Python: Align DevUI Responses parsing

[Eduard van Valkenburg (eavanvalkenburg)]

Motivation & Context

DevUI's local Responses API conversion could flatten multiple input messages into one user message, drop conversation content after the first part, and replace unsupported input with an empty user text message. Completion conversion also estimated token usage from character counts and generated unrelated response IDs during one streamed lifecycle. These mismatches make local agent behavior diverge from the OpenAI Responses contract and lose information needed by multimodal and conversation scenarios.

Description & Review Guide

• What are the major changes? Preserve supported input message roles, boundaries, text/image/file parts, hosted file IDs, image detail, and straightforward canonical MIME types across executor and conversation conversion. Reject unsupported-only message input explicitly. Reuse the server tracking ID through lifecycle and final aggregation, and omit usage when the framework did not report it.
• What is the impact of these changes? Local DevUI execution and conversation APIs retain supported Responses input semantics instead of silently flattening or fabricating content. Completed responses expose only measured usage, and streamed response IDs remain stable.
• What do you want reviewers to focus on? Please focus on the deliberately narrow supported-part boundary: message text, images, and files are preserved without adding a shared codec or expanding support to unrelated top-level Responses item types.

Related Issue

N/A — follow-up from the Responses conversion consistency sweep

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Pull request overview

Aligns DevUI’s local Responses conversion with OpenAI semantics while preserving multimodal content and response metadata.

Changes:

• Preserves message roles, boundaries, images, files, MIME types, and hosted IDs.
• Stabilizes response IDs and removes estimated token usage.
• Rejects unsupported-only inputs and expands conversion tests.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
_executor.py	Expands Responses input conversion.
_conversations.py	Preserves supported conversation parts.
_mapper.py	Stabilizes IDs and usage handling.
test_multimodal_workflow.py	Tests multimodal input conversion.
test_mapper.py	Tests IDs and missing usage.
test_conversations.py	Tests conversation round-tripping.
test_approval_validation.py	Updates rejection expectations.

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
TOTAL	48404	4503	90%

report-only-changed-files is enabled. No files were changed during this commit :)

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
9858	36 💤	0 ❌	0 🔥	2m 44s ⏱️

[github-actions]

MAF Automated Review — Iteration 1

Result: Findings reported
Scope: full PR (1 commit(s)): 3294e694cb22
Model: gpt-5.6-sol

Overview

The change substantially improves preservation of Responses message boundaries, multimodal content, stable response IDs, and measured usage, with focused regression coverage for the intended conversions. Validation of roles, server-side approval reconstruction, atomic conversation mutation, and mapper cleanup provide useful guardrails. Two lifecycle gaps remain: failed streams are finalized with a contradictory completion event, and unsupported-only messages can still disappear when another valid message is present.

Reviewed the supplied pull-request change set across correctness, security/reliability, architecture, and failure behavior.
2 verified findings remained after source verification (1 high, 1 medium) across 2 files. Details are attached to the affected lines below.

Affected areas: python/packages/devui/agent_framework_devui/_executor.py, python/packages/devui/agent_framework_devui/_mapper.py

[github-actions]

MAF Automated Review — Iteration 2

Result: Findings reported
Scope: 1 net-new commit(s): 5dfaef081c2a
Model: gpt-5.6-sol

Overview

This iteration correctly recognizes assistant text variants, rejects each unsupported-only message, preserves partial output on failed streams, and emits one terminal event with a stable response ID. The accompanying tests cover the principal message and stream regressions. One approval-lifecycle edge case remains: validating a batch can consume a pending approval before a later message causes conversion to fail.

Reviewed the supplied incremental change set across correctness, security/reliability, architecture, and failure behavior.
1 verified finding remained after source verification (1 medium) across 1 file. Details are attached to the affected lines below.

Affected areas: python/packages/devui/agent_framework_devui/_executor.py