ai-knowledge-filler 0.1.1

Knowledge engineering system — transforms LLMs into deterministic Obsidian file generators

AI Knowledge Filler

Transform any LLM into a deterministic knowledge base generator

---

Problem → Solution

Problem: LLMs generate inconsistent, unstructured responses that require manual formatting.

Solution: System prompt that transforms any LLM into a deterministic file generator — same input, same structure, every time.

Result: Production-ready Markdown files with validated YAML metadata. Zero manual post-processing.

---

⚡ Quick Start (60 seconds)

Option 1: Claude Projects (Recommended)

1. Open Claude.ai → Create new Project
2. Project Knowledge → Upload Core_System/System_Prompt_AI_Knowledge_Filler.md
3. Custom Instructions → Paste Core_System/Custom_Instructions.md
4. Prompt: "Create guide on API authentication"
5. Done. Claude generates structured files.

Option 2: CLI (Multi-LLM)

# Clone repository
git clone https://github.com/petrnzrnk-creator/ai-knowledge-filler.git
cd ai-knowledge-filler

# Install
pip install -r requirements.txt

# Configure (any LLM)
export ANTHROPIC_API_KEY="sk-ant-..."  # or GOOGLE_API_KEY, OPENAI_API_KEY

# Generate
python3 cli.py generate "Create Docker security checklist" --model claude

Output: outputs/Docker_Security_Checklist.md — production-ready, validated.

---

What You Get

Core System (6 files)

• System Prompt — Transforms LLM from chat to file generator
• Metadata Standard — YAML structure specification
• Domain Taxonomy — 30+ classification domains
• Update Protocol — File merge rules
• Validation Script — Automated quality gates
• CLI — Multi-LLM interface (Claude, Gemini, GPT-4, Ollama)

Quality Assurance

• ✅ 96% test coverage (58 tests)
• ✅ Automated YAML validation
• ✅ CI/CD pipelines (GitHub Actions)
• ✅ Type hints (100% coverage)
• ✅ Linting (Pylint 9.55/10)

---

CLI Commands

Generate Files

# Auto-select first available LLM
python3 cli.py generate "Create Kubernetes deployment guide"

# Specific model
python3 cli.py generate "Create API checklist" --model claude
python3 cli.py generate "Create Docker guide" --model gemini
python3 cli.py generate "Create REST concept" --model gpt4
python3 cli.py generate "Create microservices reference" --model ollama

Validate Files

# Single file
python3 cli.py validate --file outputs/Guide.md

# All files
python3 cli.py validate

List Available Models

python3 cli.py models

# Output:
# ✅ claude     Claude (Anthropic) — claude-sonnet-4-20250514
# ✅ gemini     Gemini (Google) — gemini-3-flash-preview
# ❌ gpt4       GPT-4 (OpenAI) — Set OPENAI_API_KEY
# ✅ ollama     Ollama (llama3.2:3b)

---

Example Output

Input:

Create guide on API rate limiting

Output:

---
title: "API Rate Limiting Strategy"
type: guide
domain: api-design
level: intermediate
status: active
version: v1.0
tags: [api, rate-limiting, performance]
related:
  - "[[API Design Principles]]"
  - "[[System Scalability]]"
created: 2026-02-12
updated: 2026-02-12
---

## Purpose
Comprehensive strategy for implementing API rate limits...

## Core Principles
[Structured content with sections, code examples]

## Implementation
[Step-by-step technical guidance]

## Conclusion
[Summary and next steps]

Every file. Same structure. Validated automatically.

---

Architecture

User Prompt
    ↓
System Prompt (behavior definition)
    ↓
LLM Provider (Claude/Gemini/GPT-4/Ollama)
    ↓
Structured Markdown + YAML
    ↓
Automated Validation
    ↓
Production-Ready File

Key Insight: System prompt is the source of truth. Same prompt works across all LLMs.

---

Model Selection

Model	Speed	Cost	Best For
Claude	Medium	$$$	Technical docs, architecture
Gemini	Fast	$	Quick drafts, summaries
GPT-4	Slow	$$$$	Versatile content
Ollama	Very Fast	Free	Privacy, offline, local

Auto-selection: CLI tries providers in order: Claude → Gemini → GPT-4 → Ollama (first available).

