fix(cli): keep AGENTS.md enabled by default context reset by zy6p · Pull Request #2082 · QwenLM/qwen-code

zy6p · 2026-03-04T08:22:43Z

TLDR

Fix default context filename reset so AGENTS.md remains effective when settings.context.fileName is not configured.

Dive Deeper

Root cause:

In loadCliConfig, when settings.context.fileName is absent, we called:
- setGeminiMdFilename(getCurrentGeminiMdFilename())
getCurrentGeminiMdFilename() returns only the first configured file name.
Default state is [QWEN.md, AGENTS.md], but this code collapsed it into only QWEN.md, unintentionally disabling default AGENTS.md loading.

What changed:

packages/cli/src/config/config.ts
- reset defaults using both filenames explicitly:
  - setGeminiMdFilename([DEFAULT_CONTEXT_FILENAME, AGENT_CONTEXT_FILENAME])
packages/cli/src/config/config.test.ts
- added regression test: default path resets to both QWEN.md and AGENTS.md
- added test: custom settings.context.fileName still overrides as expected

Reviewer Test Plan

Run without context.fileName in settings.
Verify setGeminiMdFilename receives both default names (QWEN.md, AGENTS.md).
Set context.fileName = "CUSTOM_AGENTS.md" and verify custom filename is used.
(Optional behavior check) Create AGENTS.md in project and ensure it is discovered in default configuration.

Testing Matrix

	🍏	🪟	🐧
npm run	❓	❓	✅
npx	❓	❓	❓
Docker	❓	❓	❓
Podman	❓	-	-
Seatbelt	❓	-	-

Linked issues / bugs

Fixes #2062

LaZzyMan

Great catch on the root cause! The analysis is spot-on — getCurrentGeminiMdFilename() collapses the array into a single string.

A couple of suggestions:

1. Use getAllGeminiMdFilenames() instead of hardcoding the default array

The current fix hardcodes [DEFAULT_CONTEXT_FILENAME, AGENT_CONTEXT_FILENAME], which duplicates the default value defined in memoryTool.ts. If additional context filenames are added in the future, this code won't automatically pick them up. A more robust approach:

import {
  getAllGeminiMdFilenames,
  // remove getCurrentGeminiMdFilename if no longer needed
} from '@qwen-code/qwen-code-core';

// in loadCliConfig:
} else {
  setServerGeminiMdFilename(getAllGeminiMdFilenames());
}

This keeps the single source of truth in memoryTool.ts.

2. There's a similar issue in packages/core/src/utils/ignorePatterns.ts

ignorePatterns.ts also uses getCurrentGeminiMdFilename() to build file exclusion patterns (line 163):

patterns.push(`**/${getCurrentGeminiMdFilename()}`);

This means AGENTS.md is not excluded from tool scan results (e.g., read_many_files, glob search), causing its content to potentially appear twice in the model's context — once from system memory loading and once from tool output. It should iterate over getAllGeminiMdFilenames() to exclude all configured context files:

for (const filename of getAllGeminiMdFilenames()) {
  patterns.push(`**/${filename}`);
}

zy6p · 2026-03-05T02:25:31Z

Thanks @LaZzyMan for the detailed review and the two concrete suggestions.

Both points make sense and I will update this PR accordingly:

Replace the fallback in CLI config reset with getAllGeminiMdFilenames() instead of hardcoded defaults.
Update packages/core/src/utils/ignorePatterns.ts to iterate over getAllGeminiMdFilenames() so all context files are excluded.

I’ll include test updates as well in the follow-up commit.

zy6p · 2026-03-05T02:34:13Z

Implemented in commit d7d2ffb.

Addressed both suggestions:

loadCliConfig fallback now uses getAllGeminiMdFilenames() (instead of hardcoded defaults).
ignorePatterns.ts now excludes all configured context filenames by iterating getAllGeminiMdFilenames().

Also updated tests/mocks accordingly (config.test.ts, ignorePatterns.test.ts).

Thanks again for the review. Please take another look when convenient.

LaZzyMan

LGTM!

