# Configuration Profile Comparison

This document shows how the same 5-iteration generation task would behave differently with each configuration profile.

---

## Test Scenario

**Task:** Generate 5 D3.js visualizations
**Spec:** `specs/example_spec.md`
**Output:** `test_output/`

---

## Development Profile (USED IN THIS TEST)

**File:** `.claude/config/profiles/development.json`

### Key Settings
```json
{
  "max_parallel_agents": 2,
  "batch_size": 3,
  "min_uniqueness_threshold": 0.7,
  "enable_review_stage": true,
  "enable_progressive_sophistication": false,
  "output_directory": "output_dev",
  "naming_pattern": "{theme}_dev_{iteration:03d}.html",
  "include_metadata": true,
  "web_enhancement.enabled": true,
  "web_enhancement.initial_priming_urls": 2,
  "web_enhancement.cache_web_content": true,
  "logging.level": "debug",
  "logging.verbose": true,
  "limits.max_iterations": 10,
  "limits.max_file_size_kb": 300,
  "chain_prompting.stages": 8
}
```

### Behavior for 5 Iterations
- **Execution:** 2 waves (Wave 1: 3 iterations, Wave 2: 2 iterations)
- **Parallel agents:** 2 per wave
- **Quality bar:** 70% uniqueness required
- **Review:** Enabled (adds extra chain stage)
- **Web learning:** Enabled with caching
- **Logging:** Maximum detail (debug + verbose)
- **Naming:** climate_dev_001.html, network_dev_002.html, etc.
- **Metadata:** Included in all files
- **File limits:** 300 KB per file, 10 MB total
- **Speed:** Moderate (review stage adds time)

### Use Cases
- Testing new specifications
- Debugging issues
- Learning the system
- Rapid iteration with feedback
- Quality over speed

---

## Production Profile

**File:** `.claude/config/profiles/production.json`

### Key Settings
```json
{
  "max_parallel_agents": 5,
  "batch_size": 10,
  "min_uniqueness_threshold": 0.9,
  "enable_review_stage": false,
  "enable_progressive_sophistication": true,
  "output_directory": "output_prod",
  "naming_pattern": "{theme}_{iteration:04d}_{variant}.html",
  "include_metadata": false,
  "web_enhancement.enabled": true,
  "web_enhancement.initial_priming_urls": 5,
  "web_enhancement.cache_web_content": false,
  "logging.level": "warn",
  "logging.verbose": false,
  "limits.max_iterations": 1000,
  "limits.max_file_size_kb": 500,
  "chain_prompting.stages": 7
}
```

### Behavior for 5 Iterations
- **Execution:** 1 wave (batch_size 10 > 5 iterations)
- **Parallel agents:** 5 (processes all 5 simultaneously)
- **Quality bar:** 90% uniqueness required
- **Review:** Disabled (faster execution)
- **Web learning:** Enabled, no caching (fresh fetches)
- **Logging:** Minimal (warn level only)
- **Naming:** climate_0001_interactive.html, network_0002_animated.html
- **Metadata:** Not included (smaller files)
- **File limits:** 500 KB per file, 100 MB total
- **Speed:** Fast (no review, parallel processing)

### Use Cases
- Production deployments
- Large-scale generation (100+ iterations)
- Performance-critical scenarios
- High-quality output requirements
- Minimal overhead needed

### Differences from Development
- **5x faster:** 5 agents vs 2, no review stage
- **1 wave vs 2 waves:** Larger batch size
- **Higher quality bar:** 90% vs 70% uniqueness
- **No metadata:** Cleaner, smaller files
- **Different naming:** 4-digit iteration number, variant suffix
- **No caching:** Always fresh web content
- **Minimal logging:** Less output noise
- **Higher limits:** Can handle 1000 iterations vs 10

