Skip to content

CRP Scribe

Unlimited Long-Form Document & Book Generation

Any Model, Any Length

CRP Scribe uses the continuation engine to generate documents of unlimited length — books, manuals, whitepapers, regulatory reports. The output wall doesn't exist.

Overview

Every LLM has a finite output window. Ask for a 30-chapter book and you get 8 chapters before truncation. CRP Scribe eliminates this limitation entirely.

It combines CRP's continuation engine with a template system and outline generator to produce complete, coherent, multi-chapter documents in a single command.

Quick Start

Python API

from crp import CRPOrchestrator
from crp.products.scribe import CRPScribe

orchestrator = CRPOrchestrator(
    provider_name="openai",
    model_name="gpt-4o",
)

scribe = CRPScribe(orchestrator=orchestrator)

# Generate a 10-chapter whitepaper
result = scribe.generate(
    topic="AI Governance and the EU AI Act",
    template="whitepaper",
    num_chapters=10,
    auto_save_dir="./output/",
)

print(f"Total length: {len(result.full_text):,} characters")
print(f"Chapters: {len(result.chapters)}")

CLI

# List available templates
crp scribe templates

# Generate an outline
crp scribe outline --topic "Kubernetes Security" --chapters 8

# Generate a full document
crp scribe generate \
  --topic "AI Governance and the EU AI Act" \
  --template whitepaper \
  --chapters 10 \
  --output-dir ./output/

Templates

CRP Scribe ships with 5 built-in templates, each with optimized system prompts and chapter structures:

Template Best For Default Chapters
whitepaper Industry analysis, thought leadership, regulatory briefs 8
technical_manual API docs, deployment guides, runbooks 12
guide How-to guides, tutorials, onboarding docs 8
report Audit reports, assessments, findings 6
book Full-length books, comprehensive references 15

View Templates

crp scribe templates

scribe = CRPScribe(orchestrator=orchestrator)
for name in scribe.available_templates:
    print(name)

Outline Generation

Generate a structured outline before writing. The outline can be LLM-powered (when a provider is available) or use intelligent defaults:

# LLM-powered outline
outline = scribe.generate_outline(
    topic="Blockchain in Healthcare",
    num_chapters=8,
    template="whitepaper",
)

# Or use intelligent defaults (no LLM required)
outline = CRPScribe.generate_outline(
    topic="Blockchain in Healthcare",
    num_chapters=8,
)

Full Document Generation

With Progress Tracking

def on_progress(chapter_num: int, title: str, status: str):
    icons = {"generating": "⏳", "complete": "✅", "error": "❌"}
    print(f"{icons.get(status, '•')} Chapter {chapter_num}: {title}")

result = scribe.generate(
    topic="AI Ethics in Healthcare",
    template="book",
    num_chapters=12,
    progress_callback=on_progress,
    auto_save_dir="./my-book/",
)

With Custom Outline

outline = [
    "Introduction to AI Ethics",
    "The Hippocratic Oath in the Age of AI",
    "Bias in Medical AI Systems",
    "Patient Data Privacy",
    "Regulatory Landscape",
    "Case Studies",
    "Recommendations",
    "Conclusion",
]

result = scribe.generate(
    topic="AI Ethics in Healthcare",
    outline=outline,
    template="book",
)

With Additional Context

result = scribe.generate(
    topic="Kubernetes Networking",
    template="technical_manual",
    context="Focus on CNI plugins, service mesh, and network policies.",
)

Output Format

The DocumentResult object provides:

result.topic          # The document topic
result.template       # Template used
result.chapters       # List of ChapterResult objects
result.full_text      # Complete document as a single string
result.metadata       # Generation metadata (timestamp, model, etc.)

# Save all chapters to a directory
result.save("./output/")

# Export metadata as JSON
result.to_dict()

Each ChapterResult contains:

chapter.chapter_number  # 1-indexed chapter number
chapter.title           # Chapter title
chapter.content         # Full chapter text
chapter.word_count      # Word count
chapter.quality_tier    # CRP quality assessment (S/A/B/C/D)

Provider Support

CRP Scribe works with any CRP-supported LLM provider:

Provider Setup
OpenAI export OPENAI_API_KEY=sk-...
Anthropic export ANTHROPIC_API_KEY=sk-ant-...
Ollama --provider ollama --model llama3
LM Studio --provider lmstudio --model local-model

Contact