proffesor-for-testing__agentic-qe/docs/guides/fleet-code-intelligence-integration.md

Fleet + Code Intelligence Integration Guide

Overview

Fleet initialization now includes automatic code intelligence checks to ensure agents have access to semantic search capabilities before being spawned.

Quick Start

Standard Fleet Init (Recommended)

# Run with wizard for best experience
aqe fleet init --wizard

# The wizard will:
# 1. Check for code intelligence index
# 2. Prompt to scan if missing
# 3. Guide you through fleet configuration

Skip Code Intelligence Check

# For CI/automated scenarios
aqe fleet init --skip-code-scan

# Or non-interactive mode
aqe fleet init --topology hierarchical --max-agents 15 --skip-code-scan

When to Use Code Intelligence

Use code intelligence when:

• Testing complex codebases (>1000 files)
• Need semantic code search
• Want improved agent accuracy (up to 80% better)
• Long-term project (index is reusable)

Skip code intelligence when:

• Quick testing/demos
• Small codebases (<100 files)
• CI/CD pipelines with time constraints
• Already have index from previous init

Building the Index

If fleet init detects no index, it will recommend:

aqe code index .

This command:

• Scans all source files (TypeScript, JavaScript, Python, Go, Rust, Java, Ruby, C/C++, and more)
• Builds knowledge graph in .agentic-qe/memory.db
• Takes ~30s for typical project (1000 files)
• Only needs to run once per project (use --incremental for updates)

Index Status

Check if you have an index:

# Part of fleet init output
aqe fleet init
# Shows: "✓ Code intelligence index found (X entries)"

# Or check memory backend directly
ls -la .agentic-qe/memory.db

Agent Benefits

Agents spawned with code intelligence have access to:

Semantic Search

// Find code by meaning, not just keywords
"Find authentication logic"
→ Returns: auth.service.ts, jwt.utils.ts, user.middleware.ts

Context-Aware Analysis

// Understand relationships between modules
"What calls the payment processor?"
→ Returns: Full call chain with context

Intelligent Test Generation

// Generate tests based on actual code patterns
"Generate tests for order service"
→ Returns: Tests covering all public methods + edge cases

Integration with AQE Init

The code intelligence check is shared between:

• aqe init - Runs scan as part of full initialization
• aqe fleet init - Checks for existing index, prompts if missing

Best Practice:

# Full setup (recommended for new projects)
aqe init --wizard

# Later, spawn fleet (will reuse index)
aqe fleet init --wizard

Troubleshooting

Index Not Found

# Check database exists
ls .agentic-qe/memory.db

# Rebuild index
aqe code index .

# Try fleet init again
aqe fleet init

Index Too Large

# Exclude directories from indexing
# Edit .agentic-qe/config.yaml:
domains:
  code-intelligence:
    indexing:
      exclude:
        - "test/**"
        - "build/**"
        - "dist/**"

Slow Indexing

# For large projects (>10k files)
# Use incremental indexing
aqe code index . --incremental

# Or index only recent changes
aqe code index . --git-since HEAD~10

# Or skip for now
aqe fleet init --skip-code-scan

CI/CD Integration

# GitHub Actions example
- name: Initialize AQE Fleet
  run: |
    # Skip scan in CI (use pre-built index or skip)
    aqe fleet init --skip-code-scan --topology hierarchical

Examples

Example 1: New Project

# First time setup
aqe init --wizard
# Includes code intelligence scan

# Later, spawn fleet
aqe fleet init
# Reuses existing index, no scan needed

Example 2: Existing Project (No Index)

# Start fleet init
aqe fleet init --wizard

# Prompted:
#   ⚠ No code intelligence index found
#   Run scan now? [Y/n]:

# Choose 'y', then:
aqe code index .

# Re-run fleet init
aqe fleet init
# Now shows: ✓ Code intelligence index found

Example 3: Skip for Speed

# Quick test without index
aqe fleet init --skip-code-scan --max-agents 5

Performance

Project Size	Index Time	Agents Benefit
<100 files	~5s	Minimal
100-1k files	~15s	Moderate (+40%)
1k-10k files	~45s	Significant (+70%)
>10k files	~2min	Maximum (+80%)

FAQ

Q: Do I need to rebuild the index after code changes? A: No, agents work with the snapshot. Rebuild periodically (weekly) for best accuracy.

Q: Can I use fleet without code intelligence? A: Yes! Use --skip-code-scan or choose 'n' when prompted. Agents will use basic analysis.

Q: What's stored in the index? A: AST nodes, relationships, function signatures, imports. No secrets or credentials.

Q: How much disk space does it use? A: Typically 5-10 MB for 1000 files. Stored in SQLite .agentic-qe/memory.db.