streaming: pre-assign message_id, add block_index and seq

[TirmanSidhu]

Summary

• Assigns assistant message_id on the first delta of a turn (was: at row persist).
• Adds per-conversation seq counter, stamped on every streaming event.
• Adds block_index to all text/tool events.
• Emits new message_open / block_open / block_close / message_close events alongside legacy message_complete (additive — no breaking change).

Part of plan: streaming-message-architecture.md (PR 1 of 6)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fc67ffef94

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Keep tool-use events on the persisted assistant message

For any assistant response that includes tool calls, AgentLoop.run() dispatches message_complete before it dispatches the corresponding tool_use events (see assistant/src/agent/loop.ts around the onEvent({ type: "message_complete" ... }) call followed by the for (const toolUse of toolUseBlocks) loop). Resetting currentMessageId here therefore closes the persisted assistant row before its tool blocks are streamed; the later handleToolUse allocates a new message id, so tool_use_start/tool_result are attached to the wrong message (often the follow-up text row) and their blockIndex no longer matches the persisted content. Tool-using turns will corrupt the new streaming protocol's message/block identity even though legacy events still arrive.

Useful? React with 👍 / 👎.