-
Notifications
You must be signed in to change notification settings - Fork 2
Expand file tree
/
Copy pathtutorial.html
More file actions
610 lines (518 loc) · 37.8 KB
/
tutorial.html
File metadata and controls
610 lines (518 loc) · 37.8 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Quickstart Tutorial — openrappter</title>
<meta name="description" content="From zero to a working openrappter agent in minutes. Follow along in TypeScript or Python.">
<meta property="og:title" content="Quickstart Tutorial — openrappter">
<meta property="og:description" content="From zero to a working agent in minutes. Follow along in TypeScript or Python.">
<meta property="og:type" content="website">
<meta name="twitter:card" content="summary_large_image">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link href="https://fonts.googleapis.com/css2?family=JetBrains+Mono:wght@400;500;600;700&family=Inter:wght@400;500;600;700;800;900&display=swap" rel="stylesheet">
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<!-- Nav -->
<nav>
<div class="nav-container">
<a href="./" class="logo">openrappter <span class="logo-badge">v1.9.1</span></a>
<button class="mobile-menu-btn">☰</button>
<div class="nav-links">
<a href="./docs.html">Docs</a>
<a href="./architecture.html">Architecture</a>
<a href="./tutorial.html">Tutorial</a>
<a href="./changelog.html">Changelog</a>
</div>
<div class="nav-cta">
<a href="https://github.com/kody-w/openrappter" class="btn btn-ghost">
<svg width="18" height="18" viewBox="0 0 24 24" fill="currentColor"><path d="M12 0C5.37 0 0 5.37 0 12c0 5.31 3.435 9.795 8.205 11.385.6.105.825-.255.825-.57 0-.285-.015-1.23-.015-2.235-3.015.555-3.795-.735-4.035-1.41-.135-.345-.72-1.41-1.23-1.695-.42-.225-1.02-.78-.015-.795.945-.015 1.62.87 1.845 1.23 1.08 1.815 2.805 1.305 3.495.99.105-.78.42-1.305.765-1.605-2.67-.3-5.46-1.335-5.46-5.925 0-1.305.465-2.385 1.23-3.225-.12-.3-.54-1.53.12-3.18 0 0 1.005-.315 3.3 1.23.96-.27 1.98-.405 3-.405s2.04.135 3 .405c2.295-1.56 3.3-1.23 3.3-1.23.66 1.65.24 2.88.12 3.18.765.84 1.23 1.905 1.23 3.225 0 4.605-2.805 5.625-5.475 5.925.435.375.81 1.095.81 2.22 0 1.605-.015 2.895-.015 3.3 0 .315.225.69.825.57A12.02 12.02 0 0024 12c0-6.63-5.37-12-12-12z"/></svg>
GitHub
</a>
</div>
</div>
</nav>
<!-- Hero -->
<section class="hero">
<div class="hero-bg"></div>
<h1><span class="highlight">Quickstart</span> Tutorial</h1>
<p class="hero-subtitle">From zero to a working agent in minutes. Follow along in TypeScript or Python.</p>
</section>
<!-- Main layout -->
<div style="display:flex;justify-content:center;gap:3rem;max-width:1100px;margin:0 auto;padding:0 2rem 6rem;">
<!-- Progress sidebar -->
<div class="progress-nav">
<ol>
<li><a href="#step-1">1. Install</a></li>
<li><a href="#step-2">2. Verify</a></li>
<li><a href="#step-3">3. CLI Basics</a></li>
<li><a href="#step-4">4. Memory</a></li>
<li><a href="#step-5">5. Create Agent</a></li>
<li><a href="#step-6">6. Test It</a></li>
<li><a href="#step-7">7. Data Sloshing</a></li>
<li><a href="#step-8">8. Chain Agents</a></li>
<li><a href="#step-9">9. Multi-Agent</a></li>
</ol>
</div>
<!-- Content -->
<main style="max-width:800px;width:100%;">
<!-- Step 1: Install -->
<div class="step" id="step-1">
<div class="step-number">1</div>
<h2>Install</h2>
<p>There are three ways to get openrappter running. Pick whichever fits your workflow.</p>
<p><strong>Option A — one-line installer (recommended):</strong></p>
<pre><code><span class="cmd">curl -fsSL https://kody-w.github.io/openrappter/install.sh | bash</span></code></pre>
<p>The script detects your platform, clones the repo, installs dependencies, builds the TypeScript runtime, and adds <code>openrappter</code> to your PATH.</p>
<p><strong>Option B — manual clone:</strong></p>
<pre><code><span class="comment"># TypeScript runtime</span>
<span class="cmd">git clone https://github.com/kody-w/openrappter.git</span>
<span class="cmd">cd openrappter/typescript</span>
<span class="cmd">npm install</span>
<span class="cmd">npm run build</span>
<span class="comment"># Python runtime (optional, mirrors TypeScript)</span>
<span class="cmd">cd ../python</span>
<span class="cmd">pip install -e .</span></code></pre>
<p><strong>Option C — teach your existing agent:</strong> If you already have an AI assistant, paste the <a href="https://raw.githubusercontent.com/kody-w/openrappter/main/skills.md">skills.md</a> link into your conversation and ask it to set up openrappter for you. It can handle the full install and scaffold new agents.</p>
<div class="info-box">
<strong>Prerequisites:</strong> Node.js 18+ is required for the TypeScript runtime. Python 3.10+ is required for the Python runtime. Both runtimes implement the same agent contract and are interchangeable.
