markitect-main/examples/content-generator

Content Generator Example

Demonstrates: Prompt Dependency Resolution for Systematic Content Generation

This example shows how to use MarkiTect's Prompt Dependency Resolution infrastructure to generate InfoTech primers with full dependency tracking, provenance tracing, and quality validation.

Quick Start

# Run the example
python generate_primers.py

# Or if executable:
./generate_primers.py

What This Example Demonstrates

1. Artifact Management: Store templates, topics, and guidelines as versioned artifacts
2. Dependency Resolution: Automatically resolve @{macro} references in templates
3. Provenance Tracking: Trace generated content back to its inputs
4. Incremental Updates: Detect changes and regenerate only affected outputs
5. Quality Validation: Apply gates to ensure output meets standards
6. Visualization: Export dependency graphs to Mermaid or DOT format

Files

content-generator/
├── README.md                    # This file
├── TUTORIAL.md                  # Comprehensive tutorial (START HERE)
├── generate_primers.py          # Example implementation
│
├── templates/
│   └── generate-primer.md       # PromptTemplate for primer generation
│
├── artifacts/
│   ├── topics/                  # Topic definitions
│   │   ├── etl.md
│   │   └── microservices.md
│   │
│   └── guidelines/              # Authoring standards
│       ├── authoring-rules.md
│       └── research-prompt.md
│
└── prepdr/                      # Original manual system (preserved)
    ├── README.md
    ├── ETL.md
    ├── AuthoringRules.md
    ├── AssistentPrompt.md
    └── GeneratePrimerTemplate.md

Before vs After

Before (prepdr/ - Manual System)

# Manual macro substitution
{{ETL}}
{{AuthoringRules}}

Problems:

• No automation
• No version tracking
• No dependency awareness
• Can't detect changes
• No traceability

After (With Infrastructure)

# Automatic resolution
@{topic}
@{authoring_rules}

Benefits:

• ✅ Automatic macro resolution
• ✅ SHA-256 content digests for change detection
• ✅ Full dependency graph
• ✅ Incremental recomputation
• ✅ Complete provenance traces
• ✅ CLI tools for inspection

Tutorial

Read the full tutorial →

The tutorial covers:

• Architecture and core concepts
• Step-by-step implementation walkthrough
• Advanced features (incremental execution, quality gates, visualization)
• CLI usage examples
• Best practices

CLI Commands

After running the example, you can use CLI commands to inspect the system:

# Trace provenance
markitect prompt trace <artifact-id> --database primers.db

# Visualize dependencies
markitect prompt graph <artifact-id> --format mermaid --database primers.db

# List runs
markitect prompt runs --database primers.db

# Show impact debt (stale artifacts)
markitect prompt debt --database primers.db

# Graph statistics
markitect prompt stats --database primers.db

Architecture

graph TB
    A[Topic: ETL] -->|requires| D[Run: generate-etl]
    B[Authoring Rules] -->|requires| D
    C[Research Prompt] -->|requires| D
    D -->|generates| E[Output: ETL Primer]

    style E fill:#90EE90
    style D fill:#87CEEB

Key Features

1. Content-Based Addressing

Every artifact has a SHA-256 digest that changes when content changes:

artifact.content_digest  # "9c4d6e8a..."
artifact.has_changed(new_digest)  # True if modified

2. Dependency Graph

Full graph construction enables:

• Impact analysis: "What needs regeneration?"
• Provenance: "How was this produced?"
• Build order: Topological sort for correct execution

3. Incremental Execution

Only regenerate what's affected by changes:

config = RecomputeConfig(
    max_depth=2,              # Traverse 2 levels
    impact_threshold=0.1,     # Skip minor changes
    max_recomputes=10,        # Budget limit
)

4. Quality Validation

Apply automated checks:

gate = PatternValidationGate(
    required_patterns=[r"## Definition", r"## Context"],
    forbidden_patterns=[r"TODO", r"FIXME"],
)

Extending the Example

Add New Topics

# 1. Create artifact file
cat > artifacts/topics/oauth.md << 'EOF'
---
id: topic-oauth
name: OAuth
artifact_type: content
---

# OAuth 2.0

An authorization framework that enables...
EOF

# 2. Run generator
python generate_primers.py
# Automatically picks up new topic

Custom Quality Gates

from markitect.prompts.quality.gates.schema_gate import SchemaValidationGate

# Validate primer structure
schema = {
    "type": "object",
    "required": ["definition", "context", "scope"],
}

gate = SchemaValidationGate(schema=schema, gate_id="primer-schema")

Integrate with LLM

# Replace mock generation with real LLM call
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": compiled.content}]
)

output_content = response.choices[0].message.content

Performance

Database operations are optimized:

• Indexed lookups on artifact ID, space, digest
• Indexed dependency queries by source, target, run
• Efficient graph traversal with BFS/DFS
• Content digest comparison (no full content comparison)

Troubleshooting

"Artifact not found"

• Ensure artifact files exist in correct directories
• Check file extension is .md
• Verify space IDs match between template and config

"Resolution failed"

• Check macro names match artifact names
• Verify spaces are configured in ResolutionConfig
• Use ResolutionStrategy.FIRST_MATCH for simplicity

"Circular dependency detected"

• Review dependency edges in database
• Use detect_circular_dependencies() to find cycles
• Refactor template structure to break cycles

Related Documentation

• TUTORIAL.md - Comprehensive walkthrough
• Prompt Dependency Resolution Spec - Design documentation
• Phase 8 Implementation - Source code

License

Same as MarkiTect project.