llm-toolkit 0.63.1

A low-level, unopinionated Rust toolkit for the LLM last mile problem.
Documentation
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
//! Tracing tests for ParallelOrchestrator
//!
//! These tests verify that structured tracing events and spans are properly
//! emitted during workflow execution.

use llm_toolkit::agent::{Agent, AgentError, AgentOutput, DynamicAgent, Payload};
use llm_toolkit::orchestrator::{
    BlueprintWorkflow, ParallelOrchestrator, StrategyMap, StrategyStep,
};
use serde_json::{Value as JsonValue, json};
use std::sync::Arc;
use tokio_util::sync::CancellationToken;
use tracing::Level;
use tracing_subscriber::fmt::MakeWriter;
use tracing_subscriber::fmt::format::FmtSpan;

// ============================================================================
// Test Infrastructure
// ============================================================================

/// Captures tracing output to a string for verification
#[derive(Clone)]
struct TestWriter {
    output: Arc<std::sync::Mutex<Vec<u8>>>,
}

impl TestWriter {
    fn new() -> Self {
        Self {
            output: Arc::new(std::sync::Mutex::new(Vec::new())),
        }
    }

    fn get_output(&self) -> String {
        let bytes = self.output.lock().unwrap();
        String::from_utf8_lossy(&bytes).to_string()
    }
}

impl std::io::Write for TestWriter {
    fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
        self.output.lock().unwrap().write(buf)
    }

    fn flush(&mut self) -> std::io::Result<()> {
        self.output.lock().unwrap().flush()
    }
}

impl<'a> MakeWriter<'a> for TestWriter {
    type Writer = Self;

    fn make_writer(&'a self) -> Self::Writer {
        self.clone()
    }
}

// ============================================================================
// Mock Agents
// ============================================================================

#[derive(Clone)]
struct SimpleAgent {
    name: String,
    output: JsonValue,
}

impl SimpleAgent {
    fn new(name: impl Into<String>, output: JsonValue) -> Self {
        Self {
            name: name.into(),
            output,
        }
    }
}

#[async_trait::async_trait]
impl Agent for SimpleAgent {
    type Output = JsonValue;
    type Expertise = &'static str;

    fn expertise(&self) -> &&'static str {
        const EXPERTISE: &str = "Simple test agent";
        &EXPERTISE
    }

    async fn execute(&self, _input: Payload) -> Result<Self::Output, AgentError> {
        Ok(self.output.clone())
    }
}

#[async_trait::async_trait]
impl DynamicAgent for SimpleAgent {
    fn name(&self) -> String {
        self.name.clone()
    }

    fn description(&self) -> &str {
        "Simple test agent"
    }

    async fn execute_dynamic(&self, input: Payload) -> Result<AgentOutput, AgentError> {
        let output = self.execute(input).await?;
        Ok(AgentOutput::Success(output))
    }
}

// ============================================================================
// Tracing Tests
// ============================================================================

#[tokio::test]
async fn test_top_level_span_created() {
    let writer = TestWriter::new();

    let subscriber = tracing_subscriber::fmt()
        .with_max_level(Level::DEBUG)
        .with_span_events(FmtSpan::NEW | FmtSpan::CLOSE)
        .with_ansi(false)
        .with_writer(writer.clone())
        .finish();

    let _guard = tracing::subscriber::set_default(subscriber);

    let mut strategy = StrategyMap::new("Test".to_string());
    strategy.add_step(StrategyStep::new(
        "step_1".to_string(),
        "Step 1".to_string(),
        "Agent1".to_string(),
        "{{ task }}".to_string(),
        "Output 1".to_string(),
    ));

    let blueprint = BlueprintWorkflow::new("Test Blueprint".to_string());
    let mut orchestrator = ParallelOrchestrator::new(blueprint);
    orchestrator.set_strategy(strategy);
    orchestrator.add_agent(
        "Agent1",
        Arc::new(SimpleAgent::new("Agent1", json!({"ok": true}))),
    );

    let _result = orchestrator
        .execute("test task", CancellationToken::new(), None, None)
        .await
        .unwrap();

    let output = writer.get_output();

    // Verify top-level span exists
    assert!(
        output.contains("parallel_orchestrator_execute"),
        "Top-level span 'parallel_orchestrator_execute' not found in output:\n{}",
        output
    );

    // Verify span has task attribute
    assert!(
        output.contains("task=test task") || output.contains("task=\"test task\""),
        "Span should include task attribute in output:\n{}",
        output
    );
}

#[tokio::test]
async fn test_wave_spans_created() {
    let writer = TestWriter::new();

    let subscriber = tracing_subscriber::fmt()
        .with_max_level(Level::DEBUG)
        .with_span_events(FmtSpan::NEW | FmtSpan::CLOSE)
        .with_ansi(false)
        .with_writer(writer.clone())
        .finish();

    let _guard = tracing::subscriber::set_default(subscriber);

    let mut strategy = StrategyMap::new("Wave Test".to_string());

