Struct ResponseAccumulator 

pub struct ResponseAccumulator { /* private fields */ }

Accumulates LLM response chunks from streaming or non-streaming sources.

Implementations

impl ResponseAccumulator

pub fn new(response_id: String, conversation_id: Option<String>) -> Self

Creates a new response accumulator.

pub fn from_json( body: &str, conversation_id: Option<&str>, ) -> ExecutorResult<Self>

Parses a non-streaming JSON response body.

Errors

Returns ExecutorError::ParseError if JSON parsing fails or required fields are missing.

pub async fn from_stream( stream: Pin<Box<dyn Stream<Item = Result<String, ExecutorError>> + Send>>, conversation_id: Option<&str>, ) -> ExecutorResult<Self>

Accumulates an async stream of raw SSE lines with parallel processing.

The async task feeds raw SSE lines through a channel while a spawn_blocking worker handles JSON parsing on a blocking thread — keeping the tokio executor free between chunk arrivals.

Errors

Returns ExecutorError::ParseError if chunk parsing fails, or ExecutorError::StreamError if the stream or worker encounters an error.

pub fn from_sse_lines( lines: impl IntoIterator<Item = String>, conversation_id: Option<&str>, ) -> Self

Processes pre-collected raw SSE lines synchronously.

Useful when lines have already been buffered (e.g. replaying a recorded stream). Prefer from_stream for live async streams. Line parse errors are silently skipped — this function is infallible.

pub fn mark_incomplete(&mut self, reason: impl Into<String>)

Marks the response as incomplete due to an error or interruption.

pub fn finalize( self, model: &str, previous_response_id: Option<&str>, instructions: Option<&str>, ) -> ResponsePayload

Finalizes the accumulator into a ResponsePayload.

The caller supplies fields that come from the original request, not from the LLM response stream.

Trait Implementations

impl Debug for ResponseAccumulator

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.