htm 0.0.1 → 0.0.2

data/docs/development/schema.md

@@ -1,6 +1,6 @@
 # Database Schema Documentation
 
-This document provides a comprehensive reference for HTM's PostgreSQL database schema, including all tables, indexes, and relationships.
+This document provides a comprehensive reference for HTM's PostgreSQL database schema, including query patterns, optimization strategies, and best practices.
 
 ## Schema Overview
 
@@ -22,367 +22,116 @@ CREATE EXTENSION IF NOT EXISTS vector WITH SCHEMA public;
 
 ## Entity-Relationship Diagram
 
-Here's the complete database structure:
+Here's the complete database structure (auto-generated by tbls):
 
-```svg
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 900" style="background: transparent;">
-  <defs>
-    <style>
-      .table-box { fill: #1e1e1e; stroke: #4a9eff; stroke-width: 2; }
-      .table-header { fill: #2d5a8e; }
-      .text-header { fill: #ffffff; font-family: monospace; font-size: 14px; font-weight: bold; }
-      .text-field { fill: #d4d4d4; font-family: monospace; font-size: 11px; }
-      .text-type { fill: #8cb4e8; font-family: monospace; font-size: 10px; }
-      .relation-line { stroke: #4a9eff; stroke-width: 1.5; fill: none; }
-      .arrow { fill: #4a9eff; }
-      .join-table { fill: #1e3a1e; stroke: #4a9eff; stroke-width: 2; }
-    </style>
-  </defs>
+![HTM Entity-Relationship Diagram](../database/schema.svg)
 
-  <!-- Robots Table -->
-  <rect class="table-box" x="50" y="50" width="280" height="140" rx="5"/>
-  <rect class="table-header" x="50" y="50" width="280" height="35" rx="5"/>
-  <text class="text-header" x="190" y="73" text-anchor="middle">robots</text>
+## Table Reference
 
-  <text class="text-field" x="60" y="100">id</text>
-  <text class="text-type" x="320" y="100" text-anchor="end">BIGSERIAL PK</text>
+For detailed table definitions, columns, indexes, and constraints, see the auto-generated documentation:
 
-  <text class="text-field" x="60" y="120">name</text>
-  <text class="text-type" x="320" y="120" text-anchor="end">TEXT</text>
+### Core Tables
 
-  <text class="text-field" x="60" y="140">created_at</text>
-  <text class="text-type" x="320" y="140" text-anchor="end">TIMESTAMPTZ</text>
+| Table | Description | Details |
+|-------|-------------|---------|
+| [robots](../database/public.robots.md) | Registry of all LLM robots using the HTM system | Stores robot metadata and activity tracking |
+| [nodes](../database/public.nodes.md) | Core memory storage for conversation messages and context | Vector embeddings, full-text search, deduplication |
+| [tags](../database/public.tags.md) | Unique hierarchical tag names for categorization | Colon-separated namespaces (e.g., `ai:llm:embeddings`) |
+| [working_memories](../database/public.working_memories.md) | Per-robot working memory state | Optional persistence for token-limited context |
 
-  <text class="text-field" x="60" y="160">last_active</text>
-  <text class="text-type" x="320" y="160" text-anchor="end">TIMESTAMPTZ</text>
+### Join Tables
 
-  <text class="text-field" x="60" y="180">metadata</text>
-  <text class="text-type" x="320" y="180" text-anchor="end">JSONB</text>
+| Table | Description | Details |
+|-------|-------------|---------|
+| [robot_nodes](../database/public.robot_nodes.md) | Links robots to nodes (many-to-many) | Enables "hive mind" shared memory architecture |
+| [node_tags](../database/public.node_tags.md) | Links nodes to tags (many-to-many) | Flexible multi-tag categorization |
 
-  <!-- Nodes Table -->
-  <rect class="table-box" x="50" y="250" width="280" height="400" rx="5"/>
-  <rect class="table-header" x="50" y="250" width="280" height="35" rx="5"/>
-  <text class="text-header" x="190" y="273" text-anchor="middle">nodes</text>
+### System Tables
 
-  <text class="text-field" x="60" y="300">id</text>
-  <text class="text-type" x="320" y="300" text-anchor="end">BIGSERIAL PK</text>
+| Table | Description | Details |
+|-------|-------------|---------|
+| [schema_migrations](../database/public.schema_migrations.md) | ActiveRecord migration tracking | Tracks applied migrations |
 
-  <text class="text-field" x="60" y="320">content</text>
-  <text class="text-type" x="320" y="320" text-anchor="end">TEXT NOT NULL</text>
+For the complete schema overview including all stored procedures and functions, see the [Database Tables Overview](../database/README.md).
 
