outowiki 0.5.13

AI agent knowledge management system using Wiki-based markdown storage

OutoWiki Documentation

OutoWiki is a wiki-based knowledge management system designed for AI agents. It provides a structured way to store, retrieve, and organize information that AI agents learn across interactions.

Overview

OutoWiki solves the problem of persistent memory for AI agents by organizing information in a familiar wiki structure. Instead of opaque databases, OutoWiki uses markdown documents organized in folders, making the knowledge human-readable and editable.

Wiki-Style Classification

OutoWiki follows Wikipedia/NamuWiki classification principles:

• is-a Relationship - Determine "What is this?" not "Is this similar?"
• Category Tree Navigation - Navigate hierarchical categories to find appropriate documents
• No Similarity Matching - Use topic understanding, not keyword matching
• Explicit Document Linking - Support [[Document Name]] syntax for direct connection

Key Features

• Folder-Based Classification - Categories are folders, no preset categories forced
• Dynamic Category Creation - Create new categories as needed
• Category Tree Exploration - Navigate and explore category hierarchy
• Required Title Validation - title is REQUIRED for all documents, auto-retry if missing
• LLM-Based Processing - Keyword extraction, category matching, topic splitting all use LLM
• Full Document Delivery - Entire document content delivered to LLM (no 500-character limit)
• Section-Based Editing - Wikipedia-style section editing (append, prepend, replace)
• Multi-Topic Splitting - Split content with multiple topics using LLM
• Wikilink Support - Direct document connection via [[Document Name]] syntax

Architecture

┌─────────────────────────────────────────────────────────┐
│                      OutoWiki Facade                     │
│  (OutoWiki class - main entry point for all operations)  │
└─────────────────────┬───────────────────────────────────┘
                      │
         ┌────────────┼────────────┐
         │            │            │
    ┌────▼────┐  ┌────▼────┐  ┌────▼────┐
    │Recorder │  │Searcher │  │Internal │
    │ Module  │  │ Module  │  │ Agent   │
    └────┬────┘  └────┬────┘  └────┬────┘
         │            │            │
    ┌────▼────────────▼────────────▼────┐
    │           LLM Provider            │
    │   (OpenAI or Anthropic)          │
    └──────────────────────────────────┘

The system has three main components:

• Recorder: Processes new content using Wiki-style topic classification (is-a relationship), determines document placement, manages backlinks
• Searcher: Finds relevant documents using semantic search and intent analysis
• Internal Agent: Handles complex operations like merge, split, and modify plans

Wiki Structure

OutoWiki organizes knowledge as markdown files in a folder hierarchy. No preset categories are forced - the wiki starts empty and categories are created dynamically as needed:

wiki/                    # Initially empty
├── programming/         # Created when first programming document is recorded
│   └── mobile/
│       └── camera.md
├── users/               # Created when first user document is recorded
│   └── alice/
│       └── preferences/
│           └── theme.md
└── ...                  # Categories grow organically

Each folder represents a category. When a document is recorded, the system:

1. Analyzes the content to determine its topic (is-a relationship)
2. Explores the existing category tree
3. Finds or creates the appropriate category folder
4. Records the document in that category

Documents support backlinks using the [[Document Name]] syntax. When auto_backlinks is enabled, OutoWiki automatically updates related documents when new content references existing topics.

Quick Start

from outowiki import OutoWiki, WikiConfig

# Create configuration
config = WikiConfig(
    provider="openai",
    api_key="sk-...",        # Your OpenAI API key
    model="gpt-4",
    wiki_path="./my_wiki"    # Local wiki folder
)

# Initialize the wiki
wiki = OutoWiki(config)

# Record new information
result = wiki.record({
    "type": "conversation",
    "content": "User prefers Python for web development. Suggested Flask or Django."
})
print(f"Recorded: {result.success}")
print(f"Actions: {result.actions_taken}")

# Search for information
results = wiki.search("programming preferences")
print(f"Found: {results.paths}")

# Work with a specific document
doc = wiki.get_document("concepts/web-development.md")
print(f"Title: {doc.metadata.title}")
print(doc.content[:500])

Documentation

Getting Started

• Installation - How to install OutoWiki
• Configuration - Configuration options and settings

API Reference

• OutoWiki Facade - Main entry point for all operations
• Data Models - WikiDocument, SearchQuery, etc.
• Plan Models - CreatePlan, ModifyPlan, MergePlan, etc.
• Analysis Models - IntentAnalysis, AnalysisResult
• Modules - Recorder, Searcher, InternalAgent
• Providers - OpenAI, Anthropic providers
• Exceptions - Error handling

Guides

• Recording Workflow - How to record information
• Search Strategies - How to search effectively
• Document Management - CRUD operations

License

Apache License 2.0 - see LICENSE file for details.