</div>
</div>
<!-- Step 2: Verify -->
<div class="step" id="step-2">
<div class="step-number">2</div>
<h2>Verify</h2>
<p>Confirm the installation is healthy before writing any code.</p>
<pre><code><span class="comment"># Check version — should print v1.9.1</span>
<span class="cmd">openrappter --version</span></code></pre>
<div class="terminal">
<div class="terminal-header">
<span class="terminal-dot red"></span>
<span class="terminal-dot yellow"></span>
<span class="terminal-dot green"></span>
<span class="terminal-title">openrappter --version</span>
</div>
<div class="terminal-body">
<div class="terminal-line">
<span class="terminal-prompt">$</span>
<span class="terminal-command">openrappter --version</span>
</div>
<div class="terminal-output">openrappter v1.9.1</div>
</div>
</div>
<pre><code><span class="comment"># List all auto-discovered agents</span>
<span class="cmd">openrappter --list-agents</span></code></pre>
<div class="terminal">
<div class="terminal-header">
<span class="terminal-dot red"></span>
<span class="terminal-dot yellow"></span>
<span class="terminal-dot green"></span>
<span class="terminal-title">openrappter --list-agents</span>
</div>
<div class="terminal-body">
<div class="terminal-line">
<span class="terminal-prompt">$</span>
<span class="terminal-command">openrappter --list-agents</span>
</div>
<div class="terminal-output">Available agents (3):</div>
<div class="terminal-output"> BasicAgent — Abstract base with data sloshing</div>
<div class="terminal-output"> ShellAgent — Shell commands, file read/write/list</div>
<div class="terminal-output"> MemoryAgent — Memory storage and retrieval</div>
</div>
</div>
<p>If you see agents listed, the runtime discovered your <code>src/agents/</code> (TypeScript) or <code>python/openrappter/agents/</code> (Python) directory successfully. Any custom agents you add there will appear here automatically.</p>
</div>
<!-- Step 3: CLI Basics -->
<div class="step" id="step-3">
<div class="step-number">3</div>
<h2>CLI Basics</h2>
<p>openrappter has three primary usage patterns from the command line.</p>
<p><strong>Interactive REPL mode</strong> — drop into a persistent session where the agent remembers context across messages:</p>
<pre><code><span class="cmd">openrappter</span></code></pre>
<p><strong>Single task mode</strong> — run one task and exit, useful for scripting:</p>
<pre><code><span class="cmd">openrappter --task "list files in current directory"</span></code></pre>
<p><strong>Specify an agent directly</strong> — bypass the router and run a named agent:</p>
<pre><code><span class="cmd">openrappter --agent shell "what's in my home directory?"</span>
<span class="cmd">openrappter --exec ShellAgent "read package.json"</span></code></pre>
<div class="terminal">
<div class="terminal-header">
<span class="terminal-dot red"></span>
<span class="terminal-dot yellow"></span>
<span class="terminal-dot green"></span>
<span class="terminal-title">single task example</span>
</div>
<div class="terminal-body">
<div class="terminal-line">
<span class="terminal-prompt">$</span>
<span class="terminal-command">openrappter --task "list files in current directory"</span>
</div>
<div class="terminal-output">Executing ShellAgent...</div>
<div class="terminal-output">README.md package.json src/ dist/ node_modules/</div>
<div class="terminal-success">Done.</div>
</div>
</div>
<p>Other useful flags: <code>--status</code> shows loaded agents and runtime health, <code>--version</code> prints the version, and <code>--help</code> lists all available options.</p>
</div>
<!-- Step 4: Memory -->
<div class="step" id="step-4">
<div class="step-number">4</div>
<h2>Memory</h2>
<p>openrappter has first-class persistent memory that survives across sessions. Anything you store is written locally and recalled automatically in future queries.</p>
<p><strong>Store a fact:</strong></p>
<pre><code><span class="cmd">openrappter "remember that I prefer dark mode"</span></code></pre>
<div class="terminal">
<div class="terminal-header">
<span class="terminal-dot red"></span>
<span class="terminal-dot yellow"></span>
<span class="terminal-dot green"></span>
<span class="terminal-title">storing a memory</span>
</div>
<div class="terminal-body">
<div class="terminal-line">
<span class="terminal-prompt">$</span>
<span class="terminal-command">openrappter "remember that I prefer dark mode"</span>