-  <text class="text-field" x="60" y="340">speaker</text>
-  <text class="text-type" x="320" y="340" text-anchor="end">TEXT NOT NULL</text>
+## Key Concepts
 
-  <text class="text-field" x="60" y="360">type</text>
-  <text class="text-type" x="320" y="360" text-anchor="end">TEXT</text>
+### Content Deduplication
 
-  <text class="text-field" x="60" y="380">category</text>
-  <text class="text-type" x="320" y="380" text-anchor="end">TEXT</text>
+Content deduplication is enforced via SHA-256 hashing in the `nodes` table:
 
-  <text class="text-field" x="60" y="400">importance</text>
-  <text class="text-type" x="320" y="400" text-anchor="end">DOUBLE PRECISION</text>
+1. When `remember()` is called, a SHA-256 hash of the content is computed
+2. If a node with the same `content_hash` exists, the existing node is reused
+3. A new `robot_nodes` association is created (or updated if it already exists)
+4. This ensures identical memories are stored once but can be "remembered" by multiple robots
 
-  <text class="text-field" x="60" y="420">created_at</text>
-  <text class="text-type" x="320" y="420" text-anchor="end">TIMESTAMPTZ</text>
+### Hierarchical Tags
 
-  <text class="text-field" x="60" y="440">updated_at</text>
-  <text class="text-type" x="320" y="440" text-anchor="end">TIMESTAMPTZ</text>
-
-  <text class="text-field" x="60" y="460">last_accessed</text>
-  <text class="text-type" x="320" y="460" text-anchor="end">TIMESTAMPTZ</text>
-
-  <text class="text-field" x="60" y="480">token_count</text>
-  <text class="text-type" x="320" y="480" text-anchor="end">INTEGER</text>
-
-  <text class="text-field" x="60" y="500">in_working_memory</text>
-  <text class="text-type" x="320" y="500" text-anchor="end">BOOLEAN</text>
-
-  <text class="text-field" x="60" y="520">robot_id</text>
-  <text class="text-type" x="320" y="520" text-anchor="end">BIGINT FK</text>
-
-  <text class="text-field" x="60" y="540">embedding</text>
-  <text class="text-type" x="320" y="540" text-anchor="end">vector(2000)</text>
-
-  <text class="text-field" x="60" y="560">embedding_dimension</text>
-  <text class="text-type" x="320" y="560" text-anchor="end">INTEGER</text>
-
-  <!-- Tags Table -->
-  <rect class="table-box" x="850" y="250" width="280" height="120" rx="5"/>
-  <rect class="table-header" x="850" y="250" width="280" height="35" rx="5"/>
-  <text class="text-header" x="990" y="273" text-anchor="middle">tags</text>
-
-  <text class="text-field" x="860" y="300">id</text>
-  <text class="text-type" x="1120" y="300" text-anchor="end">BIGSERIAL PK</text>
-
-  <text class="text-field" x="860" y="320">name</text>
-  <text class="text-type" x="1120" y="320" text-anchor="end">TEXT UNIQUE</text>
-
-  <text class="text-field" x="860" y="340">created_at</text>
-  <text class="text-type" x="1120" y="340" text-anchor="end">TIMESTAMPTZ</text>
-
-  <!-- nodes_tags Join Table -->
-  <rect class="join-table" x="450" y="420" width="280" height="140" rx="5"/>
-  <rect class="table-header" x="450" y="420" width="280" height="35" rx="5"/>
-  <text class="text-header" x="590" y="443" text-anchor="middle">nodes_tags</text>
-
-  <text class="text-field" x="460" y="470">id</text>
-  <text class="text-type" x="720" y="470" text-anchor="end">BIGSERIAL PK</text>
-
-  <text class="text-field" x="460" y="490">node_id</text>
-  <text class="text-type" x="720" y="490" text-anchor="end">BIGINT FK</text>
-
-  <text class="text-field" x="460" y="510">tag_id</text>
-  <text class="text-type" x="720" y="510" text-anchor="end">BIGINT FK</text>
-
-  <text class="text-field" x="460" y="530">created_at</text>
-  <text class="text-type" x="720" y="530" text-anchor="end">TIMESTAMPTZ</text>
-
-  <!-- Relationships: robots -> nodes -->
-  <path class="relation-line" d="M 190 190 L 190 250"/>
-  <polygon class="arrow" points="190,250 185,240 195,240"/>
-
-  <!-- Relationships: nodes -> nodes_tags -->
-  <path class="relation-line" d="M 330 490 L 450 490"/>
-  <polygon class="arrow" points="450,490 440,485 440,495"/>
-
-  <!-- Relationships: tags -> nodes_tags -->
-  <path class="relation-line" d="M 850 310 L 730 310 L 730 510 L 730 510"/>
-  <polygon class="arrow" points="730,510 725,500 735,500"/>
-
-  <!-- Legend -->
-  <text class="text-field" x="50" y="720" font-weight="bold">Legend:</text>
-  <text class="text-field" x="50" y="740">PK = Primary Key</text>
-  <text class="text-field" x="200" y="740">FK = Foreign Key</text>
-  <text class="text-field" x="50" y="760">Green box = Join table (many-to-many)</text>
-
-  <!-- Annotations -->
-  <text class="text-field" x="400" y="370" font-style="italic">1:N</text>
-  <text class="text-field" x="380" y="480" font-style="italic">N:M</text>
-  <text class="text-field" x="770" y="480" font-style="italic">N:M</text>
-</svg>
-```
-
-## Table Definitions
-
-### robots
-
-The robots table stores registration and metadata for all LLM agents using the HTM system.
-
-**Purpose**: Registry of all robots (LLM agents) with their configuration and activity tracking.
+Tags use colon-separated hierarchies for organization:
+- `programming:ruby:gems` - Programming > Ruby > Gems
+- `database:postgresql:extensions` - Database > PostgreSQL > Extensions
+- `ai:llm:embeddings` - AI > LLM > Embeddings
 
