ainl-persona 0.1.6

Persona evolution engine over AINL graph memory (soft axes, metadata signals)
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
//! Metadata-driven `RawSignal` extraction from graph memory nodes.

use crate::axes::PersonaAxis;
use ainl_memory::{AinlMemoryNode, EpisodicNode, PersonaNode, ProceduralNode, SemanticNode};
use serde_json::Value;
use std::collections::HashSet;
use uuid::Uuid;

#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum MemoryNodeType {
    Episodic,
    Semantic,
    Procedural,
    Persona,
}

#[derive(Debug, Clone, PartialEq)]
pub struct RawSignal {
    pub axis: PersonaAxis,
    pub reward: f32,
    pub weight: f32,
    pub source_node_id: Uuid,
    pub source_node_type: MemoryNodeType,
}

fn trace_obj(ep: &EpisodicNode) -> Option<&serde_json::Map<String, Value>> {
    ep.trace_event.as_ref()?.as_object()
}

fn trace_outcome(ep: &EpisodicNode) -> String {
    trace_obj(ep)
        .and_then(|m| m.get("outcome"))
        .and_then(|v| v.as_str())
        .unwrap_or("")
        .to_string()
}

fn trace_duration_ms(ep: &EpisodicNode) -> u64 {
    trace_obj(ep)
        .and_then(|m| m.get("duration_ms"))
        .and_then(|v| v.as_u64().or_else(|| v.as_f64().map(|f| f as u64)))
        .unwrap_or(0)
}

fn trace_tool_name(ep: &EpisodicNode) -> Option<String> {
    let from_trace = trace_obj(ep)
        .and_then(|m| m.get("tool_name"))
        .and_then(|v| v.as_str())
        .map(str::to_string);
    if from_trace.is_some() {
        return from_trace;
    }
    ep.effective_tools()
        .first()
        .map(std::string::ToString::to_string)
}

fn trace_byte_count(ep: &EpisodicNode) -> u64 {
    trace_obj(ep)
        .and_then(|m| m.get("byte_count"))
        .and_then(|v| v.as_u64().or_else(|| v.as_f64().map(|f| f as u64)))
        .unwrap_or(0)
}

fn is_instrumentality_tool(name: &str) -> bool {
    let n = name.to_ascii_lowercase();
    n.contains("shell")
        || n.contains("cli")
        || n.contains("mcp")
        || n.contains("compile")
        || n.contains("compiler")
        || n == "ainl"
        || n.contains("ainl_")
}

fn is_web_tool(name: &str) -> bool {
    let n = name.to_ascii_lowercase();
    n.contains("web_search") || n.contains("web_fetch") || n.contains("web.fetch")
}

pub fn episodic_should_process(ep: &EpisodicNode) -> bool {
    let outcome = trace_outcome(ep);
    if outcome == "success" {
        return true;
    }
    if !ep.persona_signals_emitted.is_empty() {
        return true;
    }
    false
}

/// Extract persona signals from `EpisodicNode::vitals_*` fields.
///
/// These complement the existing heuristic signals (tool names, outcomes, duration)
/// with signals derived from the LLM's own token-level confidence during generation.
///
/// Axis mappings:
/// - `Systematicity` ← high trust / reasoning/retrieval gate=pass (the agent generated
///   structured, confident output)
/// - `Curiosity` ← hallucination or creative phase (the agent was probing uncertain territory)
/// - `Persistence` ← warn/fail gate but non-zero trust (the agent continued under uncertainty)
/// - `Verbosity` ← creative phase (open-ended, expansive generation)
pub fn extract_vitals_signals(node_id: Uuid, ep: &EpisodicNode) -> Vec<RawSignal> {
    let gate = match ep.vitals_gate.as_deref() {
        Some(g) => g,
        None => return Vec::new(),
    };
    let phase = ep.vitals_phase.as_deref().unwrap_or("");
    let trust = ep.vitals_trust.unwrap_or(0.5).clamp(0.0, 1.0);

    let phase_kind = phase.split(':').next().unwrap_or("");

    let mut out = Vec::new();