</div>
<div class="terminal-success">Got it. I'll remember: "I prefer dark mode" (id: a3f1c9)</div>
</div>
</div>
<p><strong>Recall facts:</strong></p>
<pre><code><span class="cmd">openrappter "what do you know about my preferences?"</span></code></pre>
<div class="terminal">
<div class="terminal-header">
<span class="terminal-dot red"></span>
<span class="terminal-dot yellow"></span>
<span class="terminal-dot green"></span>
<span class="terminal-title">recalling memory</span>
</div>
<div class="terminal-body">
<div class="terminal-line">
<span class="terminal-prompt">$</span>
<span class="terminal-command">openrappter "what do you know about my preferences?"</span>
</div>
<div class="terminal-output">Here's what I remember about your preferences:</div>
<div class="terminal-output"> - I prefer dark mode (stored 2 minutes ago)</div>
</div>
</div>
<p>Memory entries are stored at <code>~/.openrappter/memory.json</code> as structured JSON. Each entry contains an ID, content, tags, created timestamp, and an access counter that helps the agent prioritize frequently recalled facts.</p>
<div class="info-box">
<strong>Local-first:</strong> All memory is stored on your machine. No data leaves your system. You can inspect, back up, or wipe the file at any time with <code>cat ~/.openrappter/memory.json</code> or <code>rm ~/.openrappter/memory.json</code>.
</div>
</div>
<!-- Step 5: Create Agent -->
<div class="step" id="step-5">
<div class="step-number">5</div>
<h2>Create a Custom Agent</h2>
<p>Every agent in openrappter follows the single-file pattern: one file, one agent. The metadata contract, documentation, and implementation all live together using native language constructs — no YAML, no config files, no magic parsing.</p>
<p>Create a <strong>GreeterAgent</strong> that returns a time-aware greeting and passes data downstream via <code>data_slush</code>.</p>
<div class="code-tabs">
<div class="code-tabs-header">
<button class="code-tab-btn active" onclick="switchTab(this, 'ts-step5')">TypeScript</button>
<button class="code-tab-btn" onclick="switchTab(this, 'py-step5')">Python</button>
</div>
<div class="code-tab-content active" id="ts-step5">
<pre><code><span class="comment">// typescript/src/agents/GreeterAgent.ts</span>
<span class="kw">import</span> { BasicAgent } <span class="kw">from</span> <span class="str">'./BasicAgent.js'</span>;
<span class="kw">import type</span> { AgentMetadata } <span class="kw">from</span> <span class="str">'./types.js'</span>;
<span class="kw">export class</span> GreeterAgent <span class="kw">extends</span> BasicAgent {
<span class="fn">constructor</span>() {
<span class="kw">const</span> metadata: AgentMetadata = {
name: <span class="str">'GreeterAgent'</span>,
description: <span class="str">'Greets users by name with a fun fact'</span>,
parameters: {
type: <span class="str">'object'</span>,
properties: {
name: { type: <span class="str">'string'</span>, description: <span class="str">'Name to greet'</span> }
},
required: [<span class="str">'name'</span>]
}
};
<span class="fn">super</span>(<span class="str">'GreeterAgent'</span>, metadata);
}
<span class="kw">async</span> <span class="fn">perform</span>(kwargs: Record<string, unknown>) {
<span class="kw">const</span> name = kwargs.name <span class="kw">as</span> string;
<span class="kw">const</span> hour = <span class="kw">new</span> <span class="fn">Date</span>().getHours();
<span class="kw">const</span> greeting = hour < 12 ? <span class="str">'Good morning'</span> : hour < 18 ? <span class="str">'Good afternoon'</span> : <span class="str">'Good evening'</span>;
<span class="kw">return</span> {
message: <span class="str">`${greeting}, ${name}!`</span>,
data_slush: { greeted_user: name, time_of_day: greeting }
};
}
}</code></pre>
</div>
<div class="code-tab-content" id="py-step5">
<pre><code><span class="comment"># python/openrappter/agents/greeter_agent.py</span>
<span class="kw">from</span> openrappter.agents.basic_agent <span class="kw">import</span> BasicAgent
<span class="kw">from</span> datetime <span class="kw">import</span> datetime
<span class="kw">class</span> GreeterAgent(BasicAgent):
<span class="kw">def</span> <span class="fn">__init__</span>(self):
self.name = <span class="str">'GreeterAgent'</span>
self.metadata = {
<span class="str">"name"</span>: self.name,