Config.startNewSession() reassigns this.sessionId in the same process, which is reached by /clear, /reset, /new and /resume. Previously the old <oldId>.runtime.json was left behind, falsely claiming the still- live PID for a session no longer being served, and no new sidecar was written for the incoming session. Centralize the swap by clearing the old sidecar and writing a fresh one for the new session id from inside startNewSession itself, so all same-PID transitions are covered. The refresh runs as a fire-and- forget best-effort; failures must not block the session switch. Mirrors the post-merge Codex P1 fix on kimi-cli PR QwenLM#2082 (the source of the runtime.json sidecar pattern this PR ports). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Mirrors kimi-cli PR QwenLM#2082 commit e237951f (Codex P1 r3158754463): a short-lived non-interactive invocation (qwen --prompt, ACP, etc.) that runs `/clear` would otherwise call `Config.startNewSession()`, delete a concurrent shell's runtime.json sidecar (same outgoing session id), and never write a replacement — leaving the shell discoverable to nobody. Add a `runtimeStatusEnabled` flag on Config, flipped on by the interactive UI bootstrap immediately after the first successful sidecar write, and gate the swap-time refresh in `startNewSession()` on it. Non-interactive entry points never reach the bootstrap, so they won't touch sibling sidecars. Kimi later reverted the equivalent `write only from shell mode` guard (commit 7083975a) in favor of writing from every long-lived mode, but qwen's wire point is already interactive-only, so the narrower guard is the right shape here. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat(core): write runtime.json sidecar for active sessions Port kimi-cli PR #2082 part 1 to qwen-code. On interactive session start, atomically write a small JSON sidecar at <projectDir>/chats/<sessionId>.runtime.json recording the (pid, session_id, work_dir, hostname, started_at, qwen_version) tuple. External tools (terminal multiplexers, IDE integrations, status daemons) can map a running PID to its session id and work dir without parsing argv. Write is best-effort: a read-only filesystem must not block UI startup. OS process title (was #3713) and dynamic OSC tab title (kimi #2083) remain out of scope. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(core): refresh runtime.json on same-PID session swap Config.startNewSession() reassigns this.sessionId in the same process, which is reached by /clear, /reset, /new and /resume. Previously the old <oldId>.runtime.json was left behind, falsely claiming the still- live PID for a session no longer being served, and no new sidecar was written for the incoming session. Centralize the swap by clearing the old sidecar and writing a fresh one for the new session id from inside startNewSession itself, so all same-PID transitions are covered. The refresh runs as a fire-and- forget best-effort; failures must not block the session switch. Mirrors the post-merge Codex P1 fix on kimi-cli PR #2082 (the source of the runtime.json sidecar pattern this PR ports). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(core): only refresh runtime.json when this process owns it Mirrors kimi-cli PR #2082 commit e237951f (Codex P1 r3158754463): a short-lived non-interactive invocation (qwen --prompt, ACP, etc.) that runs `/clear` would otherwise call `Config.startNewSession()`, delete a concurrent shell's runtime.json sidecar (same outgoing session id), and never write a replacement — leaving the shell discoverable to nobody. Add a `runtimeStatusEnabled` flag on Config, flipped on by the interactive UI bootstrap immediately after the first successful sidecar write, and gate the swap-time refresh in `startNewSession()` on it. Non-interactive entry points never reach the bootstrap, so they won't touch sibling sidecars. Kimi later reverted the equivalent `write only from shell mode` guard (commit 7083975a) in favor of writing from every long-lived mode, but qwen's wire point is already interactive-only, so the narrower guard is the right shape here. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ents-context-loading fix(cli): keep AGENTS.md enabled by default context reset

fix(cli): keep AGENTS.md in default context file set

0c15576

zy6p requested review from DennisYu07, LaZzyMan, Mingholy, gwinthis, pomelo-nwu and tanzhenxin as code owners March 4, 2026 08:22

zy6p mentioned this pull request Mar 4, 2026

AGENTS.md does not take effect #2062

Closed

LaZzyMan reviewed Mar 4, 2026

View reviewed changes

tanzhenxin assigned LaZzyMan Mar 4, 2026

fix: follow review for default context filename handling

d7d2ffb

LaZzyMan approved these changes Mar 6, 2026

View reviewed changes

github-actions Bot mentioned this pull request Mar 6, 2026

📊 AI CLI 工具社区动态日报 2026-03-06 rollysys/agents-radar#42

Closed

LaZzyMan merged commit 881bef1 into QwenLM:main Mar 6, 2026
14 checks passed

tanzhenxin mentioned this pull request Mar 6, 2026

fix: add default support for AGENTS.md alongside QWEN.md #2093

Closed

yeelam-gordon mentioned this pull request Apr 28, 2026

feat(cli): set OS process title to encode session identity #3713

Closed

yeelam-gordon mentioned this pull request Apr 28, 2026

feat(core): write runtime.json sidecar for active sessions #3714

Merged

xaelistic pushed a commit to xaelistic/qwen-code that referenced this pull request Jun 7, 2026

Merge pull request QwenLM#2082 from zy6p/fix/2062-preserve-default-ag…

f501a07

…ents-context-loading fix(cli): keep AGENTS.md enabled by default context reset

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(cli): keep AGENTS.md enabled by default context reset#2082

fix(cli): keep AGENTS.md enabled by default context reset#2082
LaZzyMan merged 2 commits into
QwenLM:mainfrom
zy6p:fix/2062-preserve-default-agents-context-loading

zy6p commented Mar 4, 2026

Uh oh!

LaZzyMan left a comment

Uh oh!

zy6p commented Mar 5, 2026

Uh oh!

zy6p commented Mar 5, 2026

Uh oh!

LaZzyMan left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

zy6p commented Mar 4, 2026

TLDR

Dive Deeper

Reviewer Test Plan

Testing Matrix

Linked issues / bugs

Uh oh!

LaZzyMan left a comment

Choose a reason for hiding this comment

Uh oh!

zy6p commented Mar 5, 2026

Uh oh!

zy6p commented Mar 5, 2026

Uh oh!

LaZzyMan left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants