Implement qstudio plan command

[KofTwentyTwo]

User Story

As a developer, I want to generate implementation plans using AI so I can get structured guidance on how to implement features in my QQQ project.

Design

Command Interface

# Generate plan from prompt
qstudio plan "Add a new Order entity with CRUD operations"

# Generate plan from file
qstudio plan --file requirements.md

# Specify output location
qstudio plan "Add authentication" --output docs/PLAN-auth.md

# Use specific LLM provider
qstudio plan "Add caching" --provider anthropic

# Dry run (show what would be sent to LLM)
qstudio plan "Add logging" --dry-run

# Estimate tokens/cost
qstudio plan "Add metrics" --estimate

Output Format (plan.md)

# Implementation Plan: Add Order Entity

## Overview
Create a new Order entity with full CRUD operations following QQQ patterns.

## Files to Create
| File | Purpose |
|------|---------|
| `src/.../model/Order.java` | RecordEntity for orders |
| `src/.../metadata/OrderTableMetaDataProducer.java` | Table metadata |
| `src/.../api/OrderApiMetaDataProducer.java` | API endpoints |

## Implementation Steps

### Step 1: Create Order Entity
Create `Order.java` extending QRecord with fields:
- id (Integer, primary key)
- customerId (Integer, foreign key)
- orderDate (LocalDate)
- status (String)

### Step 2: Create Table Metadata
...

## Estimated Effort
- Files: 4
- Lines of code: ~200
- Complexity: Low

## Generated
- Date: 2024-01-15T10:30:00Z
- Provider: anthropic/claude-3-opus
- Tokens: 2,450

LLM Provider Configuration

# ~/.qctl/qctl.yaml
qstudio:
  default_provider: anthropic
  providers:
    anthropic:
      model: claude-3-opus-20240229
      api_key: ${ANTHROPIC_API_KEY}
    openai:
      model: gpt-4-turbo
      api_key: ${OPENAI_API_KEY}
    local:
      endpoint: http://localhost:11434
      model: codellama:34b

Files to Create/Modify

File	Action	Purpose
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/PlanCommand.java	Create	Main plan command
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/llm/LLMClient.java	Create	LLM client interface
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/llm/AnthropicClient.java	Create	Anthropic implementation
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/llm/OpenAIClient.java	Create	OpenAI implementation
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/llm/LocalClient.java	Create	Ollama/local implementation
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/prompt/PromptBuilder.java	Create	System/user prompt builder
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/prompt/ContextAssembler.java	Create	Index context assembly
qctl-qstudio/src/main/java/io/qrun/qctl/qstudio/model/Plan.java	Create	Plan output model
qctl-qstudio/src/main/resources/prompts/plan-system.txt	Create	System prompt template

Implementation Tasks

• Create PlanCommand with Picocli annotations
• Add --file, --output, --provider, --dry-run, --estimate options
• Define LLMClient interface
• Implement AnthropicClient with Claude API
• Implement OpenAIClient with GPT API
• Implement LocalClient for Ollama
• Create PromptBuilder with system prompt
• Create ContextAssembler to include relevant index context
• Parse LLM response into structured Plan
• Write plan to markdown file
• Add token counting and cost estimation
• Implement streaming output for long responses
• Write unit tests for PromptBuilder
• Write integration tests with mock LLM

Acceptance Criteria

• qstudio plan "<prompt>" generates implementation plan
• Uses project index for context
• Outputs structured markdown plan
• Supports multiple LLM providers
• --dry-run shows prompt without calling LLM
• --estimate shows token/cost estimate
• Saves plan to ledger automatically
• Respects path policies (Add allow/deny path policies #52)