Dev CLI Development

Dev-CLI Development Guide

Purpose

The AIDB dev-cli (src/aidb_cli/) is a Click-based command-line interface for AIDB development workflows. This skill guides developers in:

• Adding new CLI commands and command groups
• Creating reusable services following BaseService pattern
• Using custom decorators (@handle_exceptions, @require_repo_context)
• Integrating with Docker, test, and adapter systems
• Following Click framework best practices
• Proper error handling and output formatting

When to Use This Skill

Auto-activates when:

• Editing files in src/aidb_cli/
• Mentioning "dev-cli", "CLI command", "Click", "CliOutput"
• Adding commands, services, or CLI utilities
• Working with test orchestration or Docker integration

Related Skills

When developing CLI commands, you may also need:

• testing-strategy - CLI orchestrates test execution via test coordinator service
• code-reuse-enforcement - CLI must use existing constants and avoid magic strings

Architecture Overview

For comprehensive architecture details, see docs/developer-guide/cli-reference.md and source code in src/aidb_cli/.

Component Structure

src/aidb_cli/
├── commands/      # CLI command definitions (13 modules)
├── services/      # Business logic services (41+ files)
├── managers/      # High-level orchestration (singleton pattern)
├── core/          # Decorators, utilities, constants, param types
└── generators/    # Code generation for test scenarios

Key Patterns

1. Commands - Thin wrappers that delegate to services
2. Services - Reusable business logic with CommandExecutor
3. Managers - Singleton orchestrators for complex workflows
4. Context Injection - Dependencies via Click's ctx.obj
5. Unified Error Handling - @handle_exceptions decorator
6. Dynamic Parameter Types - Custom types with shell completion

Quick Start: Adding a Command

Step 1: Create Command File

Create src/aidb_cli/commands/mycommand.py:

"""My new command description."""

import click

from aidb_cli.core.decorators import handle_exceptions
from aidb_logging import get_cli_logger

logger = get_cli_logger(__name__)


@click.group(name="mycommand")
@click.pass_context
def group(ctx: click.Context) -> None:
    """Command group description."""
    pass


@group.command()
@click.option("--option", "-o", help="Option description")
@click.pass_context
@handle_exceptions
def subcommand(ctx: click.Context, option: str) -> None:
    """Subcommand description."""
    # Get dependencies from context
    output = ctx.obj.output
    repo_root = ctx.obj.repo_root
    executor = ctx.obj.command_executor

    # Delegate to service
    from aidb_cli.services.myservice import MyService
    service = MyService(repo_root, executor)
    result = service.do_something(option)

    # User-facing output (via OutputStrategy)
    output.success(f"Done: {result}")

Step 2: Register Command

In src/aidb_cli/cli.py:

from aidb_cli.commands import mycommand

# In main CLI setup
cli.add_command(mycommand.group)

Step 3: Test

./dev-cli mycommand subcommand --option value

Command Development Patterns

Decorator Order (CRITICAL)

Always stack decorators in this exact order:

@group.command()                    # 1. Click command
@click.option("--opt", "-o")        # 2. Options
@click.pass_context                 # 3. Context (BEFORE handle_exceptions)
@handle_exceptions                  # 4. Error handling (LAST)
def command(ctx: click.Context, opt: str) -> None:
    """Implementation."""

Wrong order causes cryptic errors!

Custom Parameter Types

Use custom types for validation and autocompletion:

from aidb_cli.core.param_types import TestSuiteParamType

@click.option("--suite", type=TestSuiteParamType(), required=True)
def command(ctx: click.Context, suite: str) -> None:
    """Command with validated suite parameter."""

Available: TestSuiteParamType, LanguageParamType, DockerProfileParamType, TestMarkerParamType

For comprehensive Click patterns, see click-framework-patterns.md:

• Decorator stacking rules explained
• All custom parameter types with examples
• Context management patterns
• Error handling with @handle_exceptions
• Command groups and subcommands
• Advanced Click patterns

Service Development Patterns

BaseService Pattern (Recommended)

Most services extend BaseService for consistent dependency injection and utilities:

from aidb_cli.managers.base.service import BaseService

class MyService(BaseService):
    def __init__(self, repo_root: Path, command_executor=None, ctx=None):
        super().__init__(repo_root, command_executor, ctx)
        # Inherited: command_executor, resolved_env, logging methods, path utilities

    def do_something(self) -> str:
        self.log_info("Starting operation...")
        return self.command_executor.execute(["cmd"], cwd=self.repo_root).stdout

See service-patterns.md for comprehensive BaseService details, advanced patterns, and testing examples.