<span class="str">"description"</span>: <span class="str">"Greets users by name with a fun fact"</span>,
<span class="str">"parameters"</span>: {
<span class="str">"type"</span>: <span class="str">"object"</span>,
<span class="str">"properties"</span>: {
<span class="str">"name"</span>: {<span class="str">"type"</span>: <span class="str">"string"</span>, <span class="str">"description"</span>: <span class="str">"Name to greet"</span>}
},
<span class="str">"required"</span>: [<span class="str">"name"</span>]
}
}
<span class="fn">super</span>().__init__(name=self.name, metadata=self.metadata)
<span class="kw">def</span> <span class="fn">perform</span>(self, **kwargs):
name = kwargs.<span class="fn">get</span>(<span class="str">'name'</span>, <span class="str">'World'</span>)
hour = datetime.<span class="fn">now</span>().hour
greeting = <span class="str">'Good morning'</span> <span class="kw">if</span> hour < 12 <span class="kw">else</span> <span class="str">'Good afternoon'</span> <span class="kw">if</span> hour < 18 <span class="kw">else</span> <span class="str">'Good evening'</span>
<span class="kw">return</span> {
<span class="str">"message"</span>: <span class="kw">f</span><span class="str">"{greeting}, {name}!"</span>,
<span class="str">"data_slush"</span>: {<span class="str">"greeted_user"</span>: name, <span class="str">"time_of_day"</span>: greeting}
}</code></pre>
</div>
</div>
<p>Drop the file into the agents directory. The framework auto-discovers all files matching <code>*Agent.ts</code> (TypeScript) or <code>*_agent.py</code> (Python) — no registration step required.</p>
</div>
<!-- Step 6: Test It -->
<div class="step" id="step-6">
<div class="step-number">6</div>
<h2>Test It</h2>
<p>After saving the file, confirm the agent is discovered and invoke it directly by name.</p>
<pre><code><span class="comment"># Verify it was discovered</span>
<span class="cmd">openrappter --list-agents</span>
<span class="comment"># Should now include GreeterAgent in the list</span>
<span class="comment"># Invoke it directly</span>
<span class="cmd">openrappter --agent greeter "greet Alice"</span></code></pre>
<div class="terminal">
<div class="terminal-header">
<span class="terminal-dot red"></span>
<span class="terminal-dot yellow"></span>
<span class="terminal-dot green"></span>
<span class="terminal-title">invoking GreeterAgent</span>
</div>
<div class="terminal-body">
<div class="terminal-line">
<span class="terminal-prompt">$</span>
<span class="terminal-command">openrappter --agent greeter "greet Alice"</span>
</div>
<div class="terminal-output">Executing GreeterAgent...</div>
<div class="terminal-output">Data sloshing: injecting temporal + memory context</div>
<div class="terminal-success">Agent response:</div>
<div class="terminal-output">{</div>
<div class="terminal-output"> "message": "Good afternoon, Alice!",</div>
<div class="terminal-output"> "data_slush": {</div>
<div class="terminal-output"> "greeted_user": "Alice",</div>
<div class="terminal-output"> "time_of_day": "Good afternoon"</div>
<div class="terminal-output"> }</div>
<div class="terminal-output">}</div>
</div>
</div>
<p>The greeting matches the current time of day because the <code>perform()</code> method reads the system clock. In the next step, you will see how to get this and much richer context from the data sloshing pipeline instead.</p>
<p>You can also write a Vitest test alongside the agent (TypeScript) or a pytest file (Python) to validate behavior in CI:</p>
<pre><code><span class="comment">// typescript/src/__tests__/GreeterAgent.test.ts</span>
<span class="kw">import</span> { describe, it, expect } <span class="kw">from</span> <span class="str">'vitest'</span>;
<span class="kw">import</span> { GreeterAgent } <span class="kw">from</span> <span class="str">'../agents/GreeterAgent.js'</span>;
describe(<span class="str">'GreeterAgent'</span>, () => {
it(<span class="str">'returns a greeting with the provided name'</span>, <span class="kw">async</span> () => {
<span class="kw">const</span> agent = <span class="kw">new</span> <span class="fn">GreeterAgent</span>();
<span class="kw">const</span> result = <span class="kw">await</span> agent.<span class="fn">perform</span>({ name: <span class="str">'Alice'</span> });
expect(result.message).<span class="fn">toContain</span>(<span class="str">'Alice'</span>);
