Implement session:start command - AI-powered session initialization

[jordanpartridge]

Overview

Create session:start command that provides intelligent, interactive context when starting a new Claude Code session.

The Smart Hook Experience

User starts Claude Code → Hook displays full context + AI analysis + interactive options

Example Output:

╔══════════════════════════════════════════════════════════════════════════════╗
║                      🚀 Session Start - knowledge                            ║
╚══════════════════════════════════════════════════════════════════════════════╝

📍 Current Context
  Project: knowledge
  Branch: feature/61-session-start
  Last Session: 2 hours ago

✅ Recent Milestones (Last 7 Days)
  • PR #47 merged - Bidirectional sync
  • 100% test coverage on sync commands

🚧 Active Blockers
  ⚠️  ChromaDB v2 API migration (3 days old)
  ⚠️  Session hooks storage inconsistent

📋 Recent Decisions
  • Keep both endpoints vs choosing one
  • Use Laravel type system for coverage

💬 Recent User Intents (Last 10)
  1. Deploy agent teams for new commands
  2. Make startup hook intelligent
  [... 8 more]

🎯 AI Analysis (Ollama)
  Strong momentum on knowledge improvements. ChromaDB blocker 3 days old
  preventing semantic search. Recommend: Deploy agents (quick wins) or
  tackle ChromaDB (foundational).

❓ What would you like to work on?

  1. 🎯 Deploy agent teams for new commands (HIGH MOMENTUM)
  2. 🔧 Continue ChromaDB v2 migration (BLOCKER REMOVAL)
  3. 📊 Review and update recent work
  4. 🆕 Something else

Your choice [1-4]: _

Acceptance Criteria

Context Display

• Shows current project/branch/last session time
• Shows recent milestones (last 7 days, ✅ emoji)
• Shows active blockers (with age, ⚠️ emoji, last attempt info)
• Shows recent decisions (last 7 days, 📋 emoji)
• Shows recent user intents (last 10, 💬 emoji)

AI Analysis

• Optional --ai flag integrates local LLM (Ollama/LM Studio)
• AI synthesizes patterns from context
• AI provides recommendations with reasoning
• Falls back to rule-based suggestions if no AI

Interactive Options

• Presents 3-4 smart options based on context
• Each option shows: emoji, title, description, estimated impact
• Options prioritized by: blockers > momentum > maintenance
• "Something else" option always available
• User selection persists to session context

Technical Requirements

• 100% test coverage
• Passes PHPStan level 8
• Passes Laravel Pint
• PR must pass Sentinel Gate to auto-merge
• Uses SessionService for session queries
• Uses Entry model for knowledge queries
• Integrates with AskUserQuestion pattern for choices

Agent Team Workflow

Agent 1 - Architect:

• Analyze SessionService, Entry model, user-intent tagging patterns
• Design context aggregation strategy (milestones/decisions/blockers queries)
• Design AI integration interface (local LLM via Process/HTTP)
• Create detailed recommendation for option generation logic

Agent 2 - Test Writer:

• Define contracts through unit tests (TDD approach)
• Test context display formatting
• Test option generation logic
• Test AI integration (with mocks)
• Test user choice handling

Agent 3 - Implementer:

• Implement to pass all unit tests
• Ensure 100% coverage
• Pass PHPStan level 8 + Laravel Pint
• Create PR that passes Sentinel Gate

Data Sources

• Entry model (tags: user-intent, milestone, blocker, decision)
• Session model (started_at, ended_at, summary)
• Git status (Process facade)
• Local AI (optional, Ollama API)

References

• Vision issue fix: update sync API URL and improve push reliability #54 (original request)
• Entry #7578 (good session summary format with MDO pattern)
• SessionService: app/Services/SessionService.php
• Session model: app/Models/Session.php

[jordanpartridge]

Strategic pivot: Closed as part of migration to AI-first semantic context engine. Productivity tracking replaced with pure vector storage + intelligent search. See MISSION.md in commit 641d7b4