feat: expose Claude thinking/reasoning controls via ASKCC_CLAUDE_* env vars and CLI flags by monkut · Pull Request #79 · monkut/askcc-cli

monkut · 2026-04-16T02:29:06Z

Summary

Closes #78

Add ASKCC_CLAUDE_EFFORT_LEVEL, ASKCC_CLAUDE_MAX_THINKING_TOKENS, ASKCC_CLAUDE_DISABLE_THINKING, ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING env vars in askcc/settings.py
Add --effort, --max-thinking-tokens, --disable-thinking, --disable-adaptive-thinking / --no-disable-adaptive-thinking top-level CLI flags in askcc/cli.py
Wire all options through ClaudeRunner.run() in askcc/runners.py — --effort as a CLI flag, the rest as subprocess env vars
Precedence: CLI flag > ASKCC_CLAUDE_* env var > Claude Code default
Invalid ASKCC_CLAUDE_EFFORT_LEVEL values log a warning and fall through to Claude Code defaults
ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING defaults to true (adaptive thinking disabled)
DEFAULT_MAX_THINKING_TOKENS = 21000 (~5% of Max5 plan daily budget)

Settings Defaults

Setting	Default	Valid Values	Claude Code Env Var
`ASKCC_CLAUDE_EFFORT_LEVEL`	`None` (Claude Code decides)	`low`, `medium`, `high`, `max`	passed as `--effort` CLI flag
`ASKCC_CLAUDE_MAX_THINKING_TOKENS`	`21000`	positive integer	`MAX_THINKING_TOKENS`
`ASKCC_CLAUDE_DISABLE_THINKING`	`False`	`1`, `true` (truthy)	`CLAUDE_CODE_DISABLE_THINKING`
`ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING`	`True`	`0`, `false` to enable	`CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING`

Key Flows

CLI Flag Defaults

CLI Flag	Default	Env Var Override
`--effort`	`None` (Claude Code decides)	`ASKCC_CLAUDE_EFFORT_LEVEL`
`--max-thinking-tokens`	`21000`	`ASKCC_CLAUDE_MAX_THINKING_TOKENS`
`--disable-thinking`	`False`	`ASKCC_CLAUDE_DISABLE_THINKING`
`--disable-adaptive-thinking`	`True` (disabled)	`ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING`

Precedence & Resolution

flowchart TD
    A[CLI invocation] --> B{--effort flag?}
    B -- yes --> C[use CLI value]
    B -- no --> D{ASKCC_CLAUDE_EFFORT_LEVEL set?}
    D -- valid --> C
    D -- invalid --> E[log warning, skip]
    D -- unset --> E
    C --> F[pass --effort to claude subprocess]
    E --> F

    G[Settings resolved] --> H[ClaudeRunner.run]
    H --> I[Build cmd + env]
    I --> J{effort_level?}
    J -- yes --> K[cmd += --effort level]
    J -- no --> L[skip]
    I --> M{max_thinking_tokens?}
    M -- yes --> N[env MAX_THINKING_TOKENS=21000]
    I --> O{disable_thinking?}
    O -- yes --> P[env CLAUDE_CODE_DISABLE_THINKING=1]
    I --> Q{disable_adaptive_thinking?}
    Q -- yes --> R[env CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING=1]

Verification

uv run pytest tests/ -v — passed (153 tests)
uv run ruff check — passed (no issues)
uv run pyright — passed (0 errors)

Test plan

…rs and CLI flags (#78) Add env var defaults in settings.py and matching CLI flags on the top-level parser so they apply to every subcommand. Precedence: CLI flag > ASKCC_CLAUDE_* env var > Claude Code default. Settings: ASKCC_CLAUDE_EFFORT_LEVEL, ASKCC_CLAUDE_MAX_THINKING_TOKENS, ASKCC_CLAUDE_DISABLE_THINKING, ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING. CLI flags: --effort, --max-thinking-tokens, --disable-thinking, --disable-adaptive-thinking / --no-disable-adaptive-thinking.

monkut · 2026-04-16T02:51:56Z

+        "--effort",
+        choices=VALID_EFFORT_LEVELS,
+        default=settings.ASKCC_CLAUDE_EFFORT_LEVEL,
+        help="Claude thinking effort level. Env default: ASKCC_CLAUDE_EFFORT_LEVEL.",


update the help text to DISPLAY the default value as well.