expect(result.data_slush.greeted_user).<span class="fn">toBe</span>(<span class="str">'Alice'</span>);
});
});</code></pre>
</div>
<!-- Step 7: Data Sloshing -->
<div class="step" id="step-7">
<div class="step-number">7</div>
<h2>Data Sloshing</h2>
<p>Before <code>perform()</code> is ever called, the framework automatically gathers context signals and synthesizes an <em>orientation</em> for your agent. This is data sloshing — implicit context enrichment that happens on every execution without any work from you.</p>
<p>The signals are accessible in <code>perform()</code> via <code>getSignal()</code> (TypeScript) or <code>get_signal()</code> (Python) using dot-notation paths:</p>
<div class="code-tabs">
<div class="code-tabs-header">
<button class="code-tab-btn active" onclick="switchTab(this, 'ts-step7')">TypeScript</button>
<button class="code-tab-btn" onclick="switchTab(this, 'py-step7')">Python</button>
</div>
<div class="code-tab-content active" id="ts-step7">
<pre><code><span class="kw">async</span> <span class="fn">perform</span>(kwargs: Record<string, unknown>) {
<span class="comment">// Temporal awareness — injected before perform() runs</span>
<span class="kw">const</span> timeOfDay = <span class="kw">this</span>.<span class="fn">getSignal</span><string>(<span class="str">'temporal.time_of_day'</span>, <span class="str">'morning'</span>);
<span class="kw">const</span> isWeekend = <span class="kw">this</span>.<span class="fn">getSignal</span><boolean>(<span class="str">'temporal.is_weekend'</span>, <span class="kw">false</span>);
<span class="kw">const</span> fiscalPeriod = <span class="kw">this</span>.<span class="fn">getSignal</span><string>(<span class="str">'temporal.fiscal_period'</span>);
<span class="comment">// Synthesized orientation — how confident is the agent?</span>
<span class="kw">const</span> confidence = <span class="kw">this</span>.<span class="fn">getSignal</span><string>(<span class="str">'orientation.confidence'</span>, <span class="str">'medium'</span>);
<span class="kw">const</span> approach = <span class="kw">this</span>.<span class="fn">getSignal</span><string>(<span class="str">'orientation.approach'</span>);
<span class="kw">if</span> (confidence === <span class="str">'low'</span>) {
<span class="kw">return</span> { status: <span class="str">'clarify'</span>, message: <span class="str">'Could you be more specific?'</span> };
}
<span class="comment">// Query signals — understand what the user is asking</span>
<span class="kw">const</span> isQuestion = <span class="kw">this</span>.<span class="fn">getSignal</span><boolean>(<span class="str">'query_signals.is_question'</span>, <span class="kw">false</span>);
<span class="kw">const</span> specificity = <span class="kw">this</span>.<span class="fn">getSignal</span><string>(<span class="str">'query_signals.specificity'</span>);
<span class="comment">// Memory echoes — recent relevant memories surfaced automatically</span>
<span class="kw">const</span> echoes = <span class="kw">this</span>.<span class="fn">getSignal</span><any[]>(<span class="str">'memory_echoes'</span>, []);
<span class="kw">if</span> (echoes.length > 0) {
<span class="kw">const</span> lastEcho = echoes[0]; <span class="comment">// most relevant past interaction</span>
}
<span class="comment">// Upstream slush — curated signals from the previous agent in a chain</span>
<span class="kw">const</span> upstream = <span class="kw">this</span>.context.upstream_slush ?? {};
}</code></pre>
</div>
<div class="code-tab-content" id="py-step7">
<pre><code><span class="kw">def</span> <span class="fn">perform</span>(self, **kwargs):
<span class="comment"># Temporal awareness — injected before perform() runs</span>
time_of_day = self.<span class="fn">get_signal</span>(<span class="str">'temporal.time_of_day'</span>, <span class="str">'morning'</span>)
is_weekend = self.<span class="fn">get_signal</span>(<span class="str">'temporal.is_weekend'</span>, <span class="kw">False</span>)
fiscal_period = self.<span class="fn">get_signal</span>(<span class="str">'temporal.fiscal_period'</span>)
<span class="comment"># Synthesized orientation — how confident is the agent?</span>
confidence = self.<span class="fn">get_signal</span>(<span class="str">'orientation.confidence'</span>, <span class="str">'medium'</span>)
