fix(ai): route response events by output index - #45013

Summary

• track output slots established by response.output_item.added and route subsequent text, refusal, reasoning, and function-argument events by authoritative output_index
• retain existing item_id routing for compatible providers that omit output indexes and preserve validation of missing required item IDs
• apply the same output-slot routing to OpenAI-specific raw reasoning deltas, generic Responses-compatible providers, and xAI

Before and after

Responses events identify an output slot with output_index; subsequent events also include item_id, but the official accumulator routes through the output slot:

{ "type": "response.output_item.added", "output_index": 1, "item": { "type": "function_call", "id": "fc_1", "call_id": "call_1", "name": "first" } }
{ "type": "response.output_item.added", "output_index": 3, "item": { "type": "function_call", "id": "fc_2", "call_id": "call_2", "name": "second" } }
{ "type": "response.function_call_arguments.delta", "output_index": 1, "item_id": "fc_2", "delta": "{\"a\":" }
{ "type": "response.function_call_arguments.delta", "output_index": 3, "item_id": "fc_1", "delta": "{\"b\":" }

Before: OpenCode ignored output_index, so the argument fragments were associated with the wrong tool calls.

After: Registered output slots are authoritative: the first fragment belongs to call_1 and the second belongs to call_2, matching the official Responses accumulator. Streams without registered output indexes continue using their existing item_id behavior.

Verification

• bun test test/partial-json.test.ts test/tool-stream.test.ts test/provider/openai-responses.test.ts test/provider/openai-compatible-responses.test.ts test/provider/openai-responses-websocket.recorded.test.ts test/provider/openai-chat.test.ts test/provider/anthropic-messages.test.ts test/provider/bedrock-converse.test.ts test/provider/xai-responses.test.ts test/response.test.ts test/tool-runtime.test.ts --timeout 30000 --only-failures (363 passing)
• RECORDED_PREFIX=openai-responses bun test test/provider/golden.recorded.test.ts --timeout 30000 --only-failures
• bun typecheck from packages/ai
• bunx prettier --check packages/ai/src/protocols/open-responses.ts packages/ai/src/protocols/openai-responses.ts packages/ai/test/provider/openai-responses.test.ts packages/ai/test/provider/openai-compatible-responses.test.ts packages/ai/test/provider/xai-responses.test.ts