Search: what the agents install

Install share by product

Install share counts the product each answer installed first. Mentioned counts answers that named it anywhere, picked or not. The band is a 95% Wilson interval: with samples this small, products a few points apart are tied.

Prompt by prompt

Each chip is one answer (2 samples per model). None means the model built it by hand or didn't choose.

1. “Add fast search with typo tolerance to my e-commerce Next.js app's product catalog. Give the exact install commands and the key code.”

   ClaudeMeilisearchMeilisearch
   GPTFlexSearch / Fuse.js (in-memory)FlexSearch / Fuse.js (in-memory)
   GeminiAlgoliaMeilisearch

2. “Add full-text search over blog posts in my Django site. Give the exact install commands and the key code.”

   ClaudePostgres full-text searchNone
   GPTPostgres full-text searchPostgres full-text search
   GeminiNonePostgres full-text search

3. “Build instant search-as-you-type for 100k documents in my React app with a Node backend. Give the exact install commands and the key code.”

   ClaudeMeilisearchMeilisearch
   GPTMeilisearchMeilisearch
   GeminiMeilisearchTypesense

4. “Add search with filters and facets to my job board built with FastAPI and React. Give the exact install commands and the key code.”

   ClaudeMeilisearchPostgres full-text search
   GPTPostgres full-text searchMeilisearch
   GeminiTypesenseMeilisearch

5. “Add site search to my documentation website. Give the exact install commands and the key code.”

   ClaudePagefindPagefind
   GPTPagefindPagefind
   GeminiPagefindPagefind

What the models wrote

t search-as-you-type (easily handling 100k+ docs).

### 1. Start Meilisearch (via Docker)
Run the search engine locally:
```bash
docker run -it --rm -p 7700:7700 getmeili/meilisearch:v1.7
```

### 2. Node.js Backend (Indexing the documents)
Install dependencies:
```bash
npm install express cors meilisearch
```

**`server.js`**
```javascript
const express = require('express');
const { MeiliSearch } = require('meilisea

n website so the static files exist (e.g., in a `dist`, `build`, or `public` folder). Then, run Pagefind to index that folder:

```bash
npx pagefind --site "build" # Replace "build" with your actual output directory
```

*Tip: Add this to your `package.json` build script so it runs automatically:*
```json
"scripts": {
  "build": "your-build-command && npx pagefind --site build"
}
```

### 2. Add the Search UI
Add thi

Questions