Nutlope

turboseek: Build an AI Search Engine with Web Sources

TurboSeek is a TypeScript web app that answers questions using web search results and language models. It is suited to developers who want to explore or adapt a Perplexity-style search experience, with external API credentials required to run it.

TypeScriptAILLMWeb App
View on GitHub Compare
Stars
1.7k
Forks
261
Stars, last 30 days
+2

Our take

Health score

33/100

How it is scored
Activity
28
Community
39
Issues
21
Pull requests
43

+2 stars in the last 30 days

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

1,661

Forks

258

Change since Nov 16, 2026: -1 stars·0 forks

Overview

TurboSeek is an open source AI search engine that combines web search results with language-model responses. It is designed to return an answer grounded in scraped pages, show search results, and suggest related questions for follow-up.

The project is most useful as a starting point for developers learning how to build AI search or customizing a search-focused web app. It depends on external services for search and model inference rather than providing those capabilities locally.

Key Features

  • Accepts a user's question and retrieves web search results.
  • Displays the top six search results described in the README's workflow.
  • Scrapes page text and supplies it as context for generating an answer.
  • Streams the generated response to the user.
  • Generates three related questions for follow-up.
  • Uses a Next.js app router interface styled with Tailwind.
  • Includes Helicone for observability and Plausible for website analytics.

Use Cases

  • Developers can use it to study the structure of an AI search application and follow the linked Together AI tutorial.
  • A team prototyping a search assistant can adapt the app's search, page-scraping, and answer-generation flow.
  • Users looking for a web research interface can try a search experience that pairs answers with visible search results.
  • Builders evaluating retrieval-augmented answer patterns can use it as a small web app example, while accounting for its external service dependencies.

Getting Started

Clone the repository, configure API keys in a .env file using .example.env as a reference, then run:

npm install && npm run dev

Search, inference, and observability services require accounts and credentials. See the README for setup details.

Considerations

  • Running the app requires external API credentials for language-model inference and search, and the README also lists Helicone for observability.
  • The README describes Exa.ai in the tech stack, but its workflow refers to Bing and its setup instructions mention Serper or Azure Bing Search. Check the current configuration before choosing a search provider.
  • The README lists planned work around citation accuracy, token limits, caching, rate limiting, sharing, and authentication. These should not be assumed to be implemented features.

Found this useful?

Share it with someone who would like turboseek.