fix(tui_gateway): honor target profile's terminal.cwd on desktop profile switch by xxxigm · Pull Request #40892 · NousResearch/hermes-agent

xxxigm · 2026-06-07T00:59:35Z

What does this PR do?

Fixes a profile-isolation bug in the desktop app: after switching profiles, a newly-started session inherited the previous profile's working directory instead of the directory configured in the profile you switched to.

The desktop's app-global remote mode serves every local profile from one tui_gateway backend, so the process-global TERMINAL_CWD env var only ever reflects the launch profile. When a new session was bound to another profile, _completion_cwd() resolved its workspace straight from that stale env var and never consulted the target profile's own config.yaml.

This PR teaches the session-cwd resolver to read the bound profile's configured terminal.cwd first:

Adds _profile_configured_cwd(profile_home) — reads terminal.cwd from a non-launch profile's config.yaml, skipping placeholder (./auto/cwd), empty, missing, and non-existent-directory values so callers fall back cleanly (mirrors the placeholder handling in gateway/run.py's config bridge).
Wires it into _completion_cwd() with the correct precedence: explicit client cwd → existing session cwd → bound profile's configured cwd → TERMINAL_CWD env → os.getcwd().

When no profile is passed (single-profile / launch-profile sessions) _profile_home() returns None and behavior is unchanged.

Related Issue

Fixes #40334

Type of Change

🐛 Bug fix (non-breaking change that fixes an issue)
✨ New feature (non-breaking change that adds functionality)
🔒 Security fix
📝 Documentation update
✅ Tests (adding or improving test coverage)
♻️ Refactor (no behavior change)
🎯 New skill (bundled or hub)

Changes Made

tui_gateway/server.py:
- Added _profile_configured_cwd() + a shared _CWD_PLACEHOLDERS constant (+34 lines).
- _completion_cwd() now prefers the bound profile's configured terminal.cwd over the launch profile's TERMINAL_CWD (+6/-2 lines).
tests/test_tui_gateway_server.py — added 4 tests (+58 lines):
- test_profile_configured_cwd_reads_target_profile — reads terminal.cwd from a profile's own config.yaml.
- test_profile_configured_cwd_skips_placeholders_and_missing — None, missing dir, ./auto/cwd/empty placeholders, and non-existent paths all fall through.
- test_completion_cwd_prefers_profile_over_stale_env — the Desktop: profile switch ignores terminal.cwd, uses wrong profile's working directory #40334 regression: a profile-bound session uses its own cwd, not the stale launch TERMINAL_CWD; no-profile still falls back to the env var.
- test_completion_cwd_explicit_cwd_wins_over_profile — an explicit client cwd still wins.

How to Test

Check out this branch and activate the venv: source .venv/bin/activate (or venv).
Run the new tests:
```
scripts/run_tests.sh tests/test_tui_gateway_server.py
```
Expected: the four new *_cwd tests pass alongside the existing suite.
Manual repro (desktop, app-global remote mode):
- Two profiles A/B with different terminal.cwd.
- Use A, switch to B, start a new session (/reset), ask the agent to pwd.
- Before: B's terminal opens in A's directory. After: B opens in its own configured terminal.cwd.

Checklist

Code

I've read the Contributing Guide
My commit messages follow Conventional Commits (fix(tui_gateway): ... / test(tui_gateway): ...)
I searched for existing PRs to make sure this isn't a duplicate
My PR contains only changes related to this fix (no unrelated commits)
I've run scripts/run_tests.sh tests/test_tui_gateway_server.py and the new tests pass
I've added tests for my changes
I've tested on my platform: macOS (Darwin 24.6.0), Python 3.12

Documentation & Housekeeping

I've updated relevant documentation (README, docs/, docstrings) — N/A (internal resolver, no user-facing surface change)
I've updated cli-config.yaml.example if I added/changed config keys — N/A (no new config keys)
I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — N/A
I've considered cross-platform impact (Windows, macOS) per the compatibility guide — issue is Windows-reported; the fix is pure path resolution and tests are hermetic (tmp_path + monkeypatch)
I've updated tool descriptions/schemas if I changed tool behavior — N/A

Screenshots / Logs

$ python -m pytest tests/test_tui_gateway_server.py -k "profile_configured_cwd or completion_cwd" -q
....                                                                     [100%]
4 passed, 205 deselected in 0.25s

…ile switch The desktop's app-global remote mode serves every profile from one tui_gateway backend, so the process-global TERMINAL_CWD only reflects the launch profile. After switching profiles, a new session resolved its workspace from that stale env var and inherited the previous profile's directory. Add _profile_configured_cwd() to read a non-launch profile's own terminal.cwd from its config.yaml (skipping placeholder/empty/missing and non-existent paths so callers fall back cleanly), and wire it into _completion_cwd() with precedence: explicit client cwd -> existing session cwd -> bound profile's configured cwd -> TERMINAL_CWD -> os.getcwd(). Fixes NousResearch#40334