    // Two independent steps -> wave 1
    strategy.add_step(StrategyStep::new(
        "step_1".to_string(),
        "Step 1".to_string(),
        "Agent1".to_string(),
        "{{ task }}".to_string(),
        "Output 1".to_string(),
    ));

    strategy.add_step(StrategyStep::new(
        "step_2".to_string(),
        "Step 2".to_string(),
        "Agent2".to_string(),
        "{{ task }}".to_string(),
        "Output 2".to_string(),
    ));

    // Dependent step -> wave 2
    strategy.add_step(StrategyStep::new(
        "step_3".to_string(),
        "Step 3".to_string(),
        "Agent3".to_string(),
        "{{ step_1_output }}".to_string(),
        "Output 3".to_string(),
    ));

    let blueprint = BlueprintWorkflow::new("Test Blueprint".to_string());
    let mut orchestrator = ParallelOrchestrator::new(blueprint);
    orchestrator.set_strategy(strategy);
    orchestrator.add_agent(
        "Agent1",
        Arc::new(SimpleAgent::new("Agent1", json!({"ok": 1}))),
    );
    orchestrator.add_agent(
        "Agent2",
        Arc::new(SimpleAgent::new("Agent2", json!({"ok": 2}))),
    );
    orchestrator.add_agent(
        "Agent3",
        Arc::new(SimpleAgent::new("Agent3", json!({"ok": 3}))),
    );

    let _result = orchestrator
        .execute("wave test", CancellationToken::new(), None, None)
        .await
        .unwrap();

    let output = writer.get_output();

    // Verify wave spans exist
    assert!(
        output.contains("wave"),
        "Wave spans not found in output:\n{}",
        output
    );

    // Should have at least 2 waves
    assert!(
        output.matches("wave_number=1").count() >= 1,
        "Wave 1 not found in output:\n{}",
        output
    );
}

#[tokio::test]
async fn test_step_spans_created() {
    let writer = TestWriter::new();

    let subscriber = tracing_subscriber::fmt()
        .with_max_level(Level::DEBUG)
        .with_span_events(FmtSpan::NEW | FmtSpan::CLOSE)
        .with_ansi(false)
        .with_writer(writer.clone())
        .finish();

    let _guard = tracing::subscriber::set_default(subscriber);

    let mut strategy = StrategyMap::new("Step Span Test".to_string());
    strategy.add_step(StrategyStep::new(
        "step_1".to_string(),
        "Step 1".to_string(),
        "Agent1".to_string(),
        "{{ task }}".to_string(),
        "Output 1".to_string(),
    ));

    let blueprint = BlueprintWorkflow::new("Test Blueprint".to_string());
    let mut orchestrator = ParallelOrchestrator::new(blueprint);
    orchestrator.set_strategy(strategy);
    orchestrator.add_agent(
        "Agent1",
        Arc::new(SimpleAgent::new("Agent1", json!({"ok": true}))),
    );

    let _result = orchestrator
        .execute("step test", CancellationToken::new(), None, None)
        .await
        .unwrap();

    let output = writer.get_output();

    // Verify per-step span exists
    assert!(
        output.contains("parallel_step"),
        "Per-step span 'parallel_step' not found in output:\n{}",
        output
    );

    // Verify step_id attribute
    assert!(
        output.contains("step_id=step_1") || output.contains("step_id=\"step_1\""),
        "Step span should include step_id attribute in output:\n{}",
        output
    );

    // Verify agent_name attribute
    assert!(
        output.contains("agent_name=Agent1") || output.contains("agent_name=\"Agent1\""),
        "Step span should include agent_name attribute in output:\n{}",
        output
    );
}

#[tokio::test]
async fn test_state_transition_events() {
    let writer = TestWriter::new();

    let subscriber = tracing_subscriber::fmt()
        .with_max_level(Level::DEBUG)
        .with_ansi(false)
        .with_writer(writer.clone())
        .finish();

    let _guard = tracing::subscriber::set_default(subscriber);

    let mut strategy = StrategyMap::new("State Events Test".to_string());
    strategy.add_step(StrategyStep::new(
        "step_1".to_string(),
        "Step 1".to_string(),
        "Agent1".to_string(),
        "{{ task }}".to_string(),
        "Output 1".to_string(),
    ));

    let blueprint = BlueprintWorkflow::new("Test Blueprint".to_string());
    let mut orchestrator = ParallelOrchestrator::new(blueprint);
    orchestrator.set_strategy(strategy);
    orchestrator.add_agent(
        "Agent1",
        Arc::new(SimpleAgent::new("Agent1", json!({"ok": true}))),
    );

    let _result = orchestrator
        .execute("state test", CancellationToken::new(), None, None)
        .await
        .unwrap();

    let output = writer.get_output();

