agentdb-learning

Name: agentdb-learning
Author: spencermarx

301stars24forks

How to install agentdb-learning?

spencermarx/open-code-review/agentdb-learning

$npx -y skills add spencermarx/open-code-review --skill agentdb-learning

Installs into the current project.

Summary

Create and train AI learning plugins with AgentDB's 9 reinforcement learning algorithms. Includes Decision Transformer, Q-Learning, SARSA, Actor-Critic, and more. Use when building self-learning agents, implementing RL, or optimizing agent behavior through experience.

Files · 1

View on GitHub

SKILL.md

1	# AgentDB Learning Plugins
2
3	## What This Skill Does
4
5	Provides access to 9 reinforcement learning algorithms via AgentDB's plugin system. Create, train, and deploy learning plugins for autonomous agents that improve through experience. Includes offline RL (Decision Transformer), value-based learning (Q-Learning), policy gradients (Actor-Critic), and advanced techniques.
6
7	Performance: Train models 10-100x faster with WASM-accelerated neural inference.
8
9	## Prerequisites
10
11	- Node.js 18+
12	- AgentDB v1.0.7+ (via agentic-flow)
13	- Basic understanding of reinforcement learning (recommended)
14
15	---
16
17	## Quick Start with CLI
18
19	### Create Learning Plugin
20
21	```bash
22	# Interactive wizard
23	npx agentdb@latest create-plugin
24
25	# Use specific template
26	npx agentdb@latest create-plugin -t decision-transformer -n my-agent
27
28	# Preview without creating
29	npx agentdb@latest create-plugin -t q-learning --dry-run
30
31	# Custom output directory
32	npx agentdb@latest create-plugin -t actor-critic -o ./plugins
33	```
34
35	### List Available Templates
36
37	```bash
38	# Show all plugin templates
39	npx agentdb@latest list-templates
40
41	# Available templates:
42	# - decision-transformer (sequence modeling RL - recommended)
43	# - q-learning (value-based learning)
44	# - sarsa (on-policy TD learning)
45	# - actor-critic (policy gradient with baseline)
46	# - curiosity-driven (exploration-based)
47	```
48
49	### Manage Plugins
50
51	```bash
52	# List installed plugins
53	npx agentdb@latest list-plugins
54
55	# Get plugin information
56	npx agentdb@latest plugin-info my-agent
57
58	# Shows: algorithm, configuration, training status
59	```
60
61	---
62
63	## Quick Start with API
64
65	```typescript
66	import { createAgentDBAdapter } from 'agentic-flow/reasoningbank';
67
68	// Initialize with learning enabled
69	const adapter = await createAgentDBAdapter({
70	dbPath: '.agentdb/learning.db',
71	enableLearning: true, // Enable learning plugins
72	enableReasoning: true,
73	cacheSize: 1000,
74	});
75
76	// Store training experience
77	await adapter.insertPattern({
78	id: '',
79	type: 'experience',
80	domain: 'game-playing',
81	pattern_data: JSON.stringify({
82	embedding: await computeEmbedding('state-action-reward'),
83	pattern: {