+Query by prefix to find all related tags:
 ```sql
-CREATE TABLE public.robots (
-    id bigint NOT NULL,
-    name text,
-    created_at timestamp with time zone DEFAULT CURRENT_TIMESTAMP,
-    last_active timestamp with time zone DEFAULT CURRENT_TIMESTAMP,
-    metadata jsonb
-);
-
-ALTER TABLE ONLY public.robots ALTER COLUMN id SET DEFAULT nextval('public.robots_id_seq'::regclass);
-ALTER TABLE ONLY public.robots ADD CONSTRAINT robots_pkey PRIMARY KEY (id);
+SELECT * FROM tags WHERE name LIKE 'database:%';  -- All database-related tags
+SELECT * FROM tags WHERE name LIKE 'ai:llm:%';    -- All LLM-related tags
 ```
 
-**Columns**:
+### Remember Tracking
 
-| Column | Type | Nullable | Default | Description |
-|--------|------|----------|---------|-------------|
-| `id` | BIGINT | NO | AUTO | Unique identifier (primary key) |
-| `name` | TEXT | YES | NULL | Human-readable name for the robot |
-| `created_at` | TIMESTAMPTZ | YES | NOW() | When the robot was first registered |
-| `last_active` | TIMESTAMPTZ | YES | NOW() | Last time the robot accessed the system |
-| `metadata` | JSONB | YES | NULL | Robot-specific configuration and metadata |
+The `robot_nodes` table tracks per-robot remember metadata:
 
-**Indexes**:
-- `PRIMARY KEY` on `id`
+1. `first_remembered_at` - When this robot first encountered this content
+2. `last_remembered_at` - Updated each time the robot tries to remember the same content
+3. `remember_count` - Incremented each time (useful for identifying frequently reinforced memories)
 
-**Relationships**:
-- One robot has many nodes (1:N)
+This allows querying for:
+- Recently reinforced memories: `ORDER BY last_remembered_at DESC`
+- Frequently remembered content: `ORDER BY remember_count DESC`
+- New vs old memories: Compare `first_remembered_at` across robots
 
