no more inquirer or clicker

[KRRT7]

PR Type

Enhancement, Other

---

Description

• Replace Inquirer with Textual prompts

• Add themed prompt wrapper module

• Update CLI flows to new API

• Adjust dependencies and mypy config

---

Diagram Walkthrough

flowchart LR
  Inquirer["inquirer-based prompts"] -- replaced by --> ThemedPrompts["themed_prompts (inquirer-textual)"]
  ThemedPrompts -- used in --> CmdInit["cmd_init.py flows"]
  PyProject["pyproject.toml deps"] -- add/remove --> InquirerTextual["inquirer-textual"], RemoveInquirer["remove inquirer"]
  MypyCfg["mypy overrides"] -- cleanup --> RemoveInquirer

Loading

File Walkthrough

	Relevant files
Enhancement	cli_common.py Remove legacy inquirer helpers from CLI common                      codeflash/cli_cmds/cli_common.py Remove inquirer utilities and wrapping helpers Keep only exit helper and console usage +0/-82    cmd_init.py Port init workflow to inquirer-textual prompts                      codeflash/cli_cmds/cmd_init.py Replace inquirer prompts with themed_prompts APIs Add Choice usage for formatter selection Add path existence checks for text inputs Remove custom inquirer theme and wrappers +74/-156 themed_prompts.py Add themed prompt wrapper around inquirer-textual                codeflash/cli_cmds/themed_prompts.py Introduce CodeflashThemedApp with custom CSS/theme Provide select, confirm, text, checkbox helpers Wrap inquirer-textual widgets with inline app +94/-0
Dependencies	pyproject.toml Update dependencies and mypy config for new prompts            pyproject.toml Remove inquirer dependency Add inquirer-textual dependency Update mypy ignore list to drop inquirer +2/-2

---

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Type Handling In multiple prompt flows, the returned value may be a Choice or a raw string. Ensure consistent handling and typing (especially when accessing .value or .value.data) to avoid runtime errors if the prompt returns unexpected types. formatter_choices = [ Choice("⚫ black", data="black"), Choice("⚡ ruff", data="ruff"), Choice("🔧 other", data="other"), Choice("❌ don't use a formatter", data="don't use a formatter"), ] result = prompts.select("Which code formatter do you use?", choices=formatter_choices, default=formatter_choices[0]) if result.command is None: apologize_and_exit() formatter = result.value.data if isinstance(result.value, Choice) else result.value git_remote = "" Exit Flow Consistency Several prompts treat result.command is None as a cancel and immediately exit. Verify this matches desired UX in all places (e.g., module/tests selection, GH Actions setup) and won’t surprise users by exiting instead of re-prompting. default_choice = project_name if project_name in module_subdir_options else module_subdir_options[0] result = prompts.select( "Which Python module do you want me to optimize?", choices=module_subdir_options, default=default_choice ) if result.command is None: apologize_and_exit() module_root_answer = result.value if module_root_answer == curdir_option: API Stability The wrapper relies on inquirer_textual internal widget classes and return structure. Confirm these APIs and the .run(inline=True) return shape (command/value) are stable across versions to prevent breakages. def select( # noqa: ANN201 message: str, choices: list[str | Choice], default: str | Choice | None = None ): # type: ignore[no-untyped-def] widget = InquirerSelect(message, choices, default, mandatory=True) app: CodeflashThemedApp = CodeflashThemedApp(widget, shortcuts=None, show_footer=False) return app.run(inline=True) def confirm(message: str, *, default: bool = False): # noqa: ANN201 # type: ignore[no-untyped-def] widget = InquirerConfirm(message, default=default, mandatory=True) app: CodeflashThemedApp = CodeflashThemedApp(widget, shortcuts=None, show_footer=False) return app.run(inline=True) def text(message: str): # noqa: ANN201 # type: ignore[no-untyped-def] widget = InquirerText(message) app: CodeflashThemedApp = CodeflashThemedApp(widget, shortcuts=None, show_footer=False) return app.run(inline=True) def checkbox( # noqa: ANN201 message: str, choices: list[str | Choice], enabled: list[str | Choice] | None = None ): # type: ignore[no-untyped-def] widget = InquirerCheckbox(message, choices, enabled) app: CodeflashThemedApp = CodeflashThemedApp(widget, shortcuts=None, show_footer=False) return app.run(inline=True)