---

## Research Profile

**File:** `.claude/config/profiles/research.json`

### Key Settings
```json
{
  "max_parallel_agents": 3,
  "batch_size": 5,
  "min_uniqueness_threshold": 0.95,
  "enable_review_stage": true,
  "enable_progressive_sophistication": true,
  "output_directory": "output_research",
  "naming_pattern": "{theme}_research_{iteration:03d}.html",
  "include_metadata": true,
  "web_enhancement.enabled": true,
  "web_enhancement.initial_priming_urls": 8,
  "web_enhancement.cache_web_content": true,
  "web_enhancement.progressive_difficulty": true,
  "logging.level": "debug",
  "logging.verbose": true,
  "features.enable_cross_iteration_learning": true,
  "limits.max_iterations": 50,
  "limits.max_file_size_kb": 500,
  "chain_prompting.stages": 11
}
```

### Behavior for 5 Iterations
- **Execution:** 1 wave (batch_size 5 = 5 iterations)
- **Parallel agents:** 3
- **Quality bar:** 95% uniqueness required (highest)
- **Review:** Enabled (quality feedback)
- **Web learning:** Maximum (8 initial URLs, progressive difficulty)
- **Logging:** Maximum detail (debug + verbose)
- **Naming:** climate_research_001.html, network_research_002.html
- **Metadata:** Included with extensive details
- **File limits:** 500 KB per file, 50 MB total
- **Speed:** Slow (quality-focused, many stages)
- **Extra features:** Cross-iteration learning, 11 chain stages

### Use Cases
- Exploring new domains
- Maximum quality output needed
- Understanding system capabilities
- Experimenting with techniques
- Research and development
- Publication-quality results

### Differences from Development
- **Higher quality bar:** 95% vs 70% uniqueness
- **More web priming:** 8 URLs vs 2
- **Cross-iteration learning:** Enabled (learns from previous outputs)
- **Progressive difficulty:** Web sources increase in complexity
- **More chain stages:** 11 vs 8 (additional analysis stages)
- **Larger output allowance:** 500 KB vs 300 KB files
- **Quality over speed:** Sacrifices speed for maximum quality

---

## Side-by-Side Comparison

| Feature | Development | Production | Research |
|---------|------------|------------|----------|
| **Performance** |
| Parallel Agents | 2 | 5 | 3 |
| Batch Size | 3 | 10 | 5 |
| Waves for 5 items | 2 | 1 | 1 |
| Speed | Moderate | Fast | Slow |
| **Quality** |
| Uniqueness Threshold | 70% | 90% | 95% |
| Review Stage | Yes | No | Yes |
| Progressive Sophistication | No | Yes | Yes |
| Cross-Iteration Learning | Yes | No | Yes |
| **Web Enhancement** |
| Enabled | Yes | Yes | Yes |
| Initial Priming URLs | 2 | 5 | 8 |
| Cache Content | Yes | No | Yes |
| Progressive Difficulty | No | Yes | Yes |
| **Output** |
| Directory | output_dev | output_prod | output_research |
| Naming Pattern | {theme}_dev_{iteration:03d} | {theme}_{iteration:04d}_{variant} | {theme}_research_{iteration:03d} |
| Include Metadata | Yes | No | Yes |
| Max File Size | 300 KB | 500 KB | 500 KB |
| **Logging** |
| Level | debug | warn | debug |
| Verbose | Yes | No | Yes |
| Log Outputs | Yes | No | Yes |
| **Limits** |
| Max Iterations | 10 | 1000 | 50 |
| Max Total Output | 10 MB | 100 MB | 50 MB |
| Warn At | 5 | 100 | 25 |
| **Chain Prompting** |
| Stages | 8 | 7 | 11 |
| Self-Correction | Yes | Yes | Yes |
| **Features** |
| URL Strategy | Yes | Yes | Yes |
| Theme Evolution | Yes | Yes | Yes |
| Cross-Learning | Yes | No | Yes |
| Auto Indexing | Yes | Yes | Yes |