 ---
 
-### nodes
-
-The core table storing all memory nodes with vector embeddings for semantic search.
-
-**Purpose**: Stores all memories (conversation messages, facts, decisions, code, etc.) with full-text and vector search capabilities.
-
-```sql
-CREATE TABLE public.nodes (
-    id bigint NOT NULL,
-    content text NOT NULL,
-    speaker text NOT NULL,
-    type text,
-    category text,
-    importance double precision DEFAULT 1.0,
-    created_at timestamp with time zone DEFAULT CURRENT_TIMESTAMP,
-    updated_at timestamp with time zone DEFAULT CURRENT_TIMESTAMP,
-    last_accessed timestamp with time zone DEFAULT CURRENT_TIMESTAMP,
-    token_count integer,
-    in_working_memory boolean DEFAULT false,
-    robot_id bigint NOT NULL,
-    embedding public.vector(2000),
-    embedding_dimension integer,
-    CONSTRAINT check_embedding_dimension CHECK (((embedding_dimension IS NULL) OR ((embedding_dimension > 0) AND (embedding_dimension <= 2000))))
-);
-
-ALTER TABLE ONLY public.nodes ALTER COLUMN id SET DEFAULT nextval('public.nodes_id_seq'::regclass);
-ALTER TABLE ONLY public.nodes ADD CONSTRAINT nodes_pkey PRIMARY KEY (id);
-ALTER TABLE ONLY public.nodes
-    ADD CONSTRAINT fk_rails_60162e9d3a FOREIGN KEY (robot_id) REFERENCES public.robots(id) ON DELETE CASCADE;
-```
-
-**Columns**:
-
-| Column | Type | Nullable | Default | Description |
-|--------|------|----------|---------|-------------|
-| `id` | BIGINT | NO | AUTO | Unique identifier (primary key) |
-| `content` | TEXT | NO | - | The conversation message/utterance content |
-| `speaker` | TEXT | NO | - | Who said it: user or robot name |
-| `type` | TEXT | YES | NULL | Memory type: fact, context, code, preference, decision, question |
-| `category` | TEXT | YES | NULL | Optional category for organizing memories |
-| `importance` | DOUBLE PRECISION | YES | 1.0 | Importance score (0.0-1.0) for prioritizing recall |
-| `created_at` | TIMESTAMPTZ | YES | NOW() | When this memory was created |
-| `updated_at` | TIMESTAMPTZ | YES | NOW() | When this memory was last modified |
-| `last_accessed` | TIMESTAMPTZ | YES | NOW() | When this memory was last accessed |
-| `token_count` | INTEGER | YES | NULL | Number of tokens in the content (for context budget management) |
-| `in_working_memory` | BOOLEAN | YES | FALSE | Whether this memory is currently in working memory |
-| `robot_id` | BIGINT | NO | - | ID of the robot that owns this memory |
-| `embedding` | vector(2000) | YES | NULL | Vector embedding (max 2000 dimensions) for semantic search |
-| `embedding_dimension` | INTEGER | YES | NULL | Actual number of dimensions used in the embedding vector (max 2000) |
-
-**Indexes**:
-
-- `PRIMARY KEY` on `id`
-- `idx_nodes_robot_id` BTREE on `robot_id`
-- `idx_nodes_speaker` BTREE on `speaker`
-- `idx_nodes_type` BTREE on `type`
-- `idx_nodes_category` BTREE on `category`
-- `idx_nodes_created_at` BTREE on `created_at`
-- `idx_nodes_updated_at` BTREE on `updated_at`
-- `idx_nodes_last_accessed` BTREE on `last_accessed`
-- `idx_nodes_in_working_memory` BTREE on `in_working_memory`
-- `idx_nodes_embedding` HNSW on `embedding` using `vector_cosine_ops` (m=16, ef_construction=64)
-- `idx_nodes_content_gin` GIN on `to_tsvector('english', content)` for full-text search
-- `idx_nodes_content_trgm` GIN on `content` using `gin_trgm_ops` for fuzzy matching
-
-**Foreign Keys**:
-- `robot_id` references `robots(id)` ON DELETE CASCADE
-
-**Relationships**:
-- Many nodes belong to one robot (N:1)
-- Many nodes have many tags through nodes_tags (N:M)
-
-**Check Constraints**:
-- `check_embedding_dimension`: Ensures embedding_dimension is NULL or between 1 and 2000
-
----
-
-### tags
-
-The tags table stores unique hierarchical tag names for categorization.
+## Common Query Patterns
 