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Use boolean confirm instead of select Use a proper boolean confirm prompt to avoid fragile string comparisons and localization issues. Replace the select with prompts.confirm and branch on its boolean value. codeflash/cli_cmds/cmd_init.py [573-577] -result = prompts.select("Create pyproject.toml in the current directory?", choices=["Yes", "No"], default="Yes") -if result.command is None or result.value == "No": +result = prompts.confirm("Create pyproject.toml in the current directory?", default=True) +if result.command is None or result.value is False: apologize_and_exit() create_empty_pyproject_toml(pyproject_toml_path) Suggestion importance[1-10]: 7 __ Why: Replacing a Yes/No select with a boolean confirm reduces string-comparison fragility and improves UX; the improved_code aligns with the PR’s new prompt framework (prompts.confirm) and keeps behavior consistent.	Medium
	Normalize select default and value handling Ensure the default passed to select matches the widget’s expected type. Pass the underlying value (label or Choice) consistently and avoid mixing plain strings and Choice objects to prevent selection mismatch. codeflash/cli_cmds/cmd_init.py [473-484] formatter_choices = [ Choice("⚫ black", data="black"), Choice("⚡ ruff", data="ruff"), Choice("🔧 other", data="other"), Choice("❌ don't use a formatter", data="don't use a formatter"), ] -result = prompts.select("Which code formatter do you use?", choices=formatter_choices, default=formatter_choices[0]) +default_choice = formatter_choices[0] +result = prompts.select("Which code formatter do you use?", choices=formatter_choices, default=default_choice) if result.command is None: apologize_and_exit() -formatter = result.value.data if isinstance(result.value, Choice) else result.value +formatter = result.value.data Suggestion importance[1-10]: 4 __ Why: Passing a Choice as default is already done and supported; forcing formatter = result.value.data may be slightly cleaner but the current isinstance guard is safe, making the improvement minor.	Low
General	Validate non-empty choices Guard against empty choices to prevent runtime errors inside the widget and provide a clear failure path. Validate input early and raise a helpful error message. codeflash/cli_cmds/themed_prompts.py [69-75] def select( # noqa: ANN201 message: str, choices: list[str | Choice], default: str | Choice | None = None ): # type: ignore[no-untyped-def] + if not choices: + raise ValueError("select() requires at least one choice") widget = InquirerSelect(message, choices, default, mandatory=True) app: CodeflashThemedApp = CodeflashThemedApp(widget, shortcuts=None, show_footer=False) return app.run(inline=True) Suggestion importance[1-10]: 6 __ Why: Adding an early check for empty choices prevents runtime errors and clarifies failures; it's a small but useful robustness improvement that accurately updates the provided code block.	Low

[codeflash-ai]

⚡️ Codeflash found optimizations for this PR

📄 55% (0.55x) speedup for RelativePathValidator.validate in codeflash/cli_cmds/validators.py

⏱️ Runtime : 3.22 milliseconds → 2.08 milliseconds (best of 59 runs)

A dependent PR with the suggested changes has been created. Please review:

• ⚡️ Speed up method RelativePathValidator.validate by 55% in PR #980 (inquirer) #983

If you approve, it will be merged into this PR (branch inquirer).

[claude]

Code Review: PR #980 - Replace inquirer with inquirer-textual

Thanks for this refactoring! The migration to inquirer-textual improves the CLI experience. However, I've identified several issues that should be addressed.

---

🔴 Critical Issues - Must Fix

1. Missing Return Type Annotations (themed_prompts.py)

All functions suppress return type warnings with # noqa: ANN201. This defeats type safety:

# Current (problematic):
def select(message: str, choices: list[str | Choice], ...):  # noqa: ANN201

# Should be:
def select(message: str, choices: list[str | Choice], ...) -> InquirerResult:

Impact: Makes API unclear to users, prevents type checking from catching bugs.

---

2. Unsafe Exception Handling (validators.py:85)

except Exception:  # noqa: S110
    pass

Issue: Silently swallows ALL exceptions including KeyboardInterrupt, SystemExit.
Fix: Catch specific exceptions:

except (OSError, ValueError, RuntimeError) as e:
    logger.debug(f"Path resolution failed: {e}")

---

3. Path Separator Bug (cmd_init.py:107)

create_for_me_option = f"🆕 Create a new tests{os.pathsep} directory for me!"

