Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
Python agent-contextagent-memoryagent-pluginsagentic-ragcontext-databasedsh-pluginknowledge-basememorymemory-managementself-evolvingvector-database
//! RefStore trait - named references to Git objects
use async_trait::async_trait;
use gix_hash::ObjectId;
use crate::git::error::RefStoreError;
/// Storage for named references (branches, tags) pointing to Git objects.
///
/// The core operation is `cas_update` (Compare-And-Swap), which ensures
/// atomicity and consistency in the presence of concurrent writers.
#[async_trait]
pub trait RefStore: Send + Sync + 'static {
/// Read the current value of a ref.
///
/// Returns `NotFound` if the ref doesn't exist.
async fn read(&self, account: &str, ref_name: &str) -> Result<ObjectId, RefStoreError>;
/// Compare-And-Swap update: write `new` only if current value == `expected`.
///
/// - `expected = None` means "write only if ref doesn't exist yet"
/// - `expected = Some(oid)` means "write only if ref currently has this oid"
///
/// Returns `Conflict` if the expectation fails.
async fn cas_update(
&self,
account: &str,
ref_name: &str,
expected: Option<ObjectId>,
new: ObjectId,
) -> Result<(), RefStoreError>;
/// List all refs under a prefix (e.g., "refs/heads/").
///
/// Returns a list of (full_ref_name, target_oid) pairs.
async fn list(
&self,
account: &str,
prefix: &str,
) -> Result<Vec<(String, ObjectId)>, RefStoreError>;
}