feat: stream SSE tokens in-flight with delta content fallback#6
Merged
Conversation
Zero-copy token accounting for streaming responses: chunks are processed as they arrive via SSEAccumulator rather than buffering the full response body. - Add SSEAccumulator that parses SSE lines byte-by-byte across chunk boundaries, extracting usage fields or summing delta.content bytes as fallback - Update budgetTrackingBody to route streaming reads through SSEAccumulator instead of the buffer, preserving zero-copy semantics - Fall back to delta content byte estimation when no usage field is present (more accurate than full-body estimation for SSE) - Add test-streaming-tokens.sh integration script covering: streaming with usage field, streaming without usage field, and non-streaming baseline
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Zero-copy token accounting for streaming responses: chunks are
processed as they arrive via SSEAccumulator rather than buffering
the full response body.
Changes
SSEAccumulatorininternal/pricing/streaming.gothat processes SSE response chunks byte-by-byte as they arrive, building lines across chunk boundaries without buffering the full response bodybudgetTrackingBodyinproxy.goto route streaming reads throughSSEAccumulatorinstead ofb.buffer, preserving zero-copy semantics while tracking costs in-flightusagefield in final SSE chunk (exact) → sum ofchoices[0].delta.contentbytes (estimated, output only) → body-length estimation fallback (existing)data:, JSON structure, and newline overhead that inflated previous estimatesscripts/dev/test-streaming-tokens.shintegration script covering three cases: streaming with usage field, streaming without usage field (delta fallback), and non-streaming baselineTesting
go test ./...passesgo vet ./...passesbash scripts/dev/test-streaming-tokens.sh— all 3 tests pass andspent_usd > 0for each agentpricing/streaming_test.go: usage field extraction, delta content fallback, malformed chunks, partial lines spanning chunk boundariespricing_test.gonon-streaming path unchangedChecklist
feat:,fix:,docs:, etc.)