1use std::collections::{BTreeMap, BTreeSet, HashSet};
2use std::io;
3use std::path::{Path, PathBuf};
4
5use chrono::{DateTime, Utc};
6use serde::{Deserialize, Serialize};
7
8pub mod freshness;
9pub mod paths;
10pub mod recall;
11pub mod store;
12pub mod types;
13
14pub use freshness::{
15 memory_age_days, memory_age_label, memory_freshness_text, render_memory_freshness_note,
16 FreshnessKind,
17};
18pub use paths::{MemoryPathResolver, SESSIONS_DIR, TOPICS_DIR};
19pub use recall::{
20 select_relevant_memories, shortlist_relevant_memories, MemoryRecallCandidate,
21 MemoryRecallOptions, MemoryRecallRerankContext, MemoryRecallSelection, MemoryRecallStrategy,
22};
23pub use store::MemoryStore;
24pub use types::{
25 CreatedBy, DurableContentLocation, DurableMemoryDocument, DurableMemoryFrontmatter,
26 DurableMemoryRef, DurableMemoryRelations, DurableMemoryRetrieval, DurableMemorySource,
27 DurableMemoryStatus, DurableMemoryType, MemoryContradictionResult, MemoryInspectResult,
28 MemoryMergeResult, MemoryPurgeResult, MemoryQueryCursor, MemoryQueryItem, MemoryQueryOptions,
29 MemoryQueryResult, MemoryScope, SessionState,
30};
31
32pub const MEMORY_SCHEMA_VERSION: u32 = 1;
33pub const DEFAULT_SESSION_TOPIC: &str = "default";
34pub const MAX_SESSION_TOPIC_LEN: usize = 50;
35pub const MAX_MEMORY_TITLE_LEN: usize = 160;
36pub const MAX_MEMORY_TAGS: usize = 32;
37pub const DEFAULT_QUERY_LIMIT: usize = 5;
38pub const MAX_QUERY_LIMIT: usize = 20;
39pub const DEFAULT_MAX_CHARS: usize = 3_000;
40pub const MAX_MAX_CHARS: usize = 6_000;
41pub const WRITE_AUDIT_LOG: &str = "write_audit.jsonl";
42pub const MERGE_AUDIT_LOG: &str = "merge_audit.jsonl";
43pub const PURGE_AUDIT_LOG: &str = "purge_audit.jsonl";
44pub const CONTRADICTION_AUDIT_LOG: &str = "contradiction_audit.jsonl";
45pub const DREAM_VIEW_FILE: &str = "DREAM_NOTEBOOK.md";
46pub const MEMORY_VIEW_FILE: &str = "MEMORY.md";
47pub const RECENT_VIEW_FILE: &str = "RECENT.md";
48pub const STALE_VIEW_FILE: &str = "STALE.md";
49pub const LEXICAL_INDEX_FILE: &str = "lexical.json";
50pub const GRAPH_INDEX_FILE: &str = "graph.json";
51pub const RECENT_INDEX_FILE: &str = "recent.json";
52pub const STALE_CANDIDATES_INDEX_FILE: &str = "stale_candidates.json";
53pub const TAXONOMY_INDEX_FILE: &str = "taxonomy.json";
54
55pub fn validate_session_id(session_id: &str) -> io::Result<&str> {
56 let trimmed = session_id.trim();
57 if trimmed.is_empty() {
58 return Err(io::Error::new(
59 io::ErrorKind::InvalidInput,
60 "session_id cannot be empty",
61 ));
62 }
63 if trimmed.contains('/') || trimmed.contains('\\') || trimmed.contains("..") {
64 return Err(io::Error::new(
65 io::ErrorKind::InvalidInput,
66 "session_id contains invalid path characters",
67 ));
68 }
69 if !trimmed
70 .chars()
71 .all(|ch| ch.is_ascii_alphanumeric() || ch == '-' || ch == '_' || ch == '.')
