fix(chat-completions): coerce integer tool_call id in non-streaming normalize_response

[briandevans]

This is a sibling follow-up to #58502 (commit 55e7986)

• fix(providers): aux pool fallback, tool-use 404 classification, poolside wire types, codex text recovery (4-PR salvage) #58502 coerced Poolside's integer wire types: finish_reason in normalize_response, and tool_call.id in the streaming path (chat_completion_helpers.py).
• fix(providers): aux pool fallback, tool-use 404 classification, poolside wire types, codex text recovery (4-PR salvage) #58502 did NOT touch the non-streaming normalize_response tool_call.id at agent/transports/chat_completions.py:639 — the same integer-id shape reaches it there.
• This PR adds the same int → str coercion for the non-streaming path, mirroring the finish_reason coercion 25 lines above.

What does this PR do?

Coerces an integer tool_call.id to str in ChatCompletionsTransport.normalize_response so a Poolside tool turn on the non-streaming path (stream-unsupported fallback via agent._disable_streaming, ACP/MoA-no-consumer callers, gateway/health-check probes) does not persist an int id.

Left uncoerced, the assistant message's tool_calls[].id becomes a synthetic _deterministic_call_id string (because _split_responses_tool_id guards isinstance(raw_id, str) and falls through on a non-str id), while the paired tool result stores the raw int via make_tool_result_message. The two diverge, and the next replay turn is rejected by the provider with HTTP 400 ("tool_call_id did not match a preceding tool_calls"). Poolside is a first-class CANONICAL_PROVIDERS entry (CLI/TUI/desktop picker), so this path is user-reachable.

Related Issue

Sibling follow-up to #58502 (commit 55e7986). No separate issue.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• agent/transports/chat_completions.py: in normalize_response, coerce an integer tool_call.id to str before constructing ToolCall, mirroring the existing integer finish_reason coercion.
• tests/agent/transports/test_chat_completions.py: add test_tool_call_integer_id_coerced_to_string regression test.

How to Test

1. uv run --with pytest --with pytest-asyncio python3 -m pytest tests/agent/transports/test_chat_completions.py -q → 83 passed.
2. Fail-before / pass-after: the new test asserts nr.tool_calls[0].id == "24" and isinstance(... , str). With the coercion reverted it fails (assert 24 == '24'); with the fix it passes.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/agent/transports/test_chat_completions.py -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 15 (py3.9 parse-safety verified via ast.parse(feature_version=(3,9)))

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A (pure Python type coercion)
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Sibling code paths that may need the same fix: the streaming path already coerces this (chat_completion_helpers.py, #58502); no other normalize path constructs a ToolCall from a raw provider id. No further widening needed.

[Copilot]

Pull request overview

This PR fixes a provider-boundary type mismatch in the non-streaming chat-completions transport by coercing tool_call.id from int to str during response normalization, preventing tool-call ID divergence between assistant tool calls and subsequent tool-result messages (which can trigger HTTP 400 replay failures).

Changes:

• Coerce non-streaming tool_call.id values from int → str in ChatCompletionsTransport.normalize_response.
• Add a regression test covering Poolside’s integer tool_call.id wire shape on the non-streaming path.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
agent/transports/chat_completions.py	Coerces integer tool_call.id to str during normalization to keep tool-call IDs consistent for replay.
tests/agent/transports/test_chat_completions.py	Adds a regression test asserting tool_call.id is normalized to a string when provided as an integer.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[teknium1]

Thanks for the focused follow-up. The premise remains valid on current main: ChatCompletionsTransport.normalize_response() still passes raw tc.id through at agent/transports/chat_completions.py:691, although ToolCall.id is canonically str | None (agent/transports/types.py:22-35).

The proposed coercion matches the already-shipped streaming handling in agent/chat_completion_helpers.py:2472-2488. It also prevents the verified downstream mismatch: assistant-message construction replaces non-string IDs with deterministic strings (agent/chat_completion_helpers.py:1247-1260), while tool execution writes tool_call.id directly into the result message (agent/tool_executor.py:1666). The focused regression test covers the non-streaming transport boundary.

The target block and neighboring test remain present on current main; this should be straightforward to salvage despite the stale branch.

Automated hermes-sweeper review.