-**Purpose**: Provides flexible, hierarchical categorization using colon-separated namespaces (e.g., `database:postgresql:timescaledb`).
+### Finding Nodes for a Robot
 
 ```sql
-CREATE TABLE public.tags (
-    id bigint NOT NULL,
-    name text NOT NULL,
-    created_at timestamp with time zone DEFAULT CURRENT_TIMESTAMP
-);
-
-ALTER TABLE ONLY public.tags ALTER COLUMN id SET DEFAULT nextval('public.tags_id_seq'::regclass);
-ALTER TABLE ONLY public.tags ADD CONSTRAINT tags_pkey PRIMARY KEY (id);
+SELECT n.*
+FROM nodes n
+JOIN robot_nodes rn ON n.id = rn.node_id
+WHERE rn.robot_id = $1
+ORDER BY rn.last_remembered_at DESC;
 ```
 
-**Columns**:
-
-| Column | Type | Nullable | Default | Description |
-|--------|------|----------|---------|-------------|
-| `id` | BIGINT | NO | AUTO | Unique identifier (primary key) |
-| `name` | TEXT | NO | - | Hierarchical tag in format: root:level1:level2 (e.g., database:postgresql:timescaledb) |
-| `created_at` | TIMESTAMPTZ | YES | NOW() | When this tag was created |
-
-**Indexes**:
-- `PRIMARY KEY` on `id`
-- `idx_tags_name_unique` UNIQUE BTREE on `name`
-- `idx_tags_name_pattern` BTREE on `name` with `text_pattern_ops` for pattern matching
-
-**Relationships**:
-- Many tags belong to many nodes through nodes_tags (N:M)
-
-**Tag Hierarchy**:
+### Finding Robots that Share a Node
 
-Tags use colon-separated hierarchies for organization:
-- `programming:ruby:gems` - Programming > Ruby > Gems
-- `database:postgresql:extensions` - Database > PostgreSQL > Extensions
-- `ai:llm:embeddings` - AI > LLM > Embeddings
-
-This allows querying by prefix to find all related tags:
 ```sql
-SELECT * FROM tags WHERE name LIKE 'database:%';  -- All database-related tags
-SELECT * FROM tags WHERE name LIKE 'ai:llm:%';    -- All LLM-related tags
+SELECT r.*
+FROM robots r
+JOIN robot_nodes rn ON r.id = rn.robot_id
+WHERE rn.node_id = $1
+ORDER BY rn.first_remembered_at;
 ```
 
----
-
-### nodes_tags
-
-The nodes_tags join table implements the many-to-many relationship between nodes and tags.
-
-**Purpose**: Links nodes to tags, allowing each node to have multiple tags and each tag to be applied to multiple nodes.
+### Finding Frequently Remembered Content
 
 ```sql
-CREATE TABLE public.nodes_tags (
-    id bigint NOT NULL,
-    node_id bigint NOT NULL,
-    tag_id bigint NOT NULL,
-    created_at timestamp with time zone DEFAULT CURRENT_TIMESTAMP
-);
-
-ALTER TABLE ONLY public.nodes_tags ALTER COLUMN id SET DEFAULT nextval('public.node_tags_id_seq'::regclass);
-ALTER TABLE ONLY public.nodes_tags ADD CONSTRAINT node_tags_pkey PRIMARY KEY (id);
-
-ALTER TABLE ONLY public.nodes_tags
-    ADD CONSTRAINT fk_rails_b0b726ecf8 FOREIGN KEY (node_id) REFERENCES public.nodes(id) ON DELETE CASCADE;
-ALTER TABLE ONLY public.nodes_tags
-    ADD CONSTRAINT fk_rails_eccc99cec5 FOREIGN KEY (tag_id) REFERENCES public.tags(id) ON DELETE CASCADE;
+SELECT n.*, rn.remember_count, rn.first_remembered_at, rn.last_remembered_at
+FROM nodes n
+JOIN robot_nodes rn ON n.id = rn.node_id
+WHERE rn.robot_id = $1
+ORDER BY rn.remember_count DESC
+LIMIT 10;
 ```
 
