fetchium-core 1.0.0

Fetchium core library — search, extract, rank, validate, cache
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
//! Answer Extraction Engine (AXE) — extracts direct answers from search results.
//!
//! For factual queries like "What is the capital of France?", this engine
//! extracts candidate answers from result snippets using pattern matching
//! and cross-result voting.

use crate::types::ResultItem;
use std::collections::HashMap;

/// Configuration for the answer extractor.
#[derive(Debug, Clone)]
pub struct AnswerConfig {
    /// Minimum number of results that must agree on an answer.
    pub min_agreement: usize,
    /// Confidence threshold below which answers are not returned.
    pub min_confidence: f64,
    /// Maximum number of candidate answers to return.
    pub max_answers: usize,
}

impl Default for AnswerConfig {
    fn default() -> Self {
        Self {
            min_agreement: 1,
            min_confidence: 0.3,
            max_answers: 3,
        }
    }
}

/// A candidate answer extracted from results.
#[derive(Debug, Clone)]
pub struct ExtractedAnswer {
    /// The answer text.
    pub answer: String,
    /// Confidence score (0.0–1.0).
    pub confidence: f64,
    /// Number of results that support this answer.
    pub supporting_sources: usize,
    /// Extraction method used.
    pub method: ExtractionMethod,
}

/// How the answer was extracted.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum ExtractionMethod {
    /// "X is Y" pattern matching.
    DefinitionPattern,
    /// Cross-result entity voting.
    EntityVoting,
    /// Direct snippet extraction.
    SnippetExtraction,
}

/// Extract direct answers from search results for a given query.
pub fn extract_answers(
    query: &str,
    results: &[ResultItem],
    config: &AnswerConfig,
) -> Vec<ExtractedAnswer> {
    if results.is_empty() || query.is_empty() {
        return vec![];
    }

    let mut candidates: Vec<ExtractedAnswer> = Vec::new();

    // Method 1: Definition patterns ("X is Y", "X are Y")
    let definitions = extract_definitions(query, results);
    candidates.extend(definitions);

    // Method 2: Entity voting (common entities across results)
    let entities = entity_voting(query, results);
    candidates.extend(entities);

    // Filter by confidence and agreement
    candidates.retain(|a| {
        a.confidence >= config.min_confidence && a.supporting_sources >= config.min_agreement
    });

    // Sort by confidence descending
    candidates.sort_by(|a, b| {
        b.confidence
            .partial_cmp(&a.confidence)
            .unwrap_or(std::cmp::Ordering::Equal)
    });

    // Deduplicate similar answers
    deduplicate_answers(&mut candidates);

    candidates.truncate(config.max_answers);
    candidates
}

/// Check if a query is answerable (factual question).
pub fn is_answerable_query(query: &str) -> bool {
    let lower = query.to_lowercase();
    let words: Vec<&str> = lower.split_whitespace().collect();

    // Check for question patterns
    if let Some(first) = words.first() {
        if QUESTION_WORDS.contains(first) {
            return true;
        }
    }

    // Check for "define X", "meaning of X" patterns
    lower.starts_with("define ")
        || lower.contains("meaning of")
        || lower.contains("definition of")
        || lower.ends_with('?')
}

// ─── Definition pattern extraction ────────────────────────

fn extract_definitions(query: &str, results: &[ResultItem]) -> Vec<ExtractedAnswer> {
    let query_lower = query.to_lowercase();
    let query_subject = extract_subject(&query_lower);
    let mut answers = Vec::new();

    for result in results {
        let snippet_lower = result.snippet.to_lowercase();
        let sentences = split_sentences(&snippet_lower);

        for sentence in &sentences {
            // Look for "Subject is/are Definition" patterns
            if let Some(definition) = extract_is_pattern(sentence, &query_subject) {
                let clean = clean_answer(&definition);
                if !clean.is_empty() && clean.len() < 200 {
                    answers.push(ExtractedAnswer {
                        answer: clean,
                        confidence: 0.7,
                        supporting_sources: 1,
                        method: ExtractionMethod::DefinitionPattern,
                    });
                }
            }
        }
    }

    // Merge identical definitions
    merge_answers(&mut answers);
    answers
}

fn extract_subject(query: &str) -> String {
    let mut subject = query.to_string();