Bug: Uses os.pathsep (';' on Windows, ':' on Unix) instead of os.sep for directory separator.
Fix: f"🆕 Create a new tests{os.sep} directory for me!"

---

4. KeyboardInterrupt Not Propagated (github_utils.py:105)

except (KeyboardInterrupt, EOFError):
    console.print()  # Clean line for next prompt

Bug: Catches Ctrl+C but doesn't exit or re-raise. User can't cancel the operation.
Fix: Add raise or call apologize_and_exit()

---

5. Fragile Result Unpacking (cmd_init.py:161-165)

if has_multiple_remotes:
    module_root_answer, tests_root_answer, formatter_answer, enable_telemetry, git_remote = result.value
else:
    module_root_answer, tests_root_answer, formatter_answer, enable_telemetry = result.value

Risk: No validation that result.value has correct number of elements. Will raise ValueError if structure changes.
Fix: Use explicit indexing or a dataclass with named fields.

---

🟡 Important Issues - Should Fix

6. Type Checker Workarounds

Multiple places have unreachable code for type checker:

if not api_key:
    apologize_and_exit()
    return  # unreachable, satisfies type checker

Fix: Add -> NoReturn type hint to apologize_and_exit():

def apologize_and_exit() -> NoReturn:
    console.rule()
    # ...
    sys.exit(1)

---

7. Missing Subprocess Timeout (cmd_init.py:989-997)

with subprocess.Popen(command, text=True, cwd=args.module_root, ...) as process:

Issue: No timeout. Could hang indefinitely.
Fix: Add timeout parameter or use communicate(timeout=300)

---

8. Unsafe Directory Creation (cmd_init.py:428)

tests_root.mkdir()

Issue: Will raise FileExistsError if directory already exists.
Fix: tests_root.mkdir(parents=True, exist_ok=True)

---

9. Path Operation Without Error Handling (cmd_init.py:175)

tests_root = tests_root.relative_to(curdir)

Issue: Can raise ValueError if path is not relative.
Fix: Wrap in try/except with proper error message.

---

10. Insufficient Test Coverage

Missing tests:

• ❌ All of themed_prompts.py (0 tests)
• ❌ Most validators: PathExistsValidator, APIKeyValidator, NotEqualPathValidator, DirectoryOrCreatableValidator, TomlFileValidator
• ❌ New install_github_app() function
• ❌ Error/cancellation paths

Recommendation: Add tests for at least:

def test_api_key_validator_rejects_invalid_prefix()
def test_api_key_validator_allows_empty_for_browser_flow()
def test_path_exists_validator_with_nonexistent_path()
def test_select_or_exit_calls_exit_on_cancel()

---

🟢 Nice to Have

11. Extract CSS - The 50+ line CSS string in CodeflashThemedApp makes testing difficult. Consider extracting to a constant.

12. Add Retry Delays (github_utils.py:88-104) - Immediately retries without backoff could hammer the API.

13. Reduce API Key Exposure - Keys printed to console (cmd_init.py:202) could be captured in logs. Consider masking more characters.

14. Break Up Large Functions - init_codeflash() is 200+ lines, install_github_actions() is 110+ lines. These violate single responsibility principle.

---

✅ Positive Changes

• Clean separation of validators into dedicated module
• Removal of overly complex text-wrapping logic
• Good use of dataclasses for configuration
• Improved UX with themed prompts

---

📊 Overall Assessment

Risk Level: Medium

The refactoring successfully removes the inquirer dependency and introduces cleaner architecture. However, the lack of return type annotations, unsafe exception handling, and the os.pathsep bug need to be addressed before merge. The missing test coverage is also concerning.

Recommendation: Fix critical issues #1-5, add basic test coverage, then merge. Nice-to-haves can be addressed in follow-up PRs.

[aseembits93]

@KRRT7 reasoning for removing inquirer and clicker ?

[KRRT7]

@aseembits93 both of them together take up 50mb in dependencies, with inquirer taking up most of the bulk, and also I want to open the door towards using Textual bit by bit.