Pin the new contract: _profile_configured_cwd reads a profile's own terminal.cwd and rejects placeholders/missing paths, and _completion_cwd prefers a bound profile's cwd over a stale launch-profile TERMINAL_CWD while still letting an explicit client cwd win.

szqjl · 2026-06-09T19:10:57Z

感谢该 PR 解决了网关端读取 Profile 自身的问题。从文档（profile-switch-terminal-cwd-fix.md）中可以看到，Hermes 原本的设计意图是 Profile 与 Workspace 正交（模式 B），即 Profile 决定 AI/工具，Workspace 决定当前工作目录。用户实际上也存在一种 “Profile‑Project 一对一” 的使用习惯（模式 A），即切 Profile 期望自动跳到该 Profile 配置的。当前补丁在网关端实现了 profile‑first 的优先级，能够完美修复模式 A 下的 bug。但正如文档所列的对比表所示，这会在模式 B 下引入新问题：用户在侧边栏工作区列表点击 “+” 时，所选的 Workspace 会被 Profile 默认目录覆盖。为了同时兼容两种使用模式，文档提出了一种 “温和版 scheme D + explicit_cwd 标记” 的完整方案：1. Desktop 端在显式选择 Workspace 时设置 = true，切 Profile 时复位；2. 创建会话时把该标记随 RPC 发送；3. 网关端在收到 profile 参数时，只有当 explicit_cwd 为 false 时才使用 Profile 自身的 terminal.cwd；否则尊重 params.cwd。这样既保证了模式 A 的自动跳转，又不破坏模式 B 的灵活性。请问是否可以在后续的迭代中考虑加入这一标记机制？如果目前的补丁已经足够满足当前的紧急需求，我们也可以先合并此 PR，并在后续的 Issue 中跟进完整的正交方案。期待维护者的意见，并愿意为实现该标记提供代码或测试上的帮助。

szqjl · 2026-06-09T19:11:57Z

感谢该 PR 解决了网关端读取 Profile 自身 terminal.cwd 的问题。我们理解Hermes 原本的设计意图是 Profile 与 Workspace 正交（模式 B），即 Profile 决定 AI/工具，Workspace 决定当前工作目录。用户实际上也存在一种 “Profile‑Project 一对一” 的使用习惯（模式 A），即切 Profile 期望自动跳到该 Profile 配置的 terminal.cwd。当前补丁在网关端实现了 profile‑first 的优先级，能够完美修复模式 A 下的 bug。这会在模式 B 下引入新问题：当用户在侧边栏工作区列表点击 “+” 显式选择 Workspace 时，该选择会被 Profile 默认的 terminal.cwd 覆盖，导致工作区被强制切回 Profile 配置的目录。为了同时兼容两种使用模式，我们提出了一种 “温和版 scheme D + explicit_cwd 标记” 的完整方案：1. Desktop 端在显式选择 Workspace 时设置 $explicitCwd = true，切 Profile 时复位；2. 创建会话时把该标记随 RPC 发送；3. 网关端在收到 profile 参数时，只有当 explicit_cwd 为 false 时才使用 Profile 自身的 terminal.cwd；否则尊重 params.cwd。这样既保证了模式 A 的自动跳转，又不破坏模式 B 的灵活性。请问是否可以在后续的迭代中考虑加入这一标记机制？我们愿意在后续的 Issue 中跟进完整的正交方案。并也可为实现该标记提供代码或测试上的协助。

OutThisLife

Approving. I traced the resolution path rather than going off the description:

_completion_cwd is the session cwd seed, not just autocomplete: session.create stores _completion_cwd(params) as the session's cwd and passes profile in params, so this lands on the exact path #40334 exercises (/reset / new session under a switched-to profile).
Precedence reads correctly: explicit client cwd > existing session cwd > bound profile config > TERMINAL_CWD > os.getcwd(). The bound profile's explicit config should beat the launch profile's stale env var, and it does.
No-profile / launch sessions: _profile_home(None) returns None, the resolver returns None, and it falls through to TERMINAL_CWD unchanged. No regression for single-profile use.
_CWD_PLACEHOLDERS matches gateway/run.py exactly, and empty/missing/non-existent dirs fall back cleanly.
The 4 hermetic tests cover the #40334 regression and the explicit-cwd-wins case. CI is green and the branch is clean against main.

LGTM.

