1# AI Search Configuration
2 
3## Worker Setup
4 
5```jsonc
6// wrangler.jsonc
7{
8  "ai": { "binding": "AI" }
9}
10```
11 
12```typescript
13interface Env {
14  AI: Ai;
15}
16 
17const answer = await env.AI.autorag("my-instance").aiSearch({
18  query: "How do I configure caching?",
19  model: "@cf/meta/llama-3.3-70b-instruct-fp8-fast"
20});
21```
22 
23## Data Sources
24 
25### R2 Bucket
26 
27Dashboard: AI Search → Create Instance → Select R2 bucket
28 
29**Supported formats:** `.md`, `.txt`, `.html`, `.pdf`, `.doc`, `.docx`, `.csv`, `.json`
30 
31**Auto-indexed metadata:** `filename`, `folder`, `timestamp`
32 
33### Website Crawler
34 
35Requirements:
36- Domain on Cloudflare
37- `sitemap.xml` at root
38- Bot protection must allow `CloudflareAISearch` user agent
39 
40## Path Filtering (R2)
41 
42```
43docs/**/*.md          # All .md in docs/ recursively
44**/*.draft.md         # Exclude (use in exclude patterns)
45```
46 
47## Indexing
48 
49- **Automatic:** Every 6 hours
50- **Force Sync:** Dashboard button (30s rate limit between syncs)
51- **Pause:** Settings → Pause Indexing (existing index remains searchable)
52 
53## Service API Token
54 
55Dashboard: AI Search → Instance → Use AI Search → API → Create Token
56 
57Permissions:
58- **Read** - search operations
59- **Edit** - instance management
60 
61Store securely:
62```bash
63wrangler secret put AI_SEARCH_TOKEN
64```
65 
66## Multi-Environment
67 
68```toml
69# wrangler.toml
70[env.production.vars]
71AI_SEARCH_INSTANCE = "prod-docs"
72 
73[env.staging.vars]
74AI_SEARCH_INSTANCE = "staging-docs"
75```
76 
77```typescript
78const answer = await env.AI.autorag(env.AI_SEARCH_INSTANCE).aiSearch({ query });
79```
80 
81## Monitoring
82 
83```typescript
84const instances = await env.AI.autorag("_").listInstances();
85console.log(instances.find(i => i.name === "docs"));
86```
87 
88Dashboard shows: files indexed, status, last index time, storage usage.
89