PR #28795: fix(cli): preserve -t/-m/--provider/--tui/--dev before chat subcommand

• Repository: NousResearch/hermes-agent
• Author: briandevans
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/28795

Description (problem / solution / changelog)

What does this PR do?

Fixes hermes -t web chat (and the sibling -m, --provider, --tui, --dev placements before the chat subcommand) silently dropping the flag value. Reported in #28780 for -t/--toolsets; the other four are sibling failures with the same root cause and are fixed in the same commit.

Root cause. The chat subparser re-declared these flags with default=None (or default=False for store_true) on top of the matching top-level parser flags. When argparse dispatches into the subparser it shares the namespace via dest, so the subparser's default overwrites whatever the top-level parser parsed before the subcommand. -s/--skills, -r/-c/-w, --yolo, and --pass-session-id already use default=argparse.SUPPRESS for exactly this reason — the chat-subparser action becomes a no-op unless the user explicitly passes the flag after chat, and the parent value survives. This PR extends the same fix to the remaining offenders.

Reproduction (origin/main, before fix):

>>> parser.parse_known_args(["-t", "web", "chat"]).toolsets
None
>>> parser.parse_known_args(["chat", "-t", "web"]).toolsets
'web'

After fix both calls return 'web'.

Audited siblings: contract test test_chat_subparser_inherited_value_flags_use_suppress scans every chat-subparser action whose dest is also on the top-level parser and asserts default is argparse.SUPPRESS. It fails on origin/main listing all five offenders (-m/--model, -t/--toolsets, --provider, --tui, --dev) and passes after this fix, so no widening needed.

Related Issue

Fixes #28780

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

• hermes_cli/_parser.py — switched chat subparser's -m/--model, -t/--toolsets, --provider, --tui, --dev declarations from default=None/default=False to default=argparse.SUPPRESS, matching the pattern already used for -s/--skills and the relaunch-inherited flags. Added a comment block explaining the SUPPRESS rationale.
• tests/hermes_cli/test_argparse_flag_propagation.py — added TestChatSubparserInheritedValueFlags that imports the real build_top_level_parser (not the hand-rolled replica higher in the file) so future drift to default=None would re-fail. Covers: parametrized before-chat / after-chat cases for each value flag; negative case where no flag is passed; combined case with all three value flags; store_true cases for --tui and --dev; and a contract test asserting every shared-dest chat flag uses SUPPRESS.

How to Test

1. Reproduce the bug on origin/main:

   uv run --with pytest --with pytest-xdist --with pytest-asyncio python3 -m pytest tests/hermes_cli/test_argparse_flag_propagation.py::TestChatSubparserInheritedValueFlags -v

   On origin/main 9 of the 15 new tests fail (the 6 after-chat cases were already passing because the bug only affects the before-chat position). After the fix, all 15 pass.
2. Run the full file plus adjacent CLI suites — both clean:

   uv run --with pytest --with pytest-xdist --with pytest-asyncio python3 -m pytest tests/hermes_cli/test_argparse_flag_propagation.py tests/hermes_cli/test_apply_profile_override.py tests/hermes_cli/test_ignore_user_config_flags.py tests/hermes_cli/test_tui_resume_flow.py tests/hermes_cli/test_setup_model_provider.py tests/hermes_cli/test_regression_16767.py tests/hermes_cli/test_tool_token_estimation.py tests/hermes_cli/test_tools_config.py tests/hermes_cli/test_setup_noninteractive.py tests/hermes_cli/test_banner.py tests/hermes_cli/test_gmi_provider.py tests/hermes_cli/test_config_drift.py tests/hermes_cli/test_tools_disable_enable.py

   29 + 67 + 156 = 252 tests pass locally on macOS 15 (Python 3.11.14).
3. Smoke-check the user's exact reported invocation — hermes -t web chat and verify /toolsets shows only web enabled.

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 focused tests for the touched code and all pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 15.x

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — N/A (inline comment added in _parser.py)
• I've updated cli-config.yaml.example if I added/changed config keys — N/A (no config keys touched)
• 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 — argparse behavior is platform-independent
• I've updated tool descriptions/schemas if I changed tool behavior — N/A

Changed files

• hermes_cli/_parser.py (modified, +20/-5)
• tests/hermes_cli/test_argparse_flag_propagation.py (modified, +117/-0)

PR #28799: fix(cli): preserve pre-subcommand flags across the chat subparser (#28780)

• Repository: NousResearch/hermes-agent
• Author: xxxigm
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/28799

Description (problem / solution / changelog)

What does this PR do?

hermes -t web chat silently parses identically to hermes chat — all default toolsets are still enabled rather than just web — even though hermes chat -t web works correctly. The same silent clobber affects -m/--model, --provider, --tui, and --dev (e.g. hermes -m gpt5 chat ignores the model override).

Root cause. Those flags are defined on both the top-level parser (so they work with -z/--oneshot and --tui) and on the chat subparser (so the natural hermes chat -t web works). The chat subparser's defaults were None / False, which argparse applies after the top-level parse — overwriting args.toolsets (and friends) whenever the user put the flag before chat.

