Runtime Model
indexbind deliberately focuses on retrieval, not content loading.
That means the runtime returns:
docIdrelativePathcanonicalUrltitlesummarymetadatascorebestMatch
Why Retrieval-Only
This keeps the runtime portable.
If the library tried to read content directly, it would need host-specific contracts for:
- local filesystem paths
- browser fetch locations
- Cloudflare asset routing
That would weaken the common API surface and make the web runtime much less clean.
Instead, the application takes the hit metadata and decides how to render, fetch, or navigate to the source content.
Runtime Split
indexbindnative SQLite query runtime for Nodeindexbind/buildNode build APIs for canonical bundles and incremental cache exportindexbind/webwasm-backed canonical bundle runtime for browsers and standard workersindexbind/cloudflarewasm-backed canonical bundle runtime for Cloudflare Workers