feat(reasoning): full /reasoning CLI parity — show|hide + effort levels via config.yaml (#812)

Closes #461

Adds full /reasoning CLI parity to the WebUI slash command system:

- /reasoning show|on → window._showThinking = true; writes display.show_reasoning to config.yaml (same key as CLI); mirrors to settings.json for boot.js
- /reasoning hide|off → same in reverse; re-renders immediately
- /reasoning none|minimal|low|medium|high|xhigh → POST /api/reasoning → writes agent.reasoning_effort to config.yaml; takes effect next turn (matching CLI semantics)
- /reasoning (no args) → GET /api/reasoning → live status toast from config.yaml
- Autocomplete shows all 8 options: show|hide|none|minimal|low|medium|high|xhigh
- Profile-isolated: _get_config_path() is thread-local so per-profile settings never bleed across
- Boot hydration: window._showThinking initialised from settings.json show_thinking on page load
- Inspect.signature guard in streaming.py so older hermes-agent builds don't TypeError

28 new tests, 1708/1708 total passing. Full browser QA on port 8789 with isolated state. CLI/config.yaml sync verified with hermes_constants.parse_reasoning_effort().
This commit is contained in:
nesquena-hermes
2026-04-21 15:26:52 -07:00
committed by GitHub
parent f6e1612c7e
commit 811424a87b
12 changed files with 628 additions and 10 deletions

View File

@@ -18,13 +18,19 @@ BOOT_JS = (REPO_ROOT / "static" / "boot.js").read_text(encoding="utf-8")
STYLE_CSS = (REPO_ROOT / "static" / "style.css").read_text(encoding="utf-8")
def test_subarg_registry_exists_without_promoting_reasoning_to_builtin():
def test_subarg_registry_exists_and_reasoning_is_promoted_to_builtin():
# SLASH_SUBARG_SOURCES still exists for model and personality
assert "const SLASH_SUBARG_SOURCES=" in COMMANDS_JS
assert "reasoning:{desc:'Set reasoning effort', subArgs:['low','medium','high']}" in COMMANDS_JS
assert "{name:'reasoning'" not in COMMANDS_JS, \
"/reasoning suggestions must not register as a local built-in command"
assert "source:'subarg-command'" in COMMANDS_JS, \
"top-level autocomplete should still surface subarg-only commands like /reasoning"
# /reasoning is now a proper builtin command with a fn: handler (cmdReasoning)
# so it is in the COMMANDS array, not SLASH_SUBARG_SOURCES
assert "{name:'reasoning'" in COMMANDS_JS, \
"/reasoning must be registered as a local built-in command with fn: handler"
assert "fn:cmdReasoning" in COMMANDS_JS, \
"/reasoning entry must reference cmdReasoning function"
assert "function cmdReasoning" in COMMANDS_JS, \
"cmdReasoning function must be defined"
# source:'subarg-command' is still used for model/personality in SLASH_SUBARG_SOURCES
assert "source:'subarg-command'" in COMMANDS_JS
def test_model_and_personality_subargs_load_from_existing_apis():

View File

@@ -0,0 +1,397 @@
"""Tests for /reasoning show|hide slash command and show_thinking setting.
Covers:
- show_thinking in _SETTINGS_DEFAULTS and _SETTINGS_BOOL_KEYS (api/config.py)
- window._showThinking initialised in boot.js (settings and fallback paths)
- window._showThinking guard in ui.js renderMessages thinking card
- _renderLiveThinking guard in messages.js
- cmdReasoning function present in commands.js with show/hide/effort handling
- /reasoning in COMMANDS array (not just SLASH_SUBARG_SOURCES)
- show|hide present as subArgs in COMMANDS entry
"""
import pathlib
import re
REPO = pathlib.Path(__file__).parent.parent
def read(rel):
return (REPO / rel).read_text(encoding='utf-8')
# ── api/config.py ─────────────────────────────────────────────────────────────
class TestShowThinkingConfig:
"""show_thinking must appear in defaults and bool keys."""