approach = self.<span class="fn">get_signal</span>(<span class="str">'orientation.approach'</span>)
<span class="kw">if</span> confidence == <span class="str">'low'</span>:
<span class="kw">return</span> {<span class="str">"status"</span>: <span class="str">"clarify"</span>, <span class="str">"message"</span>: <span class="str">"Could you be more specific?"</span>}
<span class="comment"># Query signals — understand what the user is asking</span>
is_question = self.<span class="fn">get_signal</span>(<span class="str">'query_signals.is_question'</span>, <span class="kw">False</span>)
specificity = self.<span class="fn">get_signal</span>(<span class="str">'query_signals.specificity'</span>)
<span class="comment"># Memory echoes — recent relevant memories surfaced automatically</span>
echoes = self.<span class="fn">get_signal</span>(<span class="str">'memory_echoes'</span>, [])
<span class="kw">if</span> echoes:
last_echo = echoes[0] <span class="comment"># most relevant past interaction</span>
<span class="comment"># Upstream slush — curated signals from the previous agent in a chain</span>
upstream = self.context.<span class="fn">get</span>(<span class="str">'upstream_slush'</span>, {})</code></pre>
</div>
</div>
<p>All available signal namespaces: <code>temporal.*</code>, <code>memory_echoes</code>, <code>query_signals.*</code>, <code>behavioral_hints.*</code>, <code>orientation.*</code>, and <code>upstream_slush.*</code>. See the <a href="./architecture.html">Architecture page</a> for the full signal reference.</p>
<div class="info-box">
<strong>Data Slush:</strong> When your agent returns a <code>data_slush</code> key in its output, the framework extracts those values and stores them in <code>lastDataSlush</code> (TypeScript) or <code>last_data_slush</code> (Python). The next agent in the chain can receive this as <code>upstream_slush</code> — enabling LLM-free, deterministic agent-to-agent pipelines.
</div>
</div>
<!-- Step 8: Chain Agents -->
<div class="step" id="step-8">
<div class="step-number">8</div>
<h2>Chain Agents</h2>
<p>Agent chaining lets you build pipelines where one agent's output becomes the next agent's context. Agent A returns <code>data_slush</code> containing curated signals; Agent B receives that as <code>upstream_slush</code> via its <code>execute()</code> call.</p>
<div class="code-tabs">
<div class="code-tabs-header">
<button class="code-tab-btn active" onclick="switchTab(this, 'ts-step8')">TypeScript</button>
<button class="code-tab-btn" onclick="switchTab(this, 'py-step8')">Python</button>
</div>
<div class="code-tab-content active" id="ts-step8">
<pre><code><span class="kw">import</span> { GreeterAgent } <span class="kw">from</span> <span class="str">'./agents/GreeterAgent.js'</span>;
<span class="kw">import</span> { FollowUpAgent } <span class="kw">from</span> <span class="str">'./agents/FollowUpAgent.js'</span>;
<span class="comment">// Step 1: Agent A runs and populates lastDataSlush</span>
<span class="kw">const</span> greeter = <span class="kw">new</span> <span class="fn">GreeterAgent</span>();
<span class="kw">await</span> greeter.<span class="fn">execute</span>({ query: <span class="str">'greet Alice'</span>, name: <span class="str">'Alice'</span> });
<span class="comment">// greeter.lastDataSlush is now:</span>
<span class="comment">// { greeted_user: "Alice", time_of_day: "Good afternoon" }</span>
<span class="comment">// Step 2: Agent B receives A's slush as upstream context</span>
<span class="kw">const</span> followUp = <span class="kw">new</span> <span class="fn">FollowUpAgent</span>();
<span class="kw">const</span> result = <span class="kw">await</span> followUp.<span class="fn">execute</span>({
query: <span class="str">'suggest something for Alice'</span>,
upstream_slush: greeter.lastDataSlush <span class="comment">// pass A's slush to B</span>
});
<span class="comment">// Inside FollowUpAgent.perform():</span>
<span class="comment">// const greetedUser = this.getSignal('upstream_slush.greeted_user');</span>
<span class="comment">// const timeOfDay = this.getSignal('upstream_slush.time_of_day');</span></code></pre>
</div>
<div class="code-tab-content" id="py-step8">
<pre><code><span class="kw">from</span> openrappter.agents.greeter_agent <span class="kw">import</span> GreeterAgent
<span class="kw">from</span> openrappter.agents.follow_up_agent <span class="kw">import</span> FollowUpAgent