    match gate {
        "pass" => {
            match phase_kind {
                "reasoning" | "retrieval" => {
                    // Confident structured output → Systematicity reward.
                    out.push(RawSignal {
                        axis: PersonaAxis::Systematicity,
                        reward: 0.7 + 0.3 * trust,
                        weight: 0.65,
                        source_node_id: node_id,
                        source_node_type: MemoryNodeType::Episodic,
                    });
                }
                "creative" => {
                    out.push(RawSignal {
                        axis: PersonaAxis::Curiosity,
                        reward: 0.65,
                        weight: 0.5,
                        source_node_id: node_id,
                        source_node_type: MemoryNodeType::Episodic,
                    });
                    out.push(RawSignal {
                        axis: PersonaAxis::Verbosity,
                        reward: 0.6,
                        weight: 0.45,
                        source_node_id: node_id,
                        source_node_type: MemoryNodeType::Episodic,
                    });
                }
                _ => {}
            }
        }
        "warn" => {
            match phase_kind {
                "hallucination" => {
                    // Agent ventured into uncertain territory — mild Curiosity nudge,
                    // negative Systematicity (reduces structured-output score slightly).
                    out.push(RawSignal {
                        axis: PersonaAxis::Curiosity,
                        reward: 0.55,
                        weight: 0.4,
                        source_node_id: node_id,
                        source_node_type: MemoryNodeType::Episodic,
                    });
                    out.push(RawSignal {
                        axis: PersonaAxis::Systematicity,
                        reward: 0.2,
                        weight: 0.4,
                        source_node_id: node_id,
                        source_node_type: MemoryNodeType::Episodic,
                    });
                }
                "refusal" => {
                    // Refusal with Warn → cautious agent; mild Systematicity signal.
                    out.push(RawSignal {
                        axis: PersonaAxis::Systematicity,
                        reward: 0.5,
                        weight: 0.3,
                        source_node_id: node_id,
                        source_node_type: MemoryNodeType::Episodic,
                    });
                }
                _ => {
                    // Generic Warn: Persistence signal if trust is non-trivial.
                    if trust > 0.3 {
                        out.push(RawSignal {
                            axis: PersonaAxis::Persistence,
                            reward: 0.55,
                            weight: 0.4,
                            source_node_id: node_id,
                            source_node_type: MemoryNodeType::Episodic,
                        });
                    }
                }
            }
        }
        "fail" => {
            // Adversarial or high-entropy hallucination: suppress Systematicity,
            // weak Persistence if the agent was still producing something.
            out.push(RawSignal {
                axis: PersonaAxis::Systematicity,
                reward: 0.1,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Episodic,
            });
            if trust > 0.2 {
                out.push(RawSignal {
                    axis: PersonaAxis::Persistence,
                    reward: 0.4,
                    weight: 0.3,
                    source_node_id: node_id,
                    source_node_type: MemoryNodeType::Episodic,
                });
            }
        }
        _ => {}
    }

    out
}

pub fn extract_episodic_signals(node_id: Uuid, ep: &EpisodicNode) -> Vec<RawSignal> {
    let mut out = Vec::new();

    // Vitals-derived signals (first, so they can be overridden/complemented by heuristics).
    out.extend(extract_vitals_signals(node_id, ep));

    for hint in &ep.persona_signals_emitted {
        if let Some(sig) = parse_axis_hint(node_id, hint) {
            out.push(sig);
        }
    }

    let mut tool_names: HashSet<String> = HashSet::new();
    if let Some(t) = trace_tool_name(ep) {
        tool_names.insert(t);
    }
    for t in ep.effective_tools() {
        tool_names.insert(t.clone());
    }
    for tool in tool_names {
        if is_instrumentality_tool(&tool) {
            out.push(RawSignal {
                axis: PersonaAxis::Instrumentality,
                reward: 0.8,
                weight: 0.6,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Episodic,
            });
        }
        if is_web_tool(&tool) {
            out.push(RawSignal {
                axis: PersonaAxis::Curiosity,
                reward: 0.7,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Episodic,
            });
        }
        if tool.to_ascii_lowercase().contains("file_write") && trace_byte_count(ep) >= 4096 {
            out.push(RawSignal {
                axis: PersonaAxis::Verbosity,
                reward: 0.6,
                weight: 0.4,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Episodic,
            });
        }
    }

    let outcome = trace_outcome(ep);
    match outcome.as_str() {
        "success" => {
            out.push(RawSignal {
                axis: PersonaAxis::Systematicity,
                reward: 0.6,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Episodic,
            });
        }
        "retry" | "error" => {
            out.push(RawSignal {
                axis: PersonaAxis::Curiosity,
                reward: 0.5,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Episodic,
            });
        }
        _ => {}
    }

    if trace_duration_ms(ep) > 5000 {
        out.push(RawSignal {
            axis: PersonaAxis::Persistence,
            reward: 0.6,
            weight: 0.4,
            source_node_id: node_id,
            source_node_type: MemoryNodeType::Episodic,
        });
    }

    out
}

fn parse_axis_hint(node_id: Uuid, hint: &str) -> Option<RawSignal> {
    let (axis_part, reward_part) = hint.split_once(':')?;
    let axis = PersonaAxis::parse(axis_part)?;
    let reward: f32 = reward_part.trim().parse().ok()?;
    Some(RawSignal {
        axis,
        reward: reward.clamp(0.0, 1.0),
        weight: 0.8,
        source_node_id: node_id,
        source_node_type: MemoryNodeType::Episodic,
    })
}

