Search / indexation

Basic document indexation lives in serenade-search (#129, #255).

Serenade ships contracts, a zero-deps in-memory adapter, and (feature http) an HTTP stub toward Meilisearch/ES-class REST APIs. It does not embed Lucene, Algolia, Meilisearch, or Postgres FTS. Applications that outgrow the memory index implement their own adapters (or use the HTTP stub with their client).

Pieces

Piece	Role
SearchDocument	Stable id + named text fields
SearchQuery / SearchHit	Free-text query, optional limit/offset, ranked hit
DocumentIndex	upsert / delete / query / clear
MemorySearchAdapter	Case-insensitive whitespace-token AND match; score = occurrence sum
HttpSearchAdapter (feature http)	REST stub; apps own the HTTP client via SearchHttpPoster

Matching on the memory adapter concatenates field values. Empty / whitespace-only queries return no hits.

HTTP stub (feature http)

Enable with serenade-search feature http. Serenade shapes a small REST surface; apps own the network client (no mandatory SaaS SDK).

Piece	Role
HttpSearchConfig	Base URL + optional Bearer API key
SearchHttpPoster / MockSearchHttpPoster	Sync request trait + test double
HttpSearchAdapter	Implements DocumentIndex over HTTP JSON

Endpoints (relative to base URL):

Op	Method	Path	Body / response
upsert	PUT	/documents/{id}	{ "id", "fields" }
delete	DELETE	/documents/{id}	optional { "deleted": bool } (default true on 2xx)
query	POST	/search	request { "q", "limit?", "offset" }; response { "hits": [{ "id", "score?" }] }
clear	DELETE	/documents	empty

use serenade_search::{
    DocumentIndex, HttpSearchAdapter, HttpSearchConfig, MockSearchHttpPoster, SearchDocument,
    SearchQuery,
};

let poster = MockSearchHttpPoster::ok_empty();
let index = HttpSearchAdapter::new(
    HttpSearchConfig::new("https://search.example").api_key("tok"),
    poster,
);
index
    .upsert(SearchDocument::new("1").field("body", "hello"))
    .expect("upsert");

Ownership

Concern	Owner
Traits + memory adapter	Serenade (serenade-search)
HTTP stub shapes (feature http)	Serenade
When to index domain entities	Application
SaaS / engine clients	Application (or plug into SearchHttpPoster)
FrameworkBundle SaaS wiring	None

MyFeed dogfood

examples/demo-html keeps a MemorySearchAdapter beside SQLite:

• rebuild on boot from stored posts
• upsert on create / admin edit
• delete on admin delete
• GET /search?q= returns escaped HTML results

Apps that outgrow the memory index can plug HttpSearchAdapter (feature http) or their own DocumentIndex; indexing policy stays application-owned.

Example

use serenade_search::{DocumentIndex, MemorySearchAdapter, SearchDocument, SearchQuery};

let index = MemorySearchAdapter::new();
index
    .upsert(SearchDocument::new("1").field("body", "hello serenade"))
    .expect("upsert");
let hits = index.query(&SearchQuery::new("serenade")).expect("query");
assert_eq!(hits[0].id(), "1");