<span class="comment"># Step 1: Agent A runs and populates last_data_slush</span>
greeter = <span class="fn">GreeterAgent</span>()
greeter.<span class="fn">execute</span>(query=<span class="str">'greet Alice'</span>, name=<span class="str">'Alice'</span>)
<span class="comment"># greeter.last_data_slush is now:</span>
<span class="comment"># {"greeted_user": "Alice", "time_of_day": "Good afternoon"}</span>
<span class="comment"># Step 2: Agent B receives A's slush as upstream context</span>
follow_up = <span class="fn">FollowUpAgent</span>()
result = follow_up.<span class="fn">execute</span>(
query=<span class="str">'suggest something for Alice'</span>,
upstream_slush=greeter.last_data_slush <span class="comment"># pass A's slush to B</span>
)
<span class="comment"># Inside FollowUpAgent.perform():</span>
<span class="comment"># greeted_user = self.get_signal('upstream_slush.greeted_user')</span>
<span class="comment"># time_of_day = self.get_signal('upstream_slush.time_of_day')</span></code></pre>
</div>
</div>
<p>This pattern is the foundation of LLM-free agent pipelines. Agent A extracts structured facts; Agent B reasons from them — no large language model required in the middle. The <code>data_slush</code> key acts as a typed handoff contract between agents.</p>
<p>For more complex graphs — fan-out, fan-in, conditional routing — see the multi-agent patterns in the next step.</p>
</div>
<!-- Step 9: Multi-Agent Patterns -->
<div class="step" id="step-9">
<div class="step-number">9</div>
<h2>Multi-Agent Patterns</h2>
<p>openrappter ships three built-in orchestration primitives for coordinating multiple agents: <code>BroadcastManager</code>, <code>AgentRouter</code>, and <code>SubAgent</code>. These cover the vast majority of multi-agent topologies.</p>
<p><strong>BroadcastManager</strong> — send a message to multiple agents simultaneously and collect results:</p>
<div class="code-tabs">
<div class="code-tabs-header">
<button class="code-tab-btn active" onclick="switchTab(this, 'ts-broadcast')">TypeScript</button>
<button class="code-tab-btn" onclick="switchTab(this, 'py-broadcast')">Python</button>
</div>
<div class="code-tab-content active" id="ts-broadcast">
<pre><code><span class="kw">import</span> { BroadcastManager } <span class="kw">from</span> <span class="str">'./agents/broadcast.js'</span>;
<span class="kw">import</span> { GreeterAgent } <span class="kw">from</span> <span class="str">'./agents/GreeterAgent.js'</span>;
<span class="kw">import</span> { ShellAgent } <span class="kw">from</span> <span class="str">'./agents/ShellAgent.js'</span>;
<span class="comment">// 'all' — wait for every agent to complete</span>
<span class="kw">const</span> broadcast = <span class="kw">new</span> <span class="fn">BroadcastManager</span>([<span class="kw">new</span> <span class="fn">GreeterAgent</span>(), <span class="kw">new</span> <span class="fn">ShellAgent</span>()], <span class="str">'all'</span>);
<span class="kw">const</span> results = <span class="kw">await</span> broadcast.<span class="fn">execute</span>({ query: <span class="str">'hello'</span> });
<span class="comment">// 'race' — take the first agent to respond</span>
<span class="kw">const</span> race = <span class="kw">new</span> <span class="fn">BroadcastManager</span>([agentA, agentB, agentC], <span class="str">'race'</span>);
<span class="kw">const</span> first = <span class="kw">await</span> race.<span class="fn">execute</span>({ query: <span class="str">'who can answer this fastest?'</span> });
<span class="comment">// 'fallback' — try agents in order until one succeeds</span>
<span class="kw">const</span> fallback = <span class="kw">new</span> <span class="fn">BroadcastManager</span>([primaryAgent, backupAgent], <span class="str">'fallback'</span>);
<span class="kw">const</span> safe = <span class="kw">await</span> fallback.<span class="fn">execute</span>({ query: <span class="str">'try primary, then backup'</span> });</code></pre>
</div>
<div class="code-tab-content" id="py-broadcast">
<pre><code><span class="comment"># Python BroadcastManager mirrors the TypeScript API</span>
<span class="kw">from</span> openrappter.agents.broadcast <span class="kw">import</span> BroadcastManager
<span class="kw">from</span> openrappter.agents.greeter_agent <span class="kw">import</span> GreeterAgent