def test_show_thinking_in_defaults(self):
src = read('api/config.py')
assert '"show_thinking": True' in src, (
"show_thinking must be True in _SETTINGS_DEFAULTS"
)
def test_show_thinking_in_bool_keys(self):
src = read('api/config.py')
assert '"show_thinking"' in src
# Find the _SETTINGS_BOOL_KEYS set and confirm show_thinking is in it
m = re.search(r'_SETTINGS_BOOL_KEYS\s*=\s*\{([^}]+)\}', src, re.DOTALL)
assert m, "_SETTINGS_BOOL_KEYS not found"
assert 'show_thinking' in m.group(1), (
"show_thinking must be in _SETTINGS_BOOL_KEYS"
)
# ── static/boot.js ────────────────────────────────────────────────────────────
class TestBootJsShowThinking:
"""window._showThinking must be set in both the settings and fallback paths."""
def test_settings_path_initialises_show_thinking(self):
src = read('static/boot.js')
# Must read from the settings object, defaulting true when absent
assert 'window._showThinking=s.show_thinking!==false' in src, (
"boot.js must initialise _showThinking from settings (default true)"
)
def test_fallback_path_initialises_show_thinking_true(self):
src = read('static/boot.js')
assert 'window._showThinking=true' in src, (
"boot.js fallback path must default _showThinking to true"
)
# ── static/ui.js ──────────────────────────────────────────────────────────────
class TestUiJsThinkingGate:
"""Historical thinking cards must be gated by window._showThinking."""
def test_thinking_card_gated_in_render_messages(self):
src = read('static/ui.js')
assert 'window._showThinking!==false' in src, (
"ui.js must gate thinkingCardHtml on window._showThinking"
)
# The guard must be on the same line as _thinkingCardHtml insertion
lines = src.splitlines()
for line in lines:
if '_thinkingCardHtml' in line and 'insertAdjacentHTML' in line:
assert 'window._showThinking' in line, (
f"thinking card insertion must be gated: {line.strip()}"
)
break
# ── static/messages.js ────────────────────────────────────────────────────────
class TestMessagesJsLiveThinkingGate:
"""Live streaming thinking card must be hidden when _showThinking is false."""
def test_live_thinking_gated(self):
src = read('static/messages.js')
assert 'window._showThinking===false' in src, (
"messages.js _renderLiveThinking must early-return when _showThinking is false"
)
# Guard must be inside _renderLiveThinking
m = re.search(r'function _renderLiveThinking\(.*?\{(.*?)^\s*\}',
src, re.DOTALL | re.MULTILINE)
assert m, "_renderLiveThinking not found"
assert 'window._showThinking' in m.group(1)
# ── static/commands.js ────────────────────────────────────────────────────────
class TestReasoningCommand:
"""cmdReasoning must be wired into COMMANDS with show/hide subArgs."""