-**Columns**:
-
-| Column | Type | Nullable | Default | Description |
-|--------|------|----------|---------|-------------|
-| `id` | BIGINT | NO | AUTO | Unique identifier (primary key) |
-| `node_id` | BIGINT | NO | - | ID of the node being tagged |
-| `tag_id` | BIGINT | NO | - | ID of the tag being applied |
-| `created_at` | TIMESTAMPTZ | YES | NOW() | When this association was created |
-
-**Indexes**:
-- `PRIMARY KEY` on `id`
-- `idx_node_tags_unique` UNIQUE BTREE on `(node_id, tag_id)` - Prevents duplicate associations
-- `idx_node_tags_node_id` BTREE on `node_id` - Fast lookups of tags for a node
-- `idx_node_tags_tag_id` BTREE on `tag_id` - Fast lookups of nodes for a tag
-
-**Foreign Keys**:
-- `node_id` references `nodes(id)` ON DELETE CASCADE
-- `tag_id` references `tags(id)` ON DELETE CASCADE
-
-**Cascade Behavior**:
-- When a node is deleted, all its tag associations are automatically removed
-- When a tag is deleted, all associations to that tag are automatically removed
-- The join table ensures referential integrity between nodes and tags
-
----
-
-## Common Query Patterns
-
 ### Finding Tags for a Node
 
 ```sql
 SELECT t.name
 FROM tags t
-JOIN nodes_tags nt ON t.id = nt.tag_id
+JOIN node_tags nt ON t.id = nt.tag_id
 WHERE nt.node_id = $1
 ORDER BY t.name;
 ```
@@ -392,7 +141,7 @@ ORDER BY t.name;
 ```sql
 SELECT n.*
 FROM nodes n
-JOIN nodes_tags nt ON n.id = nt.node_id
+JOIN node_tags nt ON n.id = nt.node_id
 JOIN tags t ON nt.tag_id = t.id
 WHERE t.name = 'database:postgresql'
 ORDER BY n.created_at DESC;
@@ -403,7 +152,7 @@ ORDER BY n.created_at DESC;
 ```sql
 SELECT n.*
 FROM nodes n
-JOIN nodes_tags nt ON n.id = nt.node_id
+JOIN node_tags nt ON n.id = nt.node_id
 JOIN tags t ON nt.tag_id = t.id
 WHERE t.name LIKE 'ai:llm:%'
 ORDER BY n.created_at DESC;
@@ -417,8 +166,8 @@ SELECT
     t2.name AS topic2,
     COUNT(DISTINCT nt1.node_id) AS shared_nodes
 FROM tags t1
-JOIN nodes_tags nt1 ON t1.id = nt1.tag_id
-JOIN nodes_tags nt2 ON nt1.node_id = nt2.node_id
+JOIN node_tags nt1 ON t1.id = nt1.tag_id
+JOIN node_tags nt2 ON nt1.node_id = nt2.node_id
 JOIN tags t2 ON nt2.tag_id = t2.id
 WHERE t1.name < t2.name
 GROUP BY t1.name, t2.name
@@ -431,7 +180,7 @@ ORDER BY shared_nodes DESC;
 ```sql
 SELECT n.*, n.embedding <=> $1::vector AS distance
 FROM nodes n
-JOIN nodes_tags nt ON n.id = nt.node_id
+JOIN node_tags nt ON n.id = nt.node_id
 JOIN tags t ON nt.tag_id = t.id
 WHERE t.name = 'programming:ruby'
   AND n.embedding IS NOT NULL
@@ -444,7 +193,7 @@ LIMIT 10;
 ```sql
 SELECT n.*, ts_rank(to_tsvector('english', n.content), query) AS rank
 FROM nodes n
-JOIN nodes_tags nt ON n.id = nt.node_id
+JOIN node_tags nt ON n.id = nt.node_id
 JOIN tags t ON nt.tag_id = t.id,
      to_tsquery('english', 'database & optimization') query
 WHERE to_tsvector('english', n.content) @@ query
@@ -453,6 +202,17 @@ ORDER BY rank DESC
 LIMIT 20;
 ```
 
+### Finding Content Shared by Multiple Robots
+
+```sql
+SELECT n.*, COUNT(DISTINCT rn.robot_id) AS robot_count
+FROM nodes n
+JOIN robot_nodes rn ON n.id = rn.node_id
+GROUP BY n.id
+HAVING COUNT(DISTINCT rn.robot_id) > 1
+ORDER BY robot_count DESC;
+```
+
 ---
 
 ## Database Optimization

data/docs/development/testing.md

@@ -147,15 +147,7 @@ ruby -r debug test/htm_test.rb
 
 ### Test File Layout
 
-```
-test/
-├── test_helper.rb              # Shared test configuration
-├── htm_test.rb                 # Unit tests for HTM class
-├── embedding_service_test.rb   # Unit tests for EmbeddingService
-├── integration_test.rb         # Integration tests
-└── fixtures/                   # Test data (future)
-    └── sample_memories.json
-```