Updated: help text now uses f-string to display the resolved settings.ASKCC_CLAUDE_EFFORT_LEVEL value.

monkut · 2026-04-16T02:52:17Z

+        "--max-thinking-tokens",
+        type=int,
+        default=settings.ASKCC_CLAUDE_MAX_THINKING_TOKENS,
+        help="Thinking token budget (default: 21000). Env default: ASKCC_CLAUDE_MAX_THINKING_TOKENS.",


update the help text to DISPLAY the default value from the actual value, do not hard code

Updated: help text now uses f-string with settings.ASKCC_CLAUDE_MAX_THINKING_TOKENS instead of hardcoded 21000.

monkut · 2026-04-16T02:52:22Z

+        "--disable-thinking",
+        action="store_true",
+        default=settings.ASKCC_CLAUDE_DISABLE_THINKING,
+        help="Force-disable extended thinking. Env default: ASKCC_CLAUDE_DISABLE_THINKING.",


update the help text to DISPLAY the default value as well.

Updated: help text now uses f-string to display the resolved settings.ASKCC_CLAUDE_DISABLE_THINKING value.

monkut · 2026-04-16T02:52:35Z

+        "--disable-adaptive-thinking",
+        action=argparse.BooleanOptionalAction,
+        default=settings.ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING,
+        help="Disable adaptive reasoning (Opus 4.6, Sonnet 4.6); default: true. "


update the help text to DISPLAY the default value from the actual value, do not hard code

Updated: help text now uses f-string with settings.ASKCC_CLAUDE_DISABLE_ADAPTIVE_THINKING instead of hardcoded true.

Use f-strings to show resolved settings values in --effort, --max-thinking-tokens, --disable-thinking, and --disable-adaptive-thinking help text.

…o 0.2.4 - Define CLAUDE_ENV_* constants in settings.py for subprocess env var names - Eliminate double os.getenv read for MAX_THINKING_TOKENS - Use constants in runners.py and tests instead of string literals - Bump version to 0.2.4

- Add EffortLevel StrEnum replacing VALID_EFFORT_LEVELS tuple - Default effort level to EffortLevel.MAX when unset - Use DEFAULT_EFFORT_LEVEL constant in tests instead of raw strings

Keep original VALID_EFFORT_LEVELS name as StrEnum class. Use DEFAULT_EFFORT_LEVEL constant in test assertions.

monkut mentioned this pull request Apr 16, 2026

feat: expose Claude thinking/reasoning controls via ASKCC_CLAUDE_* env vars and CLI flags #78

Closed

monkut commented Apr 16, 2026

View reviewed changes

monkut added 5 commits April 16, 2026 11:54

💄 Display actual default values in CLI help text

c7fc784

Use f-strings to show resolved settings values in --effort, --max-thinking-tokens, --disable-thinking, and --disable-adaptive-thinking help text.

✨ Default ASKCC_CLAUDE_EFFORT_LEVEL to 'max', convert to StrEnum

35cee98

- Add EffortLevel StrEnum replacing VALID_EFFORT_LEVELS tuple - Default effort level to EffortLevel.MAX when unset - Use DEFAULT_EFFORT_LEVEL constant in tests instead of raw strings

♻️ Rename EffortLevel to VALID_EFFORT_LEVELS, use constants in tests

46f85f7

Keep original VALID_EFFORT_LEVELS name as StrEnum class. Use DEFAULT_EFFORT_LEVEL constant in test assertions.

📝 Add comment explaining VALID_EFFORT_LEVELS placement

41fe723

monkut merged commit eda2b64 into main Apr 16, 2026

Navigation Menu

feat: expose Claude thinking/reasoning controls via ASKCC_CLAUDE_* env vars and CLI flags - #79

feat: expose Claude thinking/reasoning controls via ASKCC_CLAUDE_* env vars and CLI flags#79
monkut merged 6 commits into
mainfrom
feature/78-thinking-controls

monkut commented Apr 16, 2026 •

edited

Loading

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

monkut Apr 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

monkut commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Settings Defaults

Key Flows

CLI Flag Defaults

Precedence & Resolution

Verification

Test plan

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

monkut commented Apr 16, 2026 •

edited

Loading