84	state: [0.1, 0.2, 0.3],
85	action: 2,
86	reward: 1.0,
87	next_state: [0.15, 0.25, 0.35],
88	done: false
89	}
90	}),
91	confidence: 0.9,
92	usage_count: 1,
93	success_count: 1,
94	created_at: Date.now(),
95	last_used: Date.now(),
96	});
97
98	// Train learning model
99	const metrics = await adapter.train({
100	epochs: 50,
101	batchSize: 32,
102	});
103
104	console.log('Training Loss:', metrics.loss);
105	console.log('Duration:', metrics.duration, 'ms');
106	```
107
108	---
109
110	## Available Learning Algorithms (9 Total)
111
112	### 1. Decision Transformer (Recommended)
113
114	Type: Offline Reinforcement Learning
115	Best For: Learning from logged experiences, imitation learning
116	Strengths: No online interaction needed, stable training
117
118	```bash
119	npx agentdb@latest create-plugin -t decision-transformer -n dt-agent
120	```
121
122	Use Cases:
123	- Learn from historical data
124	- Imitation learning from expert demonstrations
125	- Safe learning without environment interaction
126	- Sequence modeling tasks
127
128	Configuration:
129	```json
130	{
131	"algorithm": "decision-transformer",
132	"model_size": "base",
133	"context_length": 20,
134	"embed_dim": 128,
135	"n_heads": 8,
136	"n_layers": 6
137	}
138	```
139
140	### 2. Q-Learning
141
142	Type: Value-Based RL (Off-Policy)
143	Best For: Discrete action spaces, sample efficiency
144	Strengths: Proven, simple, works well for small/medium problems
145
146	```bash
147	npx agentdb@latest create-plugin -t q-learning -n q-agent
148	```
149
150	Use Cases:
151	- Grid worlds, board games
152	- Navigation tasks
153	- Resource allocation
154	- Discrete decision-making
155
156	Configuration:
157	```json
158	{
159	"algorithm": "q-learning",
160	"learning_rate": 0.001,
161	"gamma": 0.99,
162	"epsilon": 0.1,
163	"epsilon_decay": 0.995
164	}
165	```
166
167	### 3. SARSA
168
169	Type: Value-Based RL (On-Policy)
170	Best For: Safe exploration, risk-sensitive tasks
171	Strengths: More conservative than Q-Learning, better for safety
172
173	```bash
174	npx agentdb@latest create-plugin -t sarsa -n sarsa-agent
175	```
176
177	Use Cases:
178	- Safety-critical applications
179	- Risk-sensitive decision-making
180	- Online learning with exploration
181
182	Configuration:
183	```json
184	{
185	"algorithm": "sarsa",
186	"learning_rate": 0.001,
187	"gamma": 0.99,
188	"epsilon": 0.1
189	}
190	```
191
192	### 4. Actor-Critic
193
194	Type: Policy Gradient with Value Baseline
195	Best For: Continuous actions, variance reduction
196	Strengths: Stable, works for continuous/discrete actions
197
198	```bash
199	npx agentdb@latest create-plugin -t actor-critic -n ac-agent
200	```
201
202	Use Cases:
203	- Continuous control (robotics, simulations)
204	- Complex action spaces
205	- Multi-agent coordination
206
207	Configuration:
208	```json
209	{
210	"algorithm": "actor-critic",
211	"actor_lr": 0.0