<span class="kw">from</span> openrappter.agents.shell_agent <span class="kw">import</span> ShellAgent
<span class="comment"># 'all' — wait for every agent to complete</span>
broadcast = <span class="fn">BroadcastManager</span>([<span class="fn">GreeterAgent</span>(), <span class="fn">ShellAgent</span>()], mode=<span class="str">'all'</span>)
results = broadcast.<span class="fn">execute</span>(query=<span class="str">'hello'</span>)
<span class="comment"># 'race' — take the first agent to respond</span>
race = <span class="fn">BroadcastManager</span>([agent_a, agent_b, agent_c], mode=<span class="str">'race'</span>)
first = race.<span class="fn">execute</span>(query=<span class="str">'who can answer this fastest?'</span>)
<span class="comment"># 'fallback' — try agents in order until one succeeds</span>
fallback = <span class="fn">BroadcastManager</span>([primary_agent, backup_agent], mode=<span class="str">'fallback'</span>)
safe = fallback.<span class="fn">execute</span>(query=<span class="str">'try primary, then backup'</span>)</code></pre>
</div>
</div>
<p><strong>AgentRouter</strong> — route messages to different agents based on rules (sender, channel, group, or pattern matching), with priority and session key isolation:</p>
<pre><code><span class="kw">import</span> { AgentRouter } <span class="kw">from</span> <span class="str">'./agents/router.js'</span>;
<span class="kw">const</span> router = <span class="kw">new</span> <span class="fn">AgentRouter</span>();
router.<span class="fn">addRule</span>({ pattern: <span class="str">/^greet/i</span>, agent: <span class="kw">new</span> <span class="fn">GreeterAgent</span>(), priority: 10 });
router.<span class="fn">addRule</span>({ pattern: <span class="str">/^run|^exec/i</span>, agent: <span class="kw">new</span> <span class="fn">ShellAgent</span>(), priority: 5 });
router.<span class="fn">addRule</span>({ channel: <span class="str">'slack'</span>, agent: <span class="kw">new</span> <span class="fn">SlackAgent</span>(), priority: 1 });
<span class="kw">const</span> result = <span class="kw">await</span> router.<span class="fn">route</span>({ query: <span class="str">'greet the team'</span>, channel: <span class="str">'cli'</span> });</code></pre>
<p><strong>SubAgent</strong> — invoke nested agents from within a parent agent's <code>perform()</code> method, with configurable depth limits and built-in loop detection to prevent infinite recursion:</p>
<pre><code><span class="kw">import</span> { SubAgent } <span class="kw">from</span> <span class="str">'./agents/subagent.js'</span>;
<span class="comment">// Inside a parent agent's perform():</span>
<span class="kw">const</span> sub = <span class="kw">new</span> <span class="fn">SubAgent</span>(<span class="kw">new</span> <span class="fn">ShellAgent</span>(), { maxDepth: 3 });
<span class="kw">const</span> fileList = <span class="kw">await</span> sub.<span class="fn">invoke</span>({ action: <span class="str">'list'</span>, path: <span class="str">'.'</span> });</code></pre>
<div class="info-box">
<strong>Next:</strong> The <a href="./architecture.html">Architecture page</a> covers the full execution flow, all sloshing signal namespaces, the WebSocket gateway, multi-channel routing (Slack, Discord, Telegram, iMessage, Teams), and the ClawHub skills system in depth.
</div>
<!-- Next Steps -->
<div class="info-box" style="margin-top:2rem;">
<strong>Next Steps</strong>
<ul style="margin-top:0.5rem;padding-left:1.25rem;">
<li><a href="./docs.html">Full Documentation</a> — Complete reference for all systems</li>
<li><a href="./architecture.html">Architecture Deep Dive</a> — How data sloshing and agents work</li>
<li><a href="https://github.com/kody-w/openrappter">GitHub Repository</a> — Source code and contributing guide</li>
</ul>
</div>
</div>
</main>
</div>
<!-- Footer -->
<footer>
<div class="footer-container">
<div class="footer-links">
<a href="https://rappter.com" style="color:var(--green)">rappter.com</a>
<a href="https://rappter.beehiiv.com/" style="color:var(--green)">Newsletter</a>
<a href="https://github.com/kody-w/openrappter">GitHub</a>
<a href="./docs.html">Docs</a>
<a href="./">Home</a>
<a href="./changelog.html">Changelog</a>
</div>
<div class="footer-copy">MIT License</div>
</div>
</footer>
<script src="./nav.js"></script>
</body>
</html>