    // Strip question words
    for prefix in &[
        "what is ",
        "what are ",
        "who is ",
        "who are ",
        "where is ",
        "when is ",
        "define ",
    ] {
        if subject.starts_with(prefix) {
            subject = subject[prefix.len()..].to_string();
            break;
        }
    }

    // Strip trailing "?"
    if subject.ends_with('?') {
        subject.pop();
    }

    subject.trim().to_string()
}

fn extract_is_pattern(sentence: &str, subject: &str) -> Option<String> {
    if subject.is_empty() {
        return None;
    }

    // Look for "subject is/are definition"
    for connector in &[" is ", " are ", " was ", " refers to ", " means "] {
        if let Some(pos) = sentence.find(connector) {
            let before = &sentence[..pos];
            // Check if subject appears before the connector
            if before.contains(subject) || subject.contains(before.trim()) {
                let after = &sentence[pos + connector.len()..];
                let definition = after.trim().to_string();
                if !definition.is_empty() {
                    return Some(definition);
                }
            }
        }
    }
    None
}

// ─── Entity voting ────────────────────────────────────────

fn entity_voting(query: &str, results: &[ResultItem]) -> Vec<ExtractedAnswer> {
    let query_terms: Vec<String> = query
        .to_lowercase()
        .split_whitespace()
        .filter(|w| w.len() >= 3)
        .map(|w| w.to_string())
        .collect();

    // Extract candidate entities from snippets (capitalized phrases, numbers, etc.)
    let mut entity_counts: HashMap<String, usize> = HashMap::new();

    for result in results {
        let entities = extract_entities_from_snippet(&result.snippet, &query_terms);
        for entity in entities {
            *entity_counts.entry(entity).or_insert(0) += 1;
        }
    }

    // Convert to answers, requiring multiple sources
    let total = results.len().max(1);
    entity_counts
        .into_iter()
        .filter(|(_, count)| *count >= 1)
        .map(|(entity, count)| {
            let confidence = (count as f64 / total as f64).min(0.95);
            ExtractedAnswer {
                answer: entity,
                confidence,
                supporting_sources: count,
                method: ExtractionMethod::EntityVoting,
            }
        })
        .collect()
}

fn extract_entities_from_snippet(snippet: &str, query_terms: &[String]) -> Vec<String> {
    let mut entities = Vec::new();

    // Extract numbers with units (e.g., "384,400 km", "100°C")
    let words: Vec<&str> = snippet.split_whitespace().collect();
    for window in words.windows(2) {
        let first = window[0];
        let second = window[1];
        if first.chars().any(|c| c.is_ascii_digit()) && second.len() <= 10 {
            let candidate = format!("{} {}", first, second);
            if !has_overlap(&candidate.to_lowercase(), query_terms) {
                entities.push(candidate);
            }
        }
    }

    // Extract capitalized phrases (potential named entities)
    let mut current_phrase = Vec::new();
    for word in &words {
        if word.starts_with(|c: char| c.is_uppercase()) && word.len() >= 2 {
            current_phrase.push(*word);
        } else {
            if !current_phrase.is_empty() {
                let phrase = current_phrase.join(" ");
                if phrase.len() >= 3 && !has_overlap(&phrase.to_lowercase(), query_terms) {
                    entities.push(phrase);
                }
            }
            current_phrase.clear();
        }
    }
    if !current_phrase.is_empty() {
        let phrase = current_phrase.join(" ");
        if phrase.len() >= 3 && !has_overlap(&phrase.to_lowercase(), query_terms) {
            entities.push(phrase);
        }
    }

    entities
}

fn has_overlap(text: &str, terms: &[String]) -> bool {
    terms.iter().any(|t| text.contains(t.as_str()))
}

// ─── Helpers ─────────────────────────────────────────────

fn split_sentences(text: &str) -> Vec<String> {
    text.split(['.', '!', '?'])
        .map(|s| s.trim().to_string())
        .filter(|s| !s.is_empty())
        .collect()
}

fn clean_answer(answer: &str) -> String {
    let trimmed = answer.trim();
    // Strip trailing punctuation
    let stripped = trimmed.trim_end_matches(['.', ',', ';']);
    stripped.to_string()
}

fn merge_answers(answers: &mut Vec<ExtractedAnswer>) {
    let mut merged: Vec<ExtractedAnswer> = Vec::new();

    for answer in answers.drain(..) {
        let lower = answer.answer.to_lowercase();
        if let Some(existing) = merged.iter_mut().find(|a| a.answer.to_lowercase() == lower) {
            existing.supporting_sources += answer.supporting_sources;
            existing.confidence = (existing.confidence + answer.confidence) / 2.0;
        } else {
            merged.push(answer);
        }
    }

