Library

csp is usable as a library two ways: the Rust crate, or a JavaScript/TypeScript SDK (@pleaseai/csp-sdk) that binds the same core natively via napi-rs.

Rust

Published on crates.io as code-search-please (the short name csp was already taken). The library name stays csp, so you depend on code-search-please but still write use csp::.... It exposes CspIndex with from_path / from_git / search / find_related, plus the ContentType enum and the ranking pipeline.

The crate ships the csp CLI too (behind the default-on cli feature). As a library dependency, disable default features to skip the CLI stack (clap / tokio / rmcp):

[dependencies]
code-search-please = { version = "0.1", default-features = false }

use std::path::Path;
use csp::indexing::index::{CspIndex, LoadOptions, QueryOptions};

// Index a local directory and search it
let index = CspIndex::from_path(Path::new("./my-project"), &LoadOptions::default())?;
let results = index.search("save model to disk", &QueryOptions { top_k: Some(3), ..Default::default() });

for r in &results {
    println!("{}:{}-{}", r.chunk.file_path, r.chunk.start_line, r.chunk.end_line);
}

JavaScript / TypeScript

@pleaseai/csp-sdk is a native (napi-rs) addon that runs the same Rust search engine in-process: no subprocess, no JSON round-trip. The build entrypoints are async; the per-query calls are sync.

bun add @pleaseai/csp-sdk

import { ContentType, CspIndex } from '@pleaseai/csp-sdk'

const index = await CspIndex.fromPath('./my-project', { content: [ContentType.Code] })
const results = index.search('save model to disk', { topK: 3 })

for (const { chunk, score } of results) {
  console.log(score.toFixed(3), chunk.location) // e.g. 0.871 src/index.ts:42-58
}

CspIndex exposes fromPath, fromGit, and loadFromDisk (async), plus search, findRelated, save, and stats (sync). Chunk fields are camel-cased at the JS boundary: chunk.filePath, chunk.startLine, chunk.endLine.