Appearance
Guides & Ontology
This section provides practical manuals for building, maintaining, and auditing ARG-compliant systems.
Available Guides
1. Construction & Audit Guide
Building and validating your ARG ontology from the ground up
This guide addresses the critical bootstrap phase, how to design a solid foundational structure before letting the offline loop refine it.
What you'll learn:
- How to design context typologies correctly (stable, transversal, non-redundant)
- The "trees-first" method for rapid taxonomy construction
- Building the node graph progressively to avoid M2M explosion
- Running pre-IA audits that catch structural errors early
- Stop criteria: when to rebuild instead of patching
Key topics:
- Context typology families and anti-patterns
- Cluster and label hierarchy design rules
- Coverage, granularity, redundancy, and contradiction audits
- Practical deliverables to request from domain experts
Read the Construction & Audit Guide
2. Roles, Boundaries & Examples Guide
Understanding the "branches & leaves" model: Cluster / Label / Node / Edge
This guide clarifies the operational distinction between taxonomy (structure) and graph (resolution), and explains why this design protects long-term memory integrity.
What you'll learn:
- The "branches & leaves" metaphor: what each component actually does
- Operational definitions with real-world examples
- When to create a cluster vs label vs node vs edge
- How memory remains stable even as structure evolves
- The only sensitive case: node splits and migration tables
Key topics:
- Cluster: stable domain roots
- Label: hierarchical routing branches
- Node: terminal operational leaves (chunks + actions)
- Edge: movement grammar between leaves
- Memory safety under structural evolution
Read the Roles & Boundaries Guide
3. Action Agents Guide
How to design actions that stay governed, deterministic, and evolvable
This guide explains how ARG turns actions into first-class graph elements so an agent can execute the right action inside a bounded, auditable loop.
What you'll learn:
- When an agent should act vs answer vs abstain vs clarify
- How to model actions as nodes (and connect them safely)
- How policy gates apply before and after execution
- How actions evolve over time without breaking determinism
Key topics:
- Action space design patterns
- Authorization and execution boundaries
- Deterministic action routing via labels + binding
- Offline evolution of action definitions
4. Retrieval Agents Guide
Retrieval that stays bounded, fast, and audit-friendly
This guide shows how ARG uses taxonomy-coherent routing and bounded traversal to retrieve the right context without open-world expansion.
What you'll learn:
- How routing produces a stable
L_finaland a bounded eligible space - How landing points and neighbor scoring work in practice
- How context bundles are built for generation and post-checks
- When to use internal vs external retrieval
Key topics:
- Retrieval types (USER / DOMAIN / EXTERNAL)
- Landing point computation and traversal budgets
- Context bundle construction
- Reliability signals and post-check integration
Read the Retrieval Agents Guide
5. Short and Long-term Memory Agents Guide
Memory that remains stable while the graph evolves
This guide explains episodic writes, deduplication, and offline consolidation so you can retain useful information without polluting the ontology.
What you'll learn:
- What gets written online (episodic) vs consolidated offline (long-term)
- How canonicalization and deduplication keep information unique
- How access scope and policy constrain memory writes
- How migrations preserve continuity when nodes split/merge
Key topics:
- Info Registry and stable IDs
- Canonicalization and equivalence checks
- Lifecycle rules (active, deprecated, removed)
- Offline consolidation and governance
Read the Short and Long-term Memory Guide
Recommended Reading Order
For system architects
- Start with Roles & Boundaries to understand the model
- Then Construction & Audit to apply it
- Then Retrieval Agents to wire the online loop
- Then Action Agents to govern execution
- Finally Short and Long-term Memory for persistence and continuity
For domain experts
- Start with Construction & Audit for practical steps
- Refer to Roles & Boundaries when questions arise
- Skim Retrieval Agents to understand how your structure is used at runtime
Quick Reference
- New to ARG? Start with the Introduction
- Need the protocol? See ARG Core
- Understanding taxonomy validation? Read Context Weaver
- Policy setup? Check Policy Manager
Note: These guides focus on ontology design and operational structure. They are complementary to the ARG protocol specification and Context Weaver architecture.