People who liked this also viewed

6 picks

Type

agentdb-learning

301stars24forks

View on GitHub

How to install agentdb-learning?

spencermarx/open-code-review/agentdb-learning

$npx -y skills add spencermarx/open-code-review --skill agentdb-learning

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# AgentDB Learning Plugins
2
3	## What This Skill Does
4
5	Provides access to 9 reinforcement learning algorithms via AgentDB's plugin system. Create, train, and deploy learning plugins for autonomous agents that improve through experience. Includes offline RL (Decision Transformer), value-based learning (Q-Learning), policy gradients (Actor-Critic), and advanced techniques.
6
7	Performance: Train models 10-100x faster with WASM-accelerated neural inference.
8
9	## Prerequisites
10
11	- Node.js 18+
12	- AgentDB v1.0.7+ (via agentic-flow)
13	- Basic understanding of reinforcement learning (recommended)
14
15	---
16
17	## Quick Start with CLI
18
19	### Create Learning Plugin
20
21	```bash
22	# Interactive wizard
23	npx agentdb@latest create-plugin
24
25	# Use specific template
26	npx agentdb@latest create-plugin -t decision-transformer -n my-agent
27
28	# Preview without creating
29	npx agentdb@latest create-plugin -t q-learning --dry-run
30
31	# Custom output directory
32	npx agentdb@latest create-plugin -t actor-critic -o ./plugins
33	```
34
35	### List Available Templates
36
37	```bash
38	# Show all plugin templates
39	npx agentdb@latest list-templates
40
41	# Available templates:
42	# - decision-transformer (sequence modeling RL - recommended)
43	# - q-learning (value-based learning)
44	# - sarsa (on-policy TD learning)
45	# - actor-critic (policy gradient with baseline)
46	# - curiosity-driven (exploration-based)
47	```
48
49	### Manage Plugins
50
51	```bash
52	# List installed plugins
53	npx agentdb@latest list-plugins
54
55	# Get plugin information
56	npx agentdb@latest plugin-info my-agent
57
58	# Shows: algorithm, configuration, training status
59	```
60
61	---
62
63	## Quick Start with API
64
65	```typescript
66	import { createAgentDBAdapter } from 'agentic-flow/reasoningbank';
67
68	// Initialize with learning enabled
69	const adapter = await createAgentDBAdapter({
70	dbPath: '.agentdb/learning.db',
71	enableLearning: true, // Enable learning plugins
72	enableReasoning: true,
73	cacheSize: 1000,
74	});
75
76	// Store training experience
77	await adapter.insertPattern({
78	id: '',
79	type: 'experience',
80	domain: 'game-playing',
81	pattern_data: JSON.stringify({
82	embedding: await computeEmbedding('state-action-reward'),
83	pattern: {
84	state: [0.1, 0.2, 0.3],
85	action: 2,
86	reward: 1.0,
87	next_state: [0.15, 0.25, 0.35],
88	done: false
89	}
90	}),
91	confidence: 0.9,
92	usage_count: 1,
93	success_count: 1,
94	created_at: Date.now(),
95	last_used: Date.now(),
96	});
97
98	// Train learning model
99	const metrics = await adapter.train({
100	epochs: 50,
101	batchSize: 32,
102	});
103
104	console.log('Training Loss:', metrics.loss);
105	console.log('Duration:', metrics.duration, 'ms');
106	```
107
108	---
109
110	## Available Learning Algorithms (9 Total)
111
112	### 1. Decision Transformer (Recommended)
113
114	Type: Offline Reinforcement Learning
115	Best For: Learning from logged experiences, imitation learning
116	Strengths: No online interaction needed, stable training
117
118	```bash
119	npx agentdb@latest create-plugin -t decision-transformer -n dt-agent
120	```
121
122	Use Cases:
123	- Learn from historical data
124	- Imitation learning from expert demonstrations
125	- Safe learning without environment interaction
126	- Sequence modeling tasks
127
128	Configuration:
129	```json
130	{
131	"algorithm": "decision-transformer",
132	"model_size": "base",
133	"context_length": 20,
134	"embed_dim": 128,
135	"n_heads": 8,
136	"n_layers": 6
137	}
138	```
139
140	### 2. Q-Learning
141
142	Type: Value-Based RL (Off-Policy)
143	Best For: Discrete action spaces, sample efficiency
144	Strengths: Proven, simple, works well for small/medium problems
145
146	```bash
147	npx agentdb@latest create-plugin -t q-learning -n q-agent
148	```
149
150	Use Cases:
151	- Grid worlds, board games
152	- Navigation tasks
153	- Resource allocation
154	- Discrete decision-making
155
156	Configuration:
157	```json
158	{
159	"algorithm": "q-learning",
160	"learning_rate": 0.001,
161	"gamma": 0.99,
162	"epsilon": 0.1,
163	"epsilon_decay": 0.995
164	}
165	```
166
167	### 3. SARSA
168
169	Type: Value-Based RL (On-Policy)
170	Best For: Safe exploration, risk-sensitive tasks
171	Strengths: More conservative than Q-Learning, better for safety
172
173	```bash
174	npx agentdb@latest create-plugin -t sarsa -n sarsa-agent
175	```
176
177	Use Cases:
178	- Safety-critical applications
179	- Risk-sensitive decision-making
180	- Online learning with exploration
181
182	Configuration:
183	```json
184	{
185	"algorithm": "sarsa",
186	"learning_rate": 0.001,
187	"gamma": 0.99,
188	"epsilon": 0.1
189	}
190	```
191
192	### 4. Actor-Critic
193
194	Type: Policy Gradient with Value Baseline
195	Best For: Continuous actions, variance reduction
196	Strengths: Stable, works for continuous/discrete actions
197
198	```bash
199	npx agentdb@latest create-plugin -t actor-critic -n ac-agent
200	```
201
202	Use Cases:
203	- Continuous control (robotics, simulations)
204	- Complex action spaces
205	- Multi-agent coordination
206
207	Configuration:
208	```json
209	{
210	"algorithm": "actor-critic",
211	"actor_lr": 0.0

People who liked this also viewed

6 picks

Type