basemind 0.2.1

Full AI context layer over MCP — tree-sitter code-map, document RAG (PDF/Office/HTML/email + OCR + reranker), shared agent memory, on-demand web crawl, git history + blame + per-symbol diff. 300+ languages, 8 coding-agent harnesses, content-addressed Fjall + LanceDB.
// Prototype: rustdoc lives in the code itself and in README.md rather than per-symbol docstrings.
// Flip this off once the public API is frozen.
#![allow(missing_docs)]

pub mod config;
#[cfg(feature = "intelligence")]
pub mod embeddings;
pub mod extract;
pub mod git;
pub mod git_cache;
pub mod hashing;
pub mod index;
#[cfg(feature = "intelligence")]
pub mod lance;
pub mod lang;
pub mod mcp;
pub mod path;
pub mod query;
pub mod render;
pub mod scanner;
#[cfg(feature = "documents")]
pub mod scanner_docs;
pub mod store;
#[cfg(feature = "crawl")]
pub mod url;
pub mod version;
pub mod watcher;
#[cfg(feature = "crawl")]
pub mod web;

pub use config::Config;

/// Test-only helpers exposed from the library so integration tests can mint cursors
/// without re-implementing the base64url + msgpack encoding. Not part of the stable API.
#[doc(hidden)]
pub mod testing {
    /// Build an in-memory cursor with the given `(offset, snapshot_id)`, returning the
    /// opaque base64url string an MCP client would receive in `next_cursor`. Used by
    /// the smoke tests to forge stale cursors and verify `cursor_invalidated` plumbing.
    pub fn encode_in_memory_cursor(offset: u64, snapshot_id: u32) -> String {
        crate::mcp::cursor::Cursor::encode_in_memory(offset, snapshot_id).0
    }
}