cartulary 0.3.0-alpha.1

The knowledge layer of your project — decisions, issues, docs, all in one place.
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

use std::path::PathBuf;

use nucleo::pattern::{CaseMatching, Normalization, Pattern};
use nucleo::{Config, Matcher, Utf32Str};

use crate::domain::model::decision_record::DecisionRecord;
use crate::domain::model::entity_ref::EntityRef;
use crate::domain::model::issue::Issue;
use crate::domain::model::record_kind::RecordKind;
use crate::domain::model::search::SearchHit;
use crate::domain::model::title::Title;
use crate::domain::usecases::search::SearchRepository;
use crate::infra::driven::fs::decision_record_repository::DrParseCtx;
use crate::infra::driven::fs::repository_pipeline;

/// A candidate record to be matched against the query.
struct Candidate {
    id: EntityRef,
    kind: RecordKind,
    title: Title,
    /// The text fed to the matcher: `"<title> <body>"`.
    haystack: String,
}

/// Filesystem search adapter powered by `nucleo-matcher`.
///
/// Scans all issues and decision records on each invocation (no persistent
/// index). Acceptable for knowledge bases up to a few hundred records.
pub struct NucleoSearchRepository {
    /// Issues directory.
    pub issues_dir: PathBuf,
    /// `(kind, dir)` pairs for each configured decision record kind.
    pub decision_kinds: Vec<(String, PathBuf)>,
    /// Workspace schema version, threaded into the shared parse pipeline so
    /// canonical ids and link targets are validated against the right shape.
    pub schema_version: u32,
}

impl NucleoSearchRepository {
    fn collect_candidates(&self) -> Vec<Candidate> {
        let mut candidates: Vec<Candidate> = Vec::new();

        // ── Issues ────────────────────────────────────────────────────────────
        if let Ok(entries) = std::fs::read_dir(&self.issues_dir) {
            for entry in entries.flatten() {
                let index = entry.path().join("index.md");
                if !index.exists() {
                    continue;
                }
                if let Ok((issue, _raw_events)) =
                    repository_pipeline::parse_one::<Issue>(&index, self.schema_version, &())
                {
                    let haystack = format!("{} {}", issue.title.as_str(), issue.content.as_str());
                    if let (Ok(id), Ok(kind)) = (
                        EntityRef::new(issue.id.to_string()),
                        RecordKind::new("issue"),
                    ) {
                        candidates.push(Candidate {
                            id,
                            kind,
                            title: issue.title,
                            haystack,
                        });
                    }
                }
            }
        }

        // ── Decision records ──────────────────────────────────────────────────
        for (kind_str, dir) in &self.decision_kinds {
            if let Ok(entries) = std::fs::read_dir(dir) {
                for entry in entries.flatten() {
                    let path = entry.path();
                    if !path.is_dir() {
                        continue;
                    }
                    let index = path.join("index.md");
                    if !index.exists() {
                        continue;
                    }
                    let ctx = DrParseCtx {
                        kind: kind_str.clone(),
                    };
                    if let Ok((record, _raw_events)) = repository_pipeline::parse_one::<
                        DecisionRecord,
                    >(
                        &index, self.schema_version, &ctx
                    ) {
                        let haystack =
                            format!("{} {}", record.title.as_str(), record.content.as_str());
                        if let (Ok(id), Ok(kind)) = (
                            EntityRef::new(record.id.to_string()),
                            RecordKind::new(kind_str),
                        ) {
                            candidates.push(Candidate {
                                id,
                                kind,
                                title: record.title,
                                haystack,
                            });
                        }
                    }
                }
            }
        }

        candidates
    }
}

impl SearchRepository for NucleoSearchRepository {
    fn search(&self, query: &str) -> anyhow::Result<Vec<SearchHit>> {
        let candidates = self.collect_candidates();

        let mut matcher = Matcher::new(Config::DEFAULT);
        let pattern = Pattern::parse(query, CaseMatching::Ignore, Normalization::Smart);

        // Match each candidate individually rather than via `match_list`,
        // whose opaque item resolution collides candidates whose displayed
        // text matches by string equality.
        let mut char_buf: Vec<char> = Vec::new();
        let mut scored: Vec<(usize, u32)> = candidates
            .iter()
            .enumerate()
            .filter_map(|(i, c)| {
                char_buf.clear();
                let haystack = Utf32Str::new(&c.haystack, &mut char_buf);
                pattern.score(haystack, &mut matcher).map(|s| (i, s))
            })
            .collect();

        scored.sort_by_key(|b| std::cmp::Reverse(b.1));

        let hits: Vec<SearchHit> = scored
            .into_iter()
            .map(|(idx, _score)| {
                let c = &candidates[idx];
                SearchHit {
                    id: c.id.clone(),
                    kind: c.kind.clone(),
                    title: c.title.clone(),
                    excerpt: make_excerpt(&c.haystack, query),
                }
            })
            .collect();