def test_reasoning_in_commands_array(self):
src = read('static/commands.js')
# Must appear in COMMANDS array (not just SLASH_SUBARG_SOURCES)
m = re.search(r'const COMMANDS\s*=\s*\[(.*?)\];', src, re.DOTALL)
assert m, "COMMANDS array not found"
commands_block = m.group(1)
assert 'reasoning' in commands_block, (
"/reasoning must be in the COMMANDS array with a fn: handler"
)
assert 'fn:cmdReasoning' in commands_block or "fn: cmdReasoning" in commands_block, (
"/reasoning entry must reference cmdReasoning"
)
def test_reasoning_subargs_include_show_hide(self):
src = read('static/commands.js')
m = re.search(r'const COMMANDS\s*=\s*\[(.*?)\];', src, re.DOTALL)
assert m
commands_block = m.group(1)
# Find the reasoning entry
rm = re.search(r"\{name:'reasoning'.*?\}", commands_block, re.DOTALL)
assert rm, "reasoning entry not found in COMMANDS"
entry = rm.group(0)
assert 'show' in entry, "subArgs must include 'show'"
assert 'hide' in entry, "subArgs must include 'hide'"
def test_reasoning_not_only_in_subarg_sources(self):
src = read('static/commands.js')
# It's fine if SLASH_SUBARG_SOURCES is empty or doesn't have reasoning
# (reasoning moved to COMMANDS with a real fn)
m = re.search(r'const SLASH_SUBARG_SOURCES\s*=\s*\{(.*?)\};', src, re.DOTALL)
if m:
subarg_block = m.group(1)
assert 'reasoning' not in subarg_block, (
"reasoning must not remain in SLASH_SUBARG_SOURCES once it has a fn: handler"
)
def test_cmd_reasoning_function_exists(self):
src = read('static/commands.js')
assert 'function cmdReasoning' in src, (
"cmdReasoning function must be defined"
)
def test_cmd_reasoning_handles_show(self):
src = read('static/commands.js')
m = re.search(r'function cmdReasoning\(.*?\n\}', src, re.DOTALL)
assert m, "cmdReasoning not found"
fn = m.group(0)
# Handler must write to the UI render gate (assignment may use a
# boolean literal or a locally-computed variable) and call renderMessages.
assert re.search(r'window\._showThinking\s*=\s*(?:true|on)\b', fn), (
"cmdReasoning show branch must assign true/on to window._showThinking"
)
assert 'renderMessages' in fn, (
"show/hide branch must call renderMessages()"
)
# Persistence: POST to /api/reasoning (CLI-shared config.yaml) AND
# /api/settings (boot.js mirror).
assert "api('/api/reasoning'" in fn, (
"show/hide branch must POST to /api/reasoning so config.yaml "
"display.show_reasoning is updated (CLI parity)"
)
assert 'show_thinking' in fn, (
"show/hide branch must also mirror show_thinking into "
"WebUI settings.json for boot.js hydration"
)
def test_cmd_reasoning_handles_hide(self):
src = read('static/commands.js')
m = re.search(r'function cmdReasoning\(.*?\n\}', src, re.DOTALL)
assert m
fn = m.group(0)
# The hide branch shares logic with show via a computed `on` variable;
# the combined branch must test for both show|on and hide|off.
assert "arg==='hide'" in fn, "hide branch missing"
assert "arg==='off'" in fn, "off alias missing"
def test_cmd_reasoning_i18n_key_exists(self):
i18n = read('static/i18n.js')
assert 'cmd_reasoning' in i18n, (
"i18n.js must define the cmd_reasoning key"
)
def test_cmd_reasoning_posts_api_endpoints_not_gets(self):
"""Regression: the api() helper spreads its 2nd arg into fetch(), so
passing a plain options object without method:'POST' silently becomes
a GET and the update is dropped. Every mutating call inside
cmdReasoning (/api/reasoning, /api/settings) must be a proper POST."""
src = read('static/commands.js')
m = re.search(r'function cmdReasoning\(.*?\n\}', src, re.DOTALL)
assert m
fn = m.group(0)
# Every write call — /api/reasoning and /api/settings — must be a POST.
write_calls = re.findall(
r"api\('/api/(?:reasoning|settings)'\s*,[^)]*\)", fn
)
assert write_calls, "cmdReasoning must POST to /api/reasoning or /api/settings"
for call in write_calls:
assert "method:'POST'" in call or 'method: "POST"' in call, (
f"write call missing method:'POST' — would fall through "
f"to GET and silently drop the update: {call}"
)
assert 'JSON.stringify' in call, (
f"write call missing JSON body: {call}"
)
def test_cmd_reasoning_routes_effort_through_api_reasoning(self):
"""Effort levels (none|minimal|low|medium|high|xhigh) must POST to
/api/reasoning so the agent's config.yaml agent.reasoning_effort is
updated — the same key the CLI writes via `/reasoning <level>`.
