[Radar] Track Qwen Code export and dual-output session surfaces

[luoyuctl]

Background

Qwen Code is already listed as a supported local coding-agent session source in agenttrace. Recent official Qwen Code docs now expose additional machine-readable and shareable session surfaces: /export for session history and Dual Output for structured JSON event streams. This may affect parser coverage, docs clarity, and future interoperability, but there is not yet fixture-level evidence for agenttrace implementation work.

Evidence

• Qwen Code weekly update for 2026-02-09 says /export supports exporting session history to Markdown, JSONL, and HTML formats: https://qwenlm.github.io/qwen-code-docs/en/blog/weekly-update-2026-02-09/
• Qwen Code weekly update for 2026-05-14 says /export now defaults to HTML and mentions independently saved forked-session history plus session-list metadata loading changes: https://qwenlm.github.io/qwen-code-docs/en/blog/weekly-update-2026-05-14/
• Qwen Code Dual Output docs describe a structured JSON event stream and say session history is captured from --json-file as a canonical machine-readable transcript: https://qwenlm.github.io/qwen-code-docs/en/users/features/dual-output/
• Duplicate checks for Qwen Code session export JSONL HTML Markdown found no exact open issue in this repository.

User value

Users who work with Qwen Code sessions may benefit if agenttrace can clearly distinguish between verified local session history, exported conversation artifacts, and structured sidecar event streams when reporting cost, tool failures, health, and session evidence.

Adoption rationale

Tracking these surfaces helps keep parser coverage and Developer experience aligned with active Qwen Code workflows without claiming support for export or sidecar formats before real artifacts are validated.

Suggested scope

• Keep this as radar until a minimal Qwen Code /export artifact or --json-file Dual Output sample is available.
• Compare Qwen Code exported JSONL/JSON, HTML/Markdown, and sidecar event stream fields against existing Qwen Code parser assumptions.
• If promoted later, split into a narrow parser or docs task with fixture-backed acceptance checks.
• Clarify public docs only after validating which Qwen Code surfaces are actually supported by agenttrace.

Non-goals

• Do not add parser support without a local fixture or reproducible public sample.
• Do not claim support for Qwen Code export or Dual Output formats before validation.
• Do not parse private prompts, uploaded logs, or hosted telemetry.
• Do not create release, package, or external posting work from this radar item.

Acceptance criteria

• Maintainer or future Product review decides whether this remains radar, becomes a parser task, or becomes docs compatibility guidance.
• Any promoted task names the exact artifact type, command, and expected fields being validated.
• Parser work includes a minimal fixture or reproducible local sample.
• Public copy remains conservative and distinguishes verified Qwen Code support from unverified export or sidecar compatibility.

Suggested lane

lane/radar, priority/P2, status/needs-human

Risk

Medium. The official signal is concrete, but acting without artifacts could overstate compatibility or confuse existing Qwen Code parser support with export/sidecar support.

Source

source/radar: ecosystem scan on 2026-05-20. Tavily MCP was not exposed in this session, so Product used search fallback and prioritized official Qwen Code documentation.

[luoyuctl]

Event: Evidence
Actor: Product
Scope: #237
State change: none; keep lane/radar,status/needs-human
Evidence: Ecosystem scan on 2026-05-20 could not use Tavily because no Tavily MCP tool was exposed in this session, so Product used search fallback and rechecked Qwen Code export/Dual Output signals. Official docs still describe /export session-history export and Dual Output --json-file structured event streams, but this review did not find a local artifact, public fixture, or reproducible sample suitable for agenttrace parser work.
Product read: keep this as radar. If promoted later, require one concrete artifact type and command, such as a Qwen Code /export JSONL/HTML/Markdown file or a --json-file Dual Output transcript, then split parser/docs work with fixture-backed acceptance checks. Do not claim support for export or Dual Output formats from docs-only evidence.
Next owner: Human / Product monitor

[luoyuctl]

Event: Evidence
Actor: Product
Scope: #237
State change: none; keep lane/radar,status/needs-human
Evidence: Ecosystem scan on 2026-05-20 could not use Tavily because no Tavily MCP tool was exposed in this session, so Product used search fallback and prioritized official Qwen docs. Newer official docs strengthen the shape of the Qwen export/headless surface but still do not provide a local artifact or fixture for agenttrace parser work:

• Headless Mode documents JSON and stream-json output, including session/result message shapes, and states that project-scoped session data is JSONL under ~/.qwen/projects/<sanitized-cwd>/chats: https://qwenlm.github.io/qwen-code-docs/en/users/features/headless/
• Weekly update docs describe /export support for Markdown, JSONL, and HTML session history export: https://qwenlm.github.io/qwen-code-docs/en/blog/weekly-update-2026-02-09/
• Dual Output docs describe structured event streams through --json-file: https://qwenlm.github.io/qwen-code-docs/en/users/features/dual-output/

Product read: this is stronger adoption rationale for tracking Qwen export/headless/dual-output surfaces, but it is still docs-only. Keep this radar parked until a Qwen /export file, a --json-file Dual Output sample, or a local ~/.qwen/projects/.../chats artifact is available for fixture-backed acceptance checks. Do not claim support for these export/dual-output formats from documentation alone.