    *answers = merged;
}

fn deduplicate_answers(answers: &mut Vec<ExtractedAnswer>) {
    let mut seen: Vec<String> = Vec::new();
    answers.retain(|a| {
        let lower = a.answer.to_lowercase();
        if seen
            .iter()
            .any(|s| s == &lower || s.contains(&lower) || lower.contains(s.as_str()))
        {
            false
        } else {
            seen.push(lower);
            true
        }
    });
}

const QUESTION_WORDS: &[&str] = &["what", "who", "where", "when", "which", "how", "define"];

#[cfg(test)]
mod tests {
    use super::*;
    use crate::types::BackendId;

    fn make_item(title: &str, snippet: &str) -> ResultItem {
        ResultItem {
            title: title.into(),
            url: format!(
                "https://{}.example.com",
                title.to_lowercase().replace(' ', "-")
            ),
            snippet: snippet.into(),
            rank: 0,
            backend: BackendId::DuckDuckGo,
            score: Some(0.5),
            published_date: None,
        }
    }

    #[test]
    fn empty_input_no_answers() {
        let answers = extract_answers("", &[], &AnswerConfig::default());
        assert!(answers.is_empty());
    }

    #[test]
    fn definition_pattern_extraction() {
        let results = vec![
            make_item(
                "Rust",
                "Rust is a systems programming language focused on safety and performance.",
            ),
            make_item(
                "About Rust",
                "Rust is a multi-paradigm systems programming language.",
            ),
        ];
        let answers = extract_answers("What is Rust?", &results, &AnswerConfig::default());
        assert!(
            !answers.is_empty(),
            "should extract definition for 'what is' query"
        );
    }

    #[test]
    fn answerable_query_detection() {
        assert!(is_answerable_query("What is the capital of France?"));
        assert!(is_answerable_query("Who invented the telephone?"));
        assert!(is_answerable_query("define algorithm"));
        assert!(!is_answerable_query("rust programming"));
    }

    #[test]
    fn subject_extraction() {
        assert_eq!(extract_subject("what is rust"), "rust");
        assert_eq!(extract_subject("who is alan turing"), "alan turing");
        assert_eq!(extract_subject("define algorithm"), "algorithm");
    }

    #[test]
    fn entity_voting_finds_common_entities() {
        let results = vec![
            make_item(
                "France Capital",
                "The capital of France is Paris, a major European city.",
            ),
            make_item(
                "Paris Info",
                "Paris is the capital and largest city of France.",
            ),
            make_item(
                "French Cities",
                "Paris serves as the capital of France since centuries.",
            ),
        ];
        let answers = extract_answers("capital of France", &results, &AnswerConfig::default());

        let has_paris = answers.iter().any(|a| {
            let lower = a.answer.to_lowercase();
            lower.contains("paris")
        });
        assert!(
            has_paris,
            "Paris should be extracted as answer, got: {:?}",
            answers
        );
    }

    #[test]
    fn confidence_sorted_descending() {
        let results = vec![
            make_item("A", "Rust is a systems programming language for safety."),
            make_item("B", "Rust is a compiled language for performance."),
        ];
        let answers = extract_answers("What is Rust?", &results, &AnswerConfig::default());
        if answers.len() >= 2 {
            for i in 0..answers.len() - 1 {
                assert!(answers[i].confidence >= answers[i + 1].confidence);
            }
        }
    }

    #[test]
    fn max_answers_respected() {
        let config = AnswerConfig {
            max_answers: 1,
            min_confidence: 0.0,
            ..AnswerConfig::default()
        };
        let results = vec![make_item(
            "A",
            "Rust is safe. Rust is fast. Rust is concurrent.",
        )];
        let answers = extract_answers("What is Rust?", &results, &config);
        assert!(answers.len() <= 1);
    }

    #[test]
    fn is_pattern_extraction() {
        let def = extract_is_pattern("rust is a systems programming language", "rust");
        assert!(def.is_some());
        assert!(def.unwrap().contains("systems programming language"));
    }

    #[test]
    fn clean_answer_strips_punctuation() {
        assert_eq!(clean_answer("  hello world.  "), "hello world");
        assert_eq!(clean_answer("answer,"), "answer");
    }
}