+![Test Directory Structure](../assets/images/test-directory-structure.svg)
 
 ### Test File Template
 

data/docs/getting-started/index.md

@@ -0,0 +1,47 @@
+# Getting Started
+
+Welcome to HTM (Hierarchical Temporary Memory)! This section will help you get up and running quickly.
+
+## Overview
+
+HTM provides intelligent memory management for LLM-based applications (robots) with a two-tier architecture:
+
+- **Long-term Memory**: Durable PostgreSQL storage with vector embeddings for semantic search
+- **Working Memory**: Token-limited in-memory context for immediate LLM use
+
+## What You'll Learn
+
+<div class="grid cards" markdown>
+
+-   :material-download:{ .lg .middle } **Installation**
+
+    ---
+
+    Set up HTM in your Ruby project with all required dependencies including PostgreSQL, pgvector, and Ollama.
+
+    [:octicons-arrow-right-24: Install HTM](installation.md)
+
+-   :material-rocket-launch:{ .lg .middle } **Quick Start**
+
+    ---
+
+    Build your first memory-enabled robot in minutes with practical examples and code snippets.
+
+    [:octicons-arrow-right-24: Get started](quick-start.md)
+
+</div>
+
+## Prerequisites
+
+Before installing HTM, ensure you have:
+
+- **Ruby 3.1+** - HTM uses modern Ruby features
+- **PostgreSQL 14+** - With pgvector and pg_trgm extensions
+- **Ollama** (optional) - For local embedding generation
+
+## Next Steps
+
+1. **[Install HTM](installation.md)** - Set up the gem and database
+2. **[Quick Start](quick-start.md)** - Create your first memory-enabled robot
+3. **[Architecture Overview](../architecture/overview.md)** - Understand how HTM works
+4. **[Guides](../guides/index.md)** - Deep dive into specific features

data/docs/{installation.md → getting-started/installation.md}

@@ -453,8 +453,8 @@ psql $HTM_DBURL -c "
 Now that HTM is installed, you're ready to start building:
 
 1. **[Quick Start Guide](quick-start.md)**: Build your first HTM application in 5 minutes
-2. **[User Guide](guides/getting-started.md)**: Learn all HTM features in depth
-3. **[API Reference](api/htm.md)**: Explore the complete API documentation
+2. **[User Guide](../guides/getting-started.md)**: Learn all HTM features in depth
+3. **[API Reference](../api/htm.md)**: Explore the complete API documentation
 4. **[Examples](https://github.com/madbomber/htm/tree/main/examples)**: See real-world usage examples
 
 ## Getting Help

data/docs/{quick-start.md → getting-started/quick-start.md}

@@ -499,9 +499,9 @@ Congratulations! You've learned the basics of HTM. Here's what to explore next:
 
 ### Explore Advanced Features
 
-- **[User Guide](guides/getting-started.md)**: Deep dive into all HTM features
-- **[API Reference](api/htm.md)**: Complete API documentation
-- **[Architecture Guide](architecture/overview.md)**: Understand HTM's internals
+- **[User Guide](../guides/getting-started.md)**: Deep dive into all HTM features
+- **[API Reference](../api/htm.md)**: Complete API documentation
+- **[Architecture Guide](../architecture/overview.md)**: Understand HTM's internals
 
 ### Build Real Applications
 
@@ -648,8 +648,8 @@ memories = htm.recall(
 ## Additional Resources
 
 - **[Installation Guide](installation.md)**: Complete setup instructions
-- **[User Guide](guides/getting-started.md)**: Comprehensive feature documentation
-- **[API Reference](api/htm.md)**: Detailed API documentation
+- **[User Guide](../guides/getting-started.md)**: Comprehensive feature documentation
+- **[API Reference](../api/htm.md)**: Detailed API documentation
 - **[Examples](https://github.com/madbomber/htm/tree/main/examples)**: Real-world code examples
 
 Happy coding with HTM! 🚀