feat: #2886 add tool_name and call_id properties to ToolCallItem and ToolCallOutputItem

[SuperMarioYL]

Summary

ToolApprovalItem already exposes tool_name and call_id convenience properties, but ToolCallItem and ToolCallOutputItem force callers to write fragile getattr chains or a two-pass call_id join to correlate outputs back to their originating calls.

This PR adds the missing accessors:

• ToolCallItem.tool_name — extracts .name / "name" from the raw item. Returns None for hosted tool types that carry no name field (e.g. computer-use, file-search).
• ToolCallItem.call_id — extracts .call_id / .id / "call_id" / "id", mirroring the logic already in ToolApprovalItem._extract_call_id.
• ToolCallOutputItem.call_id — extracts "call_id" from the raw dict payload, matching the call_id on the corresponding ToolCallItem so callers can join without a second pass.

Before this change, getting all outputs for a named tool required a verbose two-pass join:

# Before — manual join
call_id_to_name = {}
for item in result.new_items:
    if isinstance(item, ToolCallItem):
        name = getattr(item.raw_item, "name", None)
        call_id = getattr(item.raw_item, "call_id", None)
        if name and call_id:
            call_id_to_name[call_id] = name

for item in result.new_items:
    if isinstance(item, ToolCallOutputItem):
        call_id = item.raw_item.get("call_id") if isinstance(item.raw_item, dict) else None
        if call_id_to_name.get(call_id) == "my_tool":
            print(item.output)

After:

# After — direct property access
calls = {item.call_id: item.tool_name for item in result.new_items if isinstance(item, ToolCallItem)}
for item in result.new_items:
    if isinstance(item, ToolCallOutputItem) and calls.get(item.call_id) == "my_tool":
        print(item.output)

Test plan

Added 8 unit tests in tests/test_items_helpers.py covering:

• tool_name from ResponseFunctionToolCall (typed object)
• tool_name from a raw dict
• tool_name returns None for ResponseComputerToolCall (no name field)
• call_id from ResponseFunctionToolCall
• call_id from a raw dict
• ToolCallOutputItem.call_id from a dict payload
• ToolCallOutputItem.call_id returns None when absent
• End-to-end join of ToolCallItem + ToolCallOutputItem via call_id

Verified locally:

• uv run pytest tests/test_items_helpers.py tests/test_run_internal_items.py — 53 passed
• uv run make lint — all checks passed
• uv run pyright src/agents/items.py — 0 errors

Issue number

Closes #2886

Checks

• I've added new tests (if relevant)
• I've added/updated the relevant documentation
• I've run make lint and make format
• I've made sure tests pass

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3d18198481

ℹ️ About Codex in GitHub

Your team has set up Codex to 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 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Support id fallback in ToolCallOutputItem.call_id

ToolCallOutputItem.call_id only reads raw_item["call_id"], but the SDK already treats id as a fallback identifier for tool payloads (for example in extract_tool_call_id and output indexing logic), so id-only provider payloads are intentionally handled elsewhere. In that supported scenario, ToolCallItem.call_id resolves to id while ToolCallOutputItem.call_id resolves to None, which breaks the new convenience join path and contradicts the property’s stated correlation behavior.

Useful? React with 👍 / 👎.