = atob(base64Audio); const bytes = new Uint8Array(binaryString.length); for (let i = 0; i < binaryString.length; i++) { bytes[i] = binaryString.charCodeAt(i); } return bytes.buffer; } catch (error) { console.error("Error decoding audio segment:", error); throw new Error("Failed to decode audio data"); } }); setAudioSegments(segments); setTranscript(data.transcript); // Generate images for all segments setIsGeneratingImages(tr [truncated — 6300 more characters] ``` [53 more indexed source files omitted to keep this export small. The full file list is in the Codebase structure section above.]