pub fn semantic_should_process(sem: &SemanticNode) -> bool {
    sem.recurrence_count >= 2
}

fn semantic_tag_strings(sem: &SemanticNode) -> Vec<String> {
    let mut tags: Vec<String> = sem.tags.iter().map(|t| t.to_ascii_lowercase()).collect();
    if let Some(tc) = &sem.topic_cluster {
        for p in tc.split([',', ';']) {
            let s = p.trim().to_ascii_lowercase();
            if !s.is_empty() {
                tags.push(s);
            }
        }
    }
    tags
}

pub fn extract_semantic_signals(node_id: Uuid, sem: &SemanticNode) -> Vec<RawSignal> {
    let mut out = Vec::new();
    let tags = semantic_tag_strings(sem);

    for t in &tags {
        if t.contains("research") || t.contains("reference") || t.contains("documentation") {
            out.push(RawSignal {
                axis: PersonaAxis::Curiosity,
                reward: 0.7,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Semantic,
            });
        }
        if t.contains("pattern") || t.contains("template") || t.contains("schema") {
            out.push(RawSignal {
                axis: PersonaAxis::Systematicity,
                reward: 0.7,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Semantic,
            });
        }
        if t.contains("summary") || t.contains("output") || t.contains("result") {
            out.push(RawSignal {
                axis: PersonaAxis::Verbosity,
                reward: 0.5,
                weight: 0.5,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Semantic,
            });
        }
    }

    if sem.recurrence_count >= 3 {
        out.push(RawSignal {
            axis: PersonaAxis::Persistence,
            reward: 0.7,
            weight: 0.6,
            source_node_id: node_id,
            source_node_type: MemoryNodeType::Semantic,
        });
    }

    out
}

pub fn procedural_should_process(proc: &ProceduralNode) -> bool {
    proc.patch_version >= 1
}

fn procedural_fitness(proc: &ProceduralNode) -> f32 {
    proc.fitness.unwrap_or(proc.success_rate).clamp(0.0, 1.0)
}

pub fn extract_procedural_signals(node_id: Uuid, proc: &ProceduralNode) -> Vec<RawSignal> {
    let mut out = Vec::new();

    if proc.patch_version >= 2 {
        out.push(RawSignal {
            axis: PersonaAxis::Systematicity,
            reward: 0.8,
            weight: 0.55,
            source_node_id: node_id,
            source_node_type: MemoryNodeType::Procedural,
        });
    }

    if procedural_fitness(proc) >= 0.7 {
        out.push(RawSignal {
            axis: PersonaAxis::Persistence,
            reward: 0.7,
            weight: 0.55,
            source_node_id: node_id,
            source_node_type: MemoryNodeType::Procedural,
        });
    }

    if !proc.declared_reads.is_empty() {
        out.push(RawSignal {
            axis: PersonaAxis::Instrumentality,
            reward: 0.6,
            weight: 0.5,
            source_node_id: node_id,
            source_node_type: MemoryNodeType::Procedural,
        });
    }

    out
}

pub fn extract_persona_priors(node_id: Uuid, persona: &PersonaNode) -> Vec<RawSignal> {
    let mut out = Vec::new();
    for (name, score) in &persona.axis_scores {
        if let Some(axis) = PersonaAxis::parse(name) {
            out.push(RawSignal {
                axis,
                reward: (*score).clamp(0.0, 1.0),
                weight: 0.3,
                source_node_id: node_id,
                source_node_type: MemoryNodeType::Persona,
            });
        }
    }
    out
}

pub fn extract_from_node(node: &AinlMemoryNode) -> Vec<RawSignal> {
    let id = node.id;
    match &node.node_type {
        ainl_memory::AinlNodeType::Episode { episodic } => {
            if episodic_should_process(episodic) {
                extract_episodic_signals(id, episodic)
            } else {
                Vec::new()
            }
        }
        ainl_memory::AinlNodeType::Semantic { semantic } => {
            if semantic_should_process(semantic) {
                extract_semantic_signals(id, semantic)
            } else {
                Vec::new()
            }
        }
        ainl_memory::AinlNodeType::Procedural { procedural } => {
            if procedural_should_process(procedural) {
                extract_procedural_signals(id, procedural)
            } else {
                Vec::new()
            }
        }
        ainl_memory::AinlNodeType::Persona { persona } => extract_persona_priors(id, persona),
        ainl_memory::AinlNodeType::RuntimeState { .. } => Vec::new(),
    }
}