Guard streamed tool terminals and retain Usage #46

Merged

weichen merged 9 commits from feature/fast-slow-u3-r04-stream-guard into main 2026-10-19 09:20:09 +00:00

Conversation 0 Commits 9 Files Changed 10

+1149 -54

9 Commits

Author	SHA1	Message	Date
agent-exec-transport	28f3ea7674	agent-exec transport d7776a79d43bd8fd44e6c0d78873761b/9585ecb5ab4b	2026-10-19 03:50:46 +00:00
agent-exec-transport	b6a843337c	agent-exec transport 4838630332e6bb1220f0e63e947ddb81/599831d96200	2026-10-19 03:02:20 +00:00
agent-exec-transport	5b21325da7	agent-exec transport ed24e191762d6f6e3b41589f552e7744/cd1765c54231	2026-10-19 02:53:01 +00:00
weichen	42b22b436b	Merge Agent-Native main into the U3 stream guard branch	2026-10-19 02:46:12 +00:00
weichen	6a1dcddb48	Record content-free stream terminal observations ... Why: The r0.4 trial could prove that the parser missed [DONE] but could not distinguish clean EOF, malformed frames, read failure, or provider error without retaining raw SSE. Evidence: RED receipts fast-slow-u3-r05-task1-terminal-red.json and task1-http-error-red.json; GREEN targeted receipt and full vet/test/race receipt fast-slow-u3-r05-task1-agent-native-full.json. Impact: Add bounded scalar diagnostics and successful HTTP status to terminal events while preserving the existing incomplete-tool rejection, returned Usage, and retry behavior. Refs: U3 r0.5 approved design and plan; Agent-Native PR #46	2026-10-14 07:50:39 +00:00
weichen	a5e3c7c817	Reject legacy empty function-call terminals ... Why: The parser also recognizes function_call as a tool finish, but cannot execute a legacy call when no deltas were collected. It must fail closed like tool_calls. Evidence: The function_call variant failed before this guard and passed afterward; full Agent-Native vet/test passed. Impact: Inconsistent legacy tool terminals fail with returned Usage preserved; valid streams are unchanged. Refs: U3 r0.4 independent review Important 2 closure	2026-10-14 04:12:04 +00:00
weichen	6e972335c9	Reject empty tool-call stream terminals ... Why: A provider can report finish_reason=tool_calls without emitting any tool call. Treating it as a successful empty answer hides an incomplete operation. Evidence: New SSE regression failed before the guard and passed afterward; full go vet and go test -count=1 ./... passed. Impact: Inconsistent tool terminals fail closed while preserving returned Usage; valid text and tool streams are unchanged. Refs: U3 r0.4 independent review issue 2	2026-10-14 03:49:06 +00:00
weichen	e86e74bc36	Preserve streamed tool delta identity ... Why: Backend emitted tool deltas from a separate goroutine, racing with serialized run events and occasionally losing a UI delta. Evidence: Backend race detector reproduced concurrent event writes; Agent-Native identity test RED→GREEN; full vet and tests pass. Impact: runtime tool delta events now carry provider slot and call ID for single-owner downstream projection. Refs: runtime/loop.go; runtime/event_vocabulary_test.go	2026-10-14 02:22:34 +00:00
weichen	94afd71f00	Reject incomplete streamed tool calls ... Why: a model tool finish could execute partial JSON or an unterminated SSE stream, while post-Usage errors could lose their charge. Evidence: missing-DONE, malformed-JSON, identity, non-object and post-Usage RED→GREEN tests; GOWORK=off go vet ./... and go test -count=1 ./... passed. Impact: invalid tool terminals fail without execution or automatic retry and retain returned Usage; valid text and multi-part tool streams keep working. Refs: runtime/llm/sse.go; runtime/loop.go	2026-10-14 02:09:44 +00:00