Struct BatchUploader 

pub struct BatchUploader {
    pub buffer: Arc<Mutex<Vec<Sample>>>,
    shutdown: Arc<AtomicBool>,
    upload_interval_secs: u64,
    sample_interval_secs: u64,
}

Fields

buffer: Arc<Mutex<Vec<Sample>>>

Buffer shared with the main thread.

shutdown: Arc<AtomicBool>

Set to true by request_shutdown() to trigger a final flush.

upload_interval_secs: u64

Polling interval for the upload thread (seconds, default 60).

sample_interval_secs: u64

Sampling interval (seconds) – needed to compute per-interval byte counts in CSV.

Implementations

impl BatchUploader

pub fn new( upload_interval_secs: u64, sample_interval_secs: u64, ) -> (Self, Arc<Mutex<Vec<Sample>>>)

Create a new BatchUploader and return the shared SampleBuffer so the main thread can push samples into it.

pub fn shutdown_flag(&self) -> Arc<AtomicBool>

Clone the shutdown flag so main.rs can signal the upload thread to flush and exit after moving self into the spawned thread.

pub fn spawn( self, ctx: Arc<Mutex<RunContext>>, agent: Agent, api_base: String, token: String, ) -> JoinHandle<Vec<String>>

Spawn the background upload thread.

The thread wakes every upload_interval_secs, drains the buffer, builds a .csv.gz batch (gzip-compressed CSV, Content-Type: application/gzip), and uploads it to S3. On shutdown signal it performs one final flush before exiting.

Returns a JoinHandle<Vec<String>> whose value is the list of all successfully uploaded S3 URIs (e.g. "s3://bucket/prefix/run-id/000000.csv.gz"). The caller uses this list to decide the /finish route:

• non-empty → data_source: "s3" with data_uris
• empty → data_source: "inline" with data_csv