---

Installation

Prerequisites

python3 --version  # 3.10+
pip3 --version

Setup

git clone https://github.com/petrnzrnk-creator/ai-knowledge-filler.git
cd ai-knowledge-filler

# Minimal (validation only)
pip install pyyaml

# Full (Multi-LLM generation)
pip install -r requirements.txt

API Keys

# Set at least one
export ANTHROPIC_API_KEY="sk-ant-..."
export GOOGLE_API_KEY="AIza..."
export OPENAI_API_KEY="sk-..."

# Or create .env file
cat > .env << 'EOF'
ANTHROPIC_API_KEY=your-key-here
GOOGLE_API_KEY=your-key-here
EOF

---

Testing

# Run all tests
pytest --cov=Scripts --cov-report=term-missing -v

# Run validation
python Scripts/validate_yaml.py

# Run linting
flake8 Scripts/ tests/
black --check Scripts/ tests/
pylint Scripts/ tests/

Coverage: 96% (58 tests)
Linting: Pylint 9.55/10
CI/CD: All checks passing

---

Use Cases

1. Technical Documentation Generate API docs, architecture decisions, deployment guides.

2. Knowledge Management Structure meeting notes, research findings, learning content.

3. Consulting Deliverables Create frameworks, methodologies, client reports.

4. Batch Processing Generate multiple files programmatically via CLI or API.

Full examples: Use_Cases_Documentation.md

---

File Types

type: concept      # Theoretical entity, definition
type: guide        # Step-by-step process
type: reference    # Specification, standard
type: checklist    # Validation criteria
type: project      # Project description
type: template     # Reusable template

30+ domains: api-design, system-design, devops, security, data-engineering, etc.

See Domain_Taxonomy.md for complete list.

---

Documentation

Core System

• System Prompt — LLM behavior definition
• Metadata Standard — YAML specification
• Domain Taxonomy — Classification system
• Update Protocol — Merge rules

Guides

• Deployment Guide — Installation options
• Use Cases — 20+ scenarios
• Workflow — 8-stage methodology

---

Advanced Usage

Programmatic Generation

from llm_providers import get_provider

# Auto-select provider
provider = get_provider("auto")

# Load system prompt
with open('Core_System/System_Prompt_AI_Knowledge_Filler.md') as f:
    system_prompt = f.read()

# Generate
content = provider.generate(
    prompt="Create API security checklist",
    system_prompt=system_prompt
)

# Save
with open('outputs/Security_Checklist.md', 'w') as f:
    f.write(content)

Batch Processing

# Create multiple files
cat > topics.txt << 'EOF'
Docker deployment best practices
Kubernetes security hardening
API authentication strategies
EOF

while read topic; do
    python3 cli.py generate "Create guide on $topic" --model gemini
done < topics.txt

---

Validation

Automated checks:

• ✅ YAML frontmatter present
• ✅ Required fields (title, type, domain, level, status, created, updated)
• ✅ Valid enum values (type, level, status)
• ✅ Domain in taxonomy
• ✅ ISO 8601 dates (YYYY-MM-DD)
• ✅ Tags array (3+ items)

Run validation:

python Scripts/validate_yaml.py
# or
python3 cli.py validate

Output:

✅ Documentation/Guide.md
❌ drafts/incomplete.md
   ERROR: Missing field: domain
   ERROR: Invalid type: document

---

Roadmap

v1.0 ✅

• System Prompt (universal LLM compatibility)
• YAML Metadata Standard
• Domain Taxonomy (30+ domains)
• Validation Script (96% test coverage)
• Multi-LLM CLI (Claude, Gemini, GPT-4, Ollama)
• CI/CD Pipelines (GitHub Actions)

v1.1 (Next)

• PyPI package (pip install akf-cli)
• VSCode extension (YAML validation)
• Obsidian plugin integration
• Enhanced documentation

---

License

MIT License — Free for commercial and personal use.

---

Philosophy

This is knowledge engineering, not chat enhancement.

LLMs are deterministic infrastructure, not conversational toys.

Before: "AI helps me write notes"
After: "AI compiles my knowledge base"

---

Created by: Petr — AI Solutions Architect
Repository: https://github.com/petrnzrnk-creator/ai-knowledge-filler
Version: 1.0.0

---

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

Quick Links:
Quick Start | CLI Commands | Documentation | Examples