Standalone Service Pattern

For simple services without BaseService:

class MyService:
    """Service for my operations."""

    def __init__(
        self,
        repo_root: Path,
        command_executor: CommandExecutor,
        logger: Logger | None = None
    ):
        self.repo_root = repo_root
        self.executor = command_executor
        self.logger = logger or get_cli_logger(__name__)

    def do_something(self, arg: str) -> str:
        """Do something useful."""
        from aidb.common.errors import AidbError

        result = self.executor.execute(["command", arg], cwd=self.repo_root)
        if result.returncode != 0:
            raise AidbError(f"Operation failed: {result.stderr}")
        return result.stdout.strip()

Key principles:

• Accept dependencies in __init__ (testable)
• Use CommandExecutor for all subprocess calls
• Log with get_cli_logger(__name__)
• Raise exceptions for errors

For comprehensive service patterns, see service-patterns.md:

• BaseService detailed patterns
• CommandExecutor comprehensive guide
• Service composition and singletons
• Testing services
• Real examples from AIDB codebase

Error Handling

@handle_exceptions Decorator

This decorator provides unified error handling:

@handle_exceptions
def command(ctx: click.Context) -> None:
    """Command with automatic error handling."""
    # Just raise errors naturally - decorator handles:
    # - KeyboardInterrupt (cleanup Docker resources)
    # - AidbError (formatted error output)
    # - FileNotFoundError (specific exit code)
    # - PermissionError (specific exit code)
    # - Generic exceptions (traceback in verbose mode)

    if not valid:
        from aidb.common.errors import AidbError
        raise AidbError("Validation failed")

Exit Codes

Use standard exit codes:

from aidb_cli.core.constants import ExitCode

if error:
    CliOutput.error("Operation failed")
    ctx.exit(ExitCode.GENERAL_ERROR)  # 1

if not found:
    ctx.exit(ExitCode.NOT_FOUND)  # 2

if config_error:
    ctx.exit(ExitCode.CONFIG_ERROR)  # 3

Output and Logging

OutputStrategy for Commands

Commands use ctx.obj.output (OutputStrategy) for verbosity-aware user-facing output:

output = ctx.obj.output

output.success("Operation completed")  # Always visible (green)
output.error("Operation failed")       # Always visible (red, stderr)
output.warning("Potential issue")      # Always visible (yellow)
output.plain("Regular message")        # Always visible (no color)
output.section("Title", Icons.ROCKET)  # Always visible (with separator)
output.info("Verbose detail")          # Only with -v flag
output.debug("Debug trace")            # Only with -vvv flag

CliOutput for Services/Managers

Services and managers (without Click context) use the static CliOutput utility:

from aidb_cli.core.utils import CliOutput

CliOutput.success("Operation completed successfully")
CliOutput.error("Operation failed")

Logger for Debugging

Use logger for debugging/trace information:

from aidb_logging import get_cli_logger

logger = get_cli_logger(__name__)

logger.debug("Detailed debugging info")
logger.info("General info")

Rule: Commands → ctx.obj.output, Services → CliOutput, Debugging → logger

Verbosity Levels

CLI supports three verbosity levels via global flags:

What each level includes:

• INFO: User-facing milestones (session started, breakpoint hit, etc.)
• DEBUG: Operation summaries, state transitions, adapter lifecycle
• TRACE: Full DAP/LSP JSON payloads, receiver timing, protocol details

Examples:

./dev-cli test run -s unit          # INFO level
./dev-cli -v test run -s unit       # DEBUG + adapter traces
./dev-cli -vvv test run -s unit     # TRACE + protocol payloads

Use -vvv for maximum observability when debugging DAP/LSP protocol issues.

Common Patterns

Import Organization

Follow project standard: stdlib → third-party → AIDB core → CLI → logging. All imports at top unless avoiding circular dependency.

Avoiding circular imports: Use TYPE_CHECKING for type-only imports:

from typing import TYPE_CHECKING

import click

if TYPE_CHECKING:
    from aidb_cli.cli import Context  # Only imported for type checking

See CLAUDE.md for full style details.