        Ok(hits)
    }
}

/// Build a short excerpt by finding the query term in `haystack` and
/// returning a window of characters around it. Returns `None` if the
/// query is not found verbatim (case-insensitive).
fn make_excerpt(haystack: &str, query: &str) -> Option<String> {
    let lower_haystack = haystack.to_lowercase();
    let lower_query = query.to_lowercase();
    let pos = lower_haystack.find(&lower_query)?;

    let start = pos.saturating_sub(30);
    let end = (pos + query.len() + 30).min(haystack.len());

    let prefix = if start > 0 { "…" } else { "" };
    let suffix = if end < haystack.len() { "…" } else { "" };

    // Snap to char boundaries.
    let snapped_start = (0..=start)
        .rev()
        .find(|&i| haystack.is_char_boundary(i))
        .unwrap_or(0);
    let snapped_end = (end..=haystack.len())
        .find(|&i| haystack.is_char_boundary(i))
        .unwrap_or(haystack.len());

    Some(format!(
        "{}{}{}",
        prefix,
        &haystack[snapped_start..snapped_end],
        suffix
    ))
}

#[cfg(test)]
mod tests {
    use super::*;
    use tempfile::TempDir;

    fn write_file(dir: &std::path::Path, path: &str, content: &str) {
        let p = dir.join(path);
        std::fs::create_dir_all(p.parent().unwrap()).unwrap();
        std::fs::write(p, content).unwrap();
    }

    #[test]
    fn search_returns_decision_record_hit() {
        let tmp = TempDir::new().unwrap();
        write_file(
            tmp.path(),
            "docs/adr/0001-use-rust/index.md",
            "---\nid: ADR-0001\ntitle: \"Use Rust\"\nstatus: accepted\ndate: 2026-01-01\n---\n\nWe chose Rust for performance.\n",
        );
        let repo = NucleoSearchRepository {
            issues_dir: tmp.path().join("docs/issues"),
            decision_kinds: vec![("adr".into(), tmp.path().join("docs/adr"))],
            schema_version: 3,
        };
        let hits = repo.search("Rust").unwrap();
        assert!(!hits.is_empty(), "expected hits for 'Rust', got none");
        assert_eq!(hits[0].id.to_string(), "ADR-0001");
    }

    #[test]
    fn search_returns_dr_hit_when_issue_also_matches() {
        let tmp = TempDir::new().unwrap();
        // ADR that contains the keyword in body
        write_file(
            tmp.path(),
            "docs/adr/0002-hexagonal/index.md",
            "---\nid: ADR-0002\ntitle: \"Hexagonal Architecture\"\nstatus: accepted\ndate: 2026-01-01\n---\n\nDomain layer is isolated.\n",
        );
        // Issue that also mentions "hexagonal" in body
        write_file(
            tmp.path(),
            "docs/issues/0001-something/index.md",
            "---\nid: ISSUE-0001\ntitle: \"Something\"\ntype: feature\nstatus: open\ndate: 2026-01-01\n---\n\nSee ADR-0002 for hexagonal details.\n",
        );
        let repo = NucleoSearchRepository {
            issues_dir: tmp.path().join("docs/issues"),
            decision_kinds: vec![("adr".into(), tmp.path().join("docs/adr"))],
            schema_version: 3,
        };
        let hits = repo.search("hexagonal").unwrap();
        let ids: Vec<String> = hits.iter().map(|h| h.id.to_string()).collect();
        assert!(
            ids.contains(&"ADR-0002".to_string()),
            "ADR-0002 missing from hits: {ids:?}"
        );
    }

    #[test]
    fn collects_decision_record_candidates() {
        let tmp = TempDir::new().unwrap();
        write_file(
            tmp.path(),
            "docs/adr/0001-use-rust/index.md",
            "---\nid: ADR-0001\ntitle: \"Use Rust\"\nstatus: accepted\ndate: 2026-01-01\n---\n\nWe chose Rust for performance.\n",
        );
        let repo = NucleoSearchRepository {
            issues_dir: tmp.path().join("docs/issues"),
            decision_kinds: vec![("adr".into(), tmp.path().join("docs/adr"))],
            schema_version: 3,
        };
        let candidates = repo.collect_candidates();
        assert_eq!(
            candidates.len(),
            1,
            "expected 1 candidate, got {}",
            candidates.len()
        );
        assert_eq!(candidates[0].id.to_string(), "ADR-0001");
        assert_eq!(candidates[0].kind.as_str(), "adr");
    }
}