…ile switch (NousResearch#40892) * fix(tui_gateway): honor target profile's terminal.cwd on desktop profile switch The desktop's app-global remote mode serves every profile from one tui_gateway backend, so the process-global TERMINAL_CWD only reflects the launch profile. After switching profiles, a new session resolved its workspace from that stale env var and inherited the previous profile's directory. Add _profile_configured_cwd() to read a non-launch profile's own terminal.cwd from its config.yaml (skipping placeholder/empty/missing and non-existent paths so callers fall back cleanly), and wire it into _completion_cwd() with precedence: explicit client cwd -> existing session cwd -> bound profile's configured cwd -> TERMINAL_CWD -> os.getcwd(). Fixes NousResearch#40334 * test(tui_gateway): cover per-profile cwd resolution (NousResearch#40334) Pin the new contract: _profile_configured_cwd reads a profile's own terminal.cwd and rejects placeholders/missing paths, and _completion_cwd prefers a bound profile's cwd over a stale launch-profile TERMINAL_CWD while still letting an explicit client cwd win.

szqjl · 2026-06-10T15:41:51Z

Thanks for the detailed review and approval! LGTM as well. 在 2026-06-10 06:13:51，brooklyn! ***@***.***> 写道： @OutThisLife approved this pull request. Approving. I traced the resolution path rather than going off the description: _completion_cwd is the session cwd seed, not just autocomplete: session.create stores _completion_cwd(params) as the session's cwd and passes profile in params, so this lands on the exact path #40334 exercises (/reset / new session under a switched-to profile). Precedence reads correctly: explicit client cwd > existing session cwd > bound profile config > TERMINAL_CWD > os.getcwd(). The bound profile's explicit config should beat the launch profile's stale env var, and it does. No-profile / launch sessions: _profile_home(None) returns None, the resolver returns None, and it falls through to TERMINAL_CWD unchanged. No regression for single-profile use. _CWD_PLACEHOLDERS matches gateway/run.py exactly, and empty/missing/non-existent dirs fall back cleanly. The 4 hermetic tests cover the #40334 regression and the explicit-cwd-wins case. CI is green and the branch is clean against main. LGTM. — Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS and Android. Download it today! You are receiving this because you commented.Message ID: ***@***.***>

…ile switch (NousResearch#40892) * fix(tui_gateway): honor target profile's terminal.cwd on desktop profile switch The desktop's app-global remote mode serves every profile from one tui_gateway backend, so the process-global TERMINAL_CWD only reflects the launch profile. After switching profiles, a new session resolved its workspace from that stale env var and inherited the previous profile's directory. Add _profile_configured_cwd() to read a non-launch profile's own terminal.cwd from its config.yaml (skipping placeholder/empty/missing and non-existent paths so callers fall back cleanly), and wire it into _completion_cwd() with precedence: explicit client cwd -> existing session cwd -> bound profile's configured cwd -> TERMINAL_CWD -> os.getcwd(). Fixes NousResearch#40334 * test(tui_gateway): cover per-profile cwd resolution (NousResearch#40334) Pin the new contract: _profile_configured_cwd reads a profile's own terminal.cwd and rejects placeholders/missing paths, and _completion_cwd prefers a bound profile's cwd over a stale launch-profile TERMINAL_CWD while still letting an explicit client cwd win.

alt-glitch added type/bug Something isn't working P3 Low — cosmetic, nice to have comp/tui Terminal UI (ui-tui/ + tui_gateway/) comp/gateway Gateway runner, session dispatch, delivery area/config Config system, migrations, profiles labels Jun 7, 2026

xxxigm added 2 commits June 9, 2026 06:46

xxxigm force-pushed the fix/desktop-profile-terminal-cwd branch from 7dc9617 to fc69e57 Compare June 8, 2026 23:46

OutThisLife approved these changes Jun 9, 2026

View reviewed changes

OutThisLife mentioned this pull request Jun 9, 2026

fix(tui): read terminal.cwd from config.yaml in _completion_cwd() #38757

Closed

23 tasks

OutThisLife merged commit 93340fa into NousResearch:main Jun 10, 2026
23 checks passed

Haderach-Ram mentioned this pull request Jun 10, 2026

Ecosystem Digest — 2026-06-10 Haderach-Ram/openclaw-radar#35

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(tui_gateway): honor target profile's terminal.cwd on desktop profile switch#40892

fix(tui_gateway): honor target profile's terminal.cwd on desktop profile switch#40892
OutThisLife merged 2 commits into
NousResearch:mainfrom
xxxigm:fix/desktop-profile-terminal-cwd

xxxigm commented Jun 7, 2026 •

edited

Loading

Uh oh!

szqjl commented Jun 9, 2026

Uh oh!

szqjl commented Jun 9, 2026 •

edited

Loading

Uh oh!

OutThisLife left a comment

Uh oh!

Uh oh!

szqjl commented Jun 10, 2026 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

xxxigm commented Jun 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Related Issue

Type of Change

Changes Made

How to Test

Checklist

Code

Documentation & Housekeeping

Screenshots / Logs

Uh oh!

szqjl commented Jun 9, 2026

Uh oh!

szqjl commented Jun 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

OutThisLife left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

szqjl commented Jun 10, 2026 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

xxxigm commented Jun 7, 2026 •

edited

Loading

szqjl commented Jun 9, 2026 •

edited

Loading