Fix. Set default=argparse.SUPPRESS on the five duplicated chat-subparser flags so an absent post-subcommand occurrence leaves the top-level value intact. Mirrors the pattern already used by --resume, --continue, --yolo, --ignore-user-config, etc., which were unaffected by the bug.

Behaviour after the fix:

Related Issue

Fixes #28780

Type of Change

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

Changes Made

• hermes_cli/_parser.py — Set default=argparse.SUPPRESS on the chat subparser's -m/--model, -t/--toolsets, --provider, --tui, and --dev arguments so pre-subcommand values are no longer clobbered. Updated the top-level help text for -m, --provider, and -t to explicitly say they work before or after chat.
• tests/hermes_cli/test_pre_subcommand_flags.py — 27 new tests covering: each affected flag in both positions, post-overrides-pre semantics, composition of all five flags stacked pre-subcommand, regression coverage for the already-working SUPPRESS'd flags (-r, -c, -w, -s, --yolo, --ignore-user-config), and a structural contract test asserting the chat subparser uses argparse.SUPPRESS for these flags.

How to Test

Reproduce the bug on main:

hermes -t web chat
# In the CLI:
/toolsets
# Expected (per issue): only 'web' is enabled.
# Actual on main:       every default toolset is enabled.

After this PR:

hermes -t web chat
/toolsets   # ← shows only the 'web' toolset enabled
exit

hermes -m gpt5 chat                            # model override works pre-subcommand
hermes --provider openai chat                  # provider override works pre-subcommand
hermes --tui chat                              # TUI launches pre-subcommand
hermes chat -t web                             # post-subcommand still works (unchanged)
hermes -t web chat -t cli                      # post wins ("cli") — explicit override

Automated coverage:

scripts/run_tests.sh tests/hermes_cli/test_pre_subcommand_flags.py -q
# 27 passed in 1.19s

The broader tests/hermes_cli/ suite was also exercised — every failure observed (PTY/systemd/kanban-DB tests) is a pre-existing platform-incompatibility flake on macOS and also fails identically on upstream/main without this PR. The parser fix introduces zero new failures.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(cli):, test(cli):)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix
• I've run the relevant tests locally and they pass
• I've added tests for my changes (27 new regression tests)
• Tested on my platform: macOS 15.2 (Darwin 24.6.0)

Documentation & Housekeeping

• Updated relevant help text (-m, --provider, -t now explicitly mention they work before or after chat)
• N/A — no config keys added or changed
• N/A — no architecture or workflow changes
• N/A — argparse fix is platform-independent
• N/A — no tool descriptions / schemas changed

Screenshots / Logs

Direct argparse trace before the fix:

>>> sys.argv = ['hermes', '-t', 'web', 'chat']
>>> args, _ = parser.parse_known_args()
>>> args.toolsets
None        # ← bug: top-level value silently dropped

After the fix:

>>> sys.argv = ['hermes', '-t', 'web', 'chat']
>>> args, _ = parser.parse_known_args()
>>> args.toolsets
'web'       # ← top-level value preserved

Changed files

• hermes_cli/_parser.py (modified, +33/-8)
• tests/hermes_cli/test_pre_subcommand_flags.py (added, +242/-0)

PR #13: fix(cli): make -t/--toolsets work before the chat subcommand (#28780)

• Repository: na-navi/hermes-agent
• Author: na-navi
• State: open | merged: False
• Link: https://github.com/na-navi/hermes-agent/pull/13

Description (problem / solution / changelog)

🟢 Merge order: 1 / 12 — trivial, 1-line change, zero risk

Closes #28780 (P2)

Problem

hermes -t web chat silently ignored the -t flag. The top-level parser consumed -t web but the chat subparser's default of None overwrote it.

Fix

Use argparse.SUPPRESS as the default for the chat_parser's -t flag so it only sets toolsets when explicitly provided, letting the parent parser's value survive.

Risk assessment

Files changed

• hermes_cli/_parser.py (+1/-1)

Changed files

• hermes_cli/_parser.py (modified, +1/-1)

PR #28907: fix(cli): make -t/--toolsets work before subcommand (fixes #28780)

• Repository: NousResearch/hermes-agent
• Author: zccyman
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/28907

Description (problem / solution / changelog)

Fixes #28780

Problem

hermes -t web chat ignores the -t flag and enables all toolsets, while hermes chat -t web correctly enables only web.

Root Cause

Argparse creates separate namespaces for the main parser and the chat subparser. When -t web is placed before chat, the main parser correctly parses it and sets toolsets='web'. However, the subparser's add_argument("-t", ..., default=None) sets a default of None which overwrites the parent's value in the shared namespace.

Solution

After parser.parse_args(), if args.command == 'chat' and args.toolsets is None, scan _processed_argv for -t/--toolsets appearing before the chat token. If found, recover the value.

Files Changed

Test Results

hermes -t web chat     → toolsets='web'  ✓
hermes chat -t web     → toolsets='web'  ✓
hermes chat            → toolsets=None   ✓
hermes -t web chat -q  → toolsets='web'  ✓

All 736 existing CLI tests pass.

Changed files

• hermes_cli/main.py (modified, +12/-0)
• tests/cli/test_toolsets_argparse_fix.py (added, +49/-0)