Service Instantiation in Commands

@group.command()
@click.pass_context
@handle_exceptions
def command(ctx: click.Context) -> None:
    """Command that uses a service."""
    # Instantiate service with dependencies from context
    from aidb_cli.services.adapter.adapter_build_service import AdapterBuildService

    service = AdapterBuildService(
        repo_root=ctx.obj.repo_root,
        command_executor=ctx.obj.command_executor
    )

    # Call service method
    result = service.build_locally(["python"], verbose=False)

    # Output result
    CliOutput.success(f"Built adapter: {result}")

Common Pitfalls

1. Decorator Order

Wrong: @handle_exceptions before @click.pass_context Right: @click.pass_context then @handle_exceptions

2. Using subprocess Directly

Wrong: subprocess.run(["cmd"]) Right: ctx.obj.command_executor.execute(["cmd"])

3. Mixing Output Types

Wrong: print("Success") or logger.info("User message") Right: CliOutput.success("Message") for users, logger.debug() for debugging

4. Hardcoded Paths

Wrong: Path("/path/to/repo") Right: ctx.obj.repo_root / "relative/path"

Real Code Examples

Example 1: Simple Command

From src/aidb_cli/commands/docker.py:

@group.command()
@click.option("--profile", "-p", type=DockerProfileParamType(), default=None)
@click.option("--no-cache", is_flag=True, help="Build without cache")
@click.pass_context
@handle_exceptions
def build(ctx: click.Context, profile: str | None, no_cache: bool) -> None:
    """Build Docker images."""
    from aidb_cli.services.docker.docker_build_service import DockerBuildService

    service = DockerBuildService(
        repo_root=ctx.obj.repo_root,
        command_executor=ctx.obj.command_executor
    )

    result = service.build_images(profile=profile, no_cache=no_cache)
    CliOutput.success(f"Build complete: {result}")

Example 2: Service with CommandExecutor

See real implementation: src/aidb_cli/services/docker/docker_build_service.py

Key pattern - services delegate to CommandExecutor and raise exceptions on failure:

class DockerBuildService:
    def __init__(self, repo_root: Path, command_executor: CommandExecutor):
        self.repo_root = repo_root
        self.executor = command_executor

    def build_images(self, profile: str | None = None) -> int:
        """Build Docker images."""
        from aidb.common.errors import AidbError

        cmd = ["docker-compose", "build"]
        if profile:
            cmd.extend(["--profile", profile])

        result = self.executor.execute(cmd, cwd=self.repo_root)
        if result.returncode != 0:
            raise AidbError(f"Build failed: {result.stderr}")
        return 0

Testing Your Changes

Run CLI Locally

./dev-cli mycommand subcommand --option value    # Run command
./dev-cli -v mycommand subcommand                # Verbose mode
./dev-cli --dry-run mycommand subcommand         # Dry run (no execution)

Unit Testing Services

Test services by mocking CommandExecutor:

from unittest.mock import Mock
from pathlib import Path

def test_my_service():
    """Test service logic without executing commands."""
    executor = Mock()
    executor.execute.return_value = Mock(returncode=0, stdout="success")

    service = MyService(Path("/tmp"), executor)
    result = service.do_something("arg")

    assert result == "success"
    executor.execute.assert_called_once_with(["command", "arg"], cwd=Path("/tmp"))

See actual test examples: src/tests/aidb_cli/commands/integration/ for command integration tests and src/tests/aidb_cli/services/ for service unit tests

Service Discovery Guide

Common tasks and which services to use:

Full service list: See src/aidb_cli/services/ subdirectories - docker/ (Docker operations), test/ (test execution), adapter/ (adapter building), docs/ (documentation), command_executor/ (subprocess execution)

Resource Files

For deeper dives into specific topics, see:

• service-patterns.md - Comprehensive service architecture patterns, CommandExecutor guide, testing, real examples
• click-framework-patterns.md - Click decorator stacking, custom parameter types, context management, error handling
• docker-compose-generation.md - Template-based docker-compose.yaml generation, Jinja2 templates, languages.yaml configuration, hash-based cache invalidation
• checksum-services.md - ChecksumServiceBase pattern, DockerImageChecksumService, FrameworkDepsChecksumService, container lifecycle tracking, creating custom checksum services