    // Verify state transition events
    assert!(
        output.contains("Step marked as Ready"),
        "Ready state event not found in output:\n{}",
        output
    );

    assert!(
        output.contains("Step execution started"),
        "Running state event not found in output:\n{}",
        output
    );

    assert!(
        output.contains("Step completed successfully"),
        "Completed state event not found in output:\n{}",
        output
    );
}

#[tokio::test]
async fn test_failure_events() {
    let writer = TestWriter::new();

    let subscriber = tracing_subscriber::fmt()
        .with_max_level(Level::DEBUG)
        .with_ansi(false)
        .with_writer(writer.clone())
        .finish();

    let _guard = tracing::subscriber::set_default(subscriber);

    #[derive(Clone)]
    struct FailingAgent;

    #[async_trait::async_trait]
    impl Agent for FailingAgent {
        type Output = JsonValue;
        type Expertise = &'static str;

        fn expertise(&self) -> &&'static str {
            const EXPERTISE: &str = "Failing agent";
            &EXPERTISE
        }

        async fn execute(&self, _input: Payload) -> Result<Self::Output, AgentError> {
            Err(AgentError::ExecutionFailed(
                "Intentional failure".to_string(),
            ))
        }
    }

    #[async_trait::async_trait]
    impl DynamicAgent for FailingAgent {
        fn name(&self) -> String {
            "FailingAgent".to_string()
        }

        fn description(&self) -> &str {
            "Failing agent"
        }

        async fn execute_dynamic(&self, input: Payload) -> Result<AgentOutput, AgentError> {
            let output = self.execute(input).await?;
            Ok(AgentOutput::Success(output))
        }
    }

    let mut strategy = StrategyMap::new("Failure Test".to_string());
    strategy.add_step(StrategyStep::new(
        "step_1".to_string(),
        "Failing Step".to_string(),
        "FailAgent".to_string(),
        "{{ task }}".to_string(),
        "Output 1".to_string(),
    ));

    strategy.add_step(StrategyStep::new(
        "step_2".to_string(),
        "Dependent Step".to_string(),
        "Agent2".to_string(),
        "{{ step_1_output }}".to_string(),
        "Output 2".to_string(),
    ));

    let blueprint = BlueprintWorkflow::new("Test Blueprint".to_string());
    let mut orchestrator = ParallelOrchestrator::new(blueprint);
    orchestrator.set_strategy(strategy);
    orchestrator.add_agent("FailAgent", Arc::new(FailingAgent));
    orchestrator.add_agent(
        "Agent2",
        Arc::new(SimpleAgent::new("Agent2", json!({"ok": 2}))),
    );

    let _result = orchestrator
        .execute("failure test", CancellationToken::new(), None, None)
        .await
        .unwrap();

    let output = writer.get_output();

    // Verify failure event
    assert!(
        output.contains("Step failed"),
        "Step failure event not found in output:\n{}",
        output
    );

    // Verify skipped event
    assert!(
        output.contains("Step skipped due to failed dependency"),
        "Skipped dependency event not found in output:\n{}",
        output
    );
}

#[tokio::test]
async fn test_span_hierarchy() {
    let writer = TestWriter::new();

    let subscriber = tracing_subscriber::fmt()
        .with_max_level(Level::DEBUG)
        .with_span_events(FmtSpan::NEW | FmtSpan::ENTER | FmtSpan::EXIT | FmtSpan::CLOSE)
        .with_ansi(false)
        .with_writer(writer.clone())
        .finish();

    let _guard = tracing::subscriber::set_default(subscriber);

    let mut strategy = StrategyMap::new("Hierarchy Test".to_string());
    strategy.add_step(StrategyStep::new(
        "step_1".to_string(),
        "Step 1".to_string(),
        "Agent1".to_string(),
        "{{ task }}".to_string(),
        "Output 1".to_string(),
    ));

    let blueprint = BlueprintWorkflow::new("Test Blueprint".to_string());
    let mut orchestrator = ParallelOrchestrator::new(blueprint);
    orchestrator.set_strategy(strategy);
    orchestrator.add_agent(
        "Agent1",
        Arc::new(SimpleAgent::new("Agent1", json!({"ok": true}))),
    );

    let _result = orchestrator
        .execute("hierarchy test", CancellationToken::new(), None, None)
        .await
        .unwrap();

    let output = writer.get_output();

    // Verify all three levels of spans exist
    assert!(
        output.contains("parallel_orchestrator_execute"),
        "Top-level span missing"
    );
    assert!(output.contains("wave"), "Wave span missing");
    assert!(output.contains("parallel_step"), "Step span missing");

    // The output should show nested structure (exact format depends on subscriber configuration)
    // We just verify all levels are present
}