---

## Execution Time Estimates

**For 5 Iterations:**

| Profile | Estimated Time | Reasoning |
|---------|---------------|-----------|
| Development | 3-5 minutes | 2 waves, review stage, moderate logging |
| Production | 1-2 minutes | 1 wave, 5 agents, no review, minimal logging |
| Research | 8-12 minutes | 1 wave, 11 stages, extensive web priming, cross-learning |

**For 100 Iterations:**

| Profile | Estimated Time | Reasoning |
|---------|---------------|-----------|
| Development | N/A | Max 10 iterations (would need override) |
| Production | 15-25 minutes | 10 waves (batch 10), 5 agents, optimized |
| Research | 60-90 minutes | 20 waves (batch 5), extensive quality checks |

---

## Quality vs Speed Trade-offs

### Development Profile
- **Priority:** Balance + Learning
- **Quality:** Medium (70% uniqueness)
- **Speed:** Medium (review enabled)
- **Best for:** Testing, debugging, iterating

### Production Profile
- **Priority:** Speed + Scale
- **Quality:** High (90% uniqueness)
- **Speed:** High (no review, max agents)
- **Best for:** Deployments, large batches

### Research Profile
- **Priority:** Maximum Quality
- **Quality:** Very High (95% uniqueness)
- **Speed:** Low (many stages, extensive learning)
- **Best for:** Exploration, publication, research

---

## Switching Profiles

### How to Switch

**Command Syntax:**
```bash
/project:infinite-config <spec> <output> <count> [profile]
```

**Examples:**

```bash
# Development (testing)
/project:infinite-config specs/example_spec.md test_out 5 development

# Production (deployment)
/project:infinite-config specs/example_spec.md prod_out 100 production

# Research (exploration)
/project:infinite-config specs/example_spec.md research_out 20 research
```

### No Code Changes Required

Switching profiles changes:
- ✓ Batch sizes and wave structure
- ✓ Parallel agent counts
- ✓ Quality thresholds
- ✓ Review stages
- ✓ Web enhancement settings
- ✓ Logging detail
- ✓ File naming patterns
- ✓ Resource limits
- ✓ Chain prompting stages

**All without touching a single line of code.**

---

## Custom Configurations

You can also create custom configurations:

```bash
# Create custom config
/project:configure create production my_custom.json

# Use custom config
/project:infinite-config specs/example_spec.md output 50 custom my_custom.json

# Override specific values inline
/project:infinite-config specs/example_spec.md output 10 development '{"logging":{"level":"info"}}'
```

---

## Recommendation Guide

### Choose Development When:
- ✓ Testing new specifications
- ✓ Debugging generation issues
- ✓ Learning the system
- ✓ Need detailed logging
- ✓ Want review feedback
- ✓ Small iteration counts (< 10)

### Choose Production When:
- ✓ Deploying to production
- ✓ Large batch generation (50-1000 iterations)
- ✓ Speed is critical
- ✓ Quality requirements known and high (90%)
- ✓ Minimal logging preferred
- ✓ System is well-tested

### Choose Research When:
- ✓ Exploring new domains
- ✓ Maximum quality needed (95%+)
- ✓ Publication or presentation outputs
- ✓ Willing to sacrifice speed for quality
- ✓ Want cross-iteration learning
- ✓ Need extensive web research

---

## Conclusion

The configuration-driven architecture enables the same codebase to behave completely differently based on the selected profile. This provides:

1. **Flexibility:** Switch use cases without code changes
2. **Optimization:** Profiles tuned for specific scenarios
3. **Safety:** Validation ensures configurations are valid
4. **Reproducibility:** Version-controlled configs guarantee consistent results
5. **Scalability:** Easy to add new profiles for new use cases

**The power of configuration over code.**

---

**Document Generated:** 2025-10-10
**Test Profile Used:** Development
**Files Generated:** 5 visualizations + 2 documentation files