Previous design stored in a dead client variable, which did nothing."""
src = read('static/commands.js')
m = re.search(r'function cmdReasoning\(.*?\n\}', src, re.DOTALL)
assert m
fn = m.group(0)
assert "api('/api/reasoning'" in fn, (
"cmdReasoning must POST effort levels to /api/reasoning so "
"config.yaml agent.reasoning_effort is updated (CLI parity)"
)
assert "'effort:'" in fn or 'effort:arg' in fn or 'effort: arg' in fn, (
"effort-level branch must send {effort: arg} to /api/reasoning"
)
# Must NOT still hold a dead local-only variable for effort.
assert '_reasoningEffort=' not in fn, (
"cmdReasoning must not keep a dead client-side _reasoningEffort "
"(effort now round-trips through /api/reasoning → config.yaml)"
)
def test_cmd_reasoning_routes_display_through_api_reasoning(self):
"""show|hide|on|off must POST to /api/reasoning (config.yaml
display.show_reasoning — the CLI's key) in addition to mirroring
into WebUI settings.json for boot.js hydration."""
src = read('static/commands.js')
m = re.search(r'function cmdReasoning\(.*?\n\}', src, re.DOTALL)
assert m
fn = m.group(0)
assert 'display:arg' in fn or 'display: arg' in fn or "'display'" in fn, (
"show|hide branch must send {display: arg} to /api/reasoning so "
"config.yaml display.show_reasoning matches the CLI's source"
)
def test_cmd_reasoning_supports_all_cli_effort_levels(self):
"""The effort-level set must match hermes_constants.VALID_REASONING_EFFORTS
+ 'none' — i.e. the exact set the CLI accepts in /reasoning."""
src = read('static/commands.js')
m = re.search(r'function cmdReasoning\(.*?\n\}', src, re.DOTALL)
assert m
fn = m.group(0)
for level in ('none', 'minimal', 'low', 'medium', 'high', 'xhigh'):
assert f"'{level}'" in fn, (
f"cmdReasoning must accept '{level}' (CLI parity with "
f"hermes_constants.parse_reasoning_effort)"
)
def test_reasoning_subargs_match_cli_levels(self):
"""Autocomplete subArgs must expose every CLI effort level + show/hide."""
src = read('static/commands.js')
m = re.search(r"\{name:'reasoning'[^}]*\}", src, re.DOTALL)
assert m, "reasoning COMMANDS entry not found"
entry = m.group(0)
for suggestion in (
'show', 'hide', 'none', 'minimal', 'low', 'medium', 'high', 'xhigh'
):
assert f"'{suggestion}'" in entry, (
f"reasoning subArgs must include '{suggestion}' for CLI parity"
)
# ── api/config.py — reasoning helpers ────────────────────────────────────────
class TestReasoningConfigHelpers:
"""Validate that api/config.py exposes the CLI-parity helpers and that
they read/write the same keys the CLI uses."""
def test_parse_reasoning_effort_matches_cli_semantics(self):
from api.config import parse_reasoning_effort, VALID_REASONING_EFFORTS
# Empty → None
assert parse_reasoning_effort('') is None
assert parse_reasoning_effort(None) is None
# none → disabled
assert parse_reasoning_effort('none') == {'enabled': False}
# Each valid level → {enabled, effort}
for level in VALID_REASONING_EFFORTS:
assert parse_reasoning_effort(level) == {'enabled': True, 'effort': level}
# Unknown → None (fall back to default)
assert parse_reasoning_effort('garbage') is None
# Case-insensitive + trimmed
assert parse_reasoning_effort(' HIGH ') == {'enabled': True, 'effort': 'high'}
def test_valid_reasoning_efforts_matches_hermes_constants(self):
"""Ensure our mirror stays in sync with hermes_constants."""
from api.config import VALID_REASONING_EFFORTS
# Snapshot-style assertion: if hermes_constants adds a level, this
# test will fail fast so we know to update WebUI too.
assert VALID_REASONING_EFFORTS == (
'minimal', 'low', 'medium', 'high', 'xhigh'
)
def test_set_reasoning_effort_persists_to_config_yaml(self, tmp_path, monkeypatch):
"""set_reasoning_effort writes agent.reasoning_effort to the active
profile's config.yaml — the same key the CLI writes."""
