Struct ThreadStore 

pub struct ThreadStore { /* private fields */ }

Disk-backed store for threads, messages, and runs.

All methods are synchronous and are intended to be called from within a tokio::task::spawn_blocking context (see the route handlers and worker).

Implementations

impl ThreadStore

pub fn new(dir: PathBuf) -> ServerResult<Self>

Open (or create) the thread store root directory.

Creates the directory and any missing parents if they do not exist.

pub fn create_thread(&self, thread: &Thread) -> ServerResult<()>

Persist a new thread to disk.

Creates {root}/{thread_id}/meta.json atomically. Fails if the directory already exists (duplicate ID).

pub fn get_thread(&self, id: &str) -> ServerResult<Thread>

Read a thread’s metadata from disk.

Returns ServerError::ThreadNotFound if no directory or meta.json exists for the given ID.

pub fn list_thread_ids(&self) -> ServerResult<Vec<String>>

List all thread IDs stored in the root directory.

pub fn append_message( &self, thread_id: &str, msg: &ThreadMessage, ) -> ServerResult<()>

Append a single message to the thread’s messages.jsonl file.

Appends are not atomic at the OS level (no fsync fence), but each line is a complete JSON object, so a reader will never observe a partially-written message — incomplete trailing lines are filtered out by list_messages.

pub fn list_messages(&self, thread_id: &str) -> ServerResult<Vec<ThreadMessage>>

Read all messages for a thread in append order (oldest first).

Blank lines and lines that fail to parse as JSON are silently skipped so that a partial write at the end of a previous session does not break future reads.

pub fn create_run(&self, thread_id: &str, run: &Run) -> ServerResult<()>

Persist a new run to disk.

Creates {root}/{thread_id}/runs/{run_id}/status.json atomically. Returns ThreadNotFound if the thread does not exist.

pub fn get_run(&self, thread_id: &str, run_id: &str) -> ServerResult<Run>

Read a run’s status from disk.

Returns RunNotFound if no status.json exists for the given IDs.

pub fn update_run_status( &self, thread_id: &str, run_id: &str, status: RunStatus, error: Option<RunError>, ) -> ServerResult<()>

Atomically update a run’s status (and optionally set last_error).

Returns RunNotFound if the run does not exist, or RunInTerminalState if the run is already in a terminal state.

pub fn force_update_run_status( &self, thread_id: &str, run_id: &str, status: RunStatus, error: Option<RunError>, ) -> ServerResult<()>

Force-update a run’s status bypassing terminal-state guard.

Used by the cancel handler to transition a queued/in-progress run to Cancelled even if no worker is currently processing it.

pub fn steps_dir(&self, thread_id: &str, run_id: &str) -> PathBuf

Return the path to a run’s steps directory.

pub fn append_step( &self, thread_id: &str, run_id: &str, step: &RunStep, ) -> ServerResult<()>

Persist a new run step to disk.

Creates {root}/{thread_id}/runs/{run_id}/steps/{step_id}.json atomically. Returns RunNotFound if the run does not exist.

pub fn list_steps( &self, thread_id: &str, run_id: &str, ) -> ServerResult<Vec<RunStep>>

Read all steps for a run, sorted by created_at ascending.

Returns RunNotFound if the run does not exist.

pub fn get_step( &self, thread_id: &str, run_id: &str, step_id: &str, ) -> ServerResult<RunStep>

Read a single run step by ID.

Returns RunStepNotFound if no step with this ID exists.

pub fn update_step_status( &self, thread_id: &str, run_id: &str, step_id: &str, status: RunStepStatus, ) -> ServerResult<()>

Atomically update a run step’s status (and optionally timestamps/error).

Returns RunStepNotFound if no step with this ID exists.

pub fn thread_dir(&self, thread_id: &str) -> PathBuf

Return the path to a thread’s subdirectory.

pub fn run_dir(&self, thread_id: &str, run_id: &str) -> PathBuf

Return the path to a run’s subdirectory.