import api.config as cfg
cfgfile = tmp_path / 'config.yaml'
monkeypatch.setattr(cfg, '_get_config_path', lambda: cfgfile)
cfg.set_reasoning_effort('high')
import yaml as _yaml
data = _yaml.safe_load(cfgfile.read_text(encoding='utf-8'))
assert data.get('agent', {}).get('reasoning_effort') == 'high', (
"agent.reasoning_effort must be persisted to config.yaml"
)
def test_set_reasoning_display_persists_to_config_yaml(self, tmp_path, monkeypatch):
"""set_reasoning_display writes display.show_reasoning to the same
config.yaml the CLI writes."""
import api.config as cfg
cfgfile = tmp_path / 'config.yaml'
monkeypatch.setattr(cfg, '_get_config_path', lambda: cfgfile)
cfg.set_reasoning_display(False)
import yaml as _yaml
data = _yaml.safe_load(cfgfile.read_text(encoding='utf-8'))
assert data.get('display', {}).get('show_reasoning') is False, (
"display.show_reasoning must be persisted to config.yaml"
)
cfg.set_reasoning_display(True)
data = _yaml.safe_load(cfgfile.read_text(encoding='utf-8'))
assert data.get('display', {}).get('show_reasoning') is True
def test_set_reasoning_effort_rejects_invalid(self, tmp_path, monkeypatch):
import api.config as cfg
monkeypatch.setattr(cfg, '_get_config_path', lambda: tmp_path / 'config.yaml')
import pytest as _pt
with _pt.raises(ValueError):
cfg.set_reasoning_effort('garbage')
with _pt.raises(ValueError):
cfg.set_reasoning_effort('')
def test_get_reasoning_status_defaults_to_show_true(self, tmp_path, monkeypatch):
"""When config.yaml has no display section, show_reasoning defaults
to True (matches CLI default where the setting is opt-in)."""
import api.config as cfg
monkeypatch.setattr(cfg, '_get_config_path', lambda: tmp_path / 'config.yaml')
st = cfg.get_reasoning_status()
assert st['show_reasoning'] is True
assert st['reasoning_effort'] == ''
# ── api/streaming.py — AIAgent receives reasoning_config ──────────────────────
class TestStreamingReasoningWiring:
"""Confirm api/streaming.py reads agent.reasoning_effort from config and
passes parsed reasoning_config to AIAgent (so effort changes take effect
on the next session)."""
def test_streaming_reads_reasoning_effort_from_config(self):
src = read('api/streaming.py')
assert 'parse_reasoning_effort' in src, (
"api/streaming.py must import parse_reasoning_effort to translate "
"config.yaml agent.reasoning_effort into AIAgent reasoning_config"
)
assert "reasoning_config" in src and "'reasoning_config' in _agent_params" in src, (
"api/streaming.py must guard the reasoning_config kwarg with "
"inspect.signature so older hermes-agent builds don't TypeError"
)
# ── api/routes.py — /api/reasoning endpoints ──────────────────────────────────
class TestReasoningRoutes:
def test_get_api_reasoning_route_exists(self):
src = read('api/routes.py')
assert 'parsed.path == "/api/reasoning"' in src, (
"GET /api/reasoning route must exist"
)
assert 'get_reasoning_status' in src, (
"api/routes.py must import and call get_reasoning_status"
)
def test_post_api_reasoning_accepts_display(self):
src = read('api/routes.py')
# The POST branch reads 'display' from body and dispatches to
# set_reasoning_display.
assert 'set_reasoning_display' in src, (
"POST /api/reasoning must route display toggles through "
"set_reasoning_display"
)
def test_post_api_reasoning_accepts_effort(self):
src = read('api/routes.py')
assert 'set_reasoning_effort' in src, (
"POST /api/reasoning must route effort changes through "
"set_reasoning_effort"
)