72 {
73 return Err(io::Error::new(
74 io::ErrorKind::InvalidInput,
75 "session_id contains unsupported characters",
76 ));
77 }
78 Ok(trimmed)
79}
80
81pub fn validate_session_topic(topic: &str) -> io::Result<&str> {
82 let trimmed = topic.trim();
83 if trimmed.is_empty() {
84 return Err(io::Error::new(
85 io::ErrorKind::InvalidInput,
86 "topic cannot be empty",
87 ));
88 }
89 if trimmed.len() > MAX_SESSION_TOPIC_LEN {
90 return Err(io::Error::new(
91 io::ErrorKind::InvalidInput,
92 format!(
93 "topic name too long (max {} chars, got {})",
94 MAX_SESSION_TOPIC_LEN,
95 trimmed.len()
96 ),
97 ));
98 }
99 if trimmed.contains('/') || trimmed.contains('\\') || trimmed.contains("..") {
100 return Err(io::Error::new(
101 io::ErrorKind::InvalidInput,
102 "topic contains invalid path characters",
103 ));
104 }
105 if !trimmed
106 .chars()
107 .all(|ch| ch.is_ascii_alphanumeric() || ch == '-' || ch == '_')
108 {
109 return Err(io::Error::new(
110 io::ErrorKind::InvalidInput,
111 "topic must contain only alphanumeric, dash, or underscore characters",
112 ));
113 }
114 Ok(trimmed)
115}
116
117pub fn validate_memory_title(title: &str) -> io::Result<&str> {
118 let trimmed = title.trim();
119 if trimmed.is_empty() {
120 return Err(io::Error::new(
121 io::ErrorKind::InvalidInput,
122 "title cannot be empty",
123 ));
124 }
125 if trimmed.chars().count() > MAX_MEMORY_TITLE_LEN {
126 return Err(io::Error::new(
127 io::ErrorKind::InvalidInput,
128 format!("title too long (max {} chars)", MAX_MEMORY_TITLE_LEN),
129 ));
130 }
131 Ok(trimmed)
132}
133
134pub fn normalize_tag(tag: &str) -> Option<String> {
135 let trimmed = tag.trim();
136 if trimmed.is_empty() {
137 return None;
138 }
139 let mut out = String::with_capacity(trimmed.len());
140 let mut prev_dash = false;
141 for ch in trimmed.chars() {
142 let normalized = match ch {
143 'A'..='Z' => ch.to_ascii_lowercase(),
144 'a'..='z' | '0'..='9' => ch,
145 '-' | '_' | ' ' | '.' | '/' => '-',
146 _ => continue,
147 };
148 if normalized == '-' {
149 if prev_dash {
150 continue;
151 }
152 prev_dash = true;
153 out.push(normalized);
154 } else {
155 prev_dash = false;
156 out.push(normalized);
157 }
158 }
159 let normalized = out.trim_matches('-').to_string();
160 (!normalized.is_empty()).then_some(normalized)
161}
162
163pub fn normalize_tags<I, S>(tags: I) -> Vec<String>
164where
165 I: IntoIterator<Item = S>,
166 S: AsRef<str>,
167{
168 let mut seen = BTreeSet::new();
169 for tag in tags {
170 if let Some(tag) = normalize_tag(tag.as_ref()) {
171 seen.insert(tag);
172 if seen.len() >= MAX_MEMORY_TAGS {
173 break;
174 }
175 }
176 }
177 seen.into_iter().collect()
178}
179
180pub fn truncate_chars(value: &str, max_chars: usize) -> (String, bool) {
181 let mut out = String::new();
182 for (count, ch) in value.chars().enumerate() {
183 if count >= max_chars {
184 return (out, true);
185 }
186 out.push(ch);
187 }
188 (out, false)
189}
190
191pub fn count_chars(value: &str) -> usize {
192 value.chars().count()
193}
194
195pub fn now_rfc3339() -> String {
196 Utc::now().to_rfc3339()
197}
198
199pub fn derive_summary(content: &str, max_chars: usize) -> String {
200 let collapsed = content
201 .lines()
202 .map(str::trim)
203 .filter(|line| !line.is_empty())
204 .collect::<Vec<_>>()
205 .join(" ");
206 let (summary, truncated) = truncate_chars(&collapsed, max_chars);
207 if truncated {
208 format!("{}...", summary.trim_end())
209 } else {
210 summary
211 }
212}
213
214pub fn extract_keywords(title: &str, content: &str, tags: &[String]) -> Vec<String> {
215 let mut seen = BTreeSet::new();
216 for tag in tags {
217 if let Some(tag) = normalize_tag(tag) {
218 seen.insert(tag);
219 }
220 }
221
222 let combined = format!("{}\n{}", title, content);
223 let mut current = String::new();
224 for ch in combined.chars() {
225 if ch.is_ascii_alphanumeric() {
226 current.push(ch.to_ascii_lowercase());
227 continue;
228 }
229 if current.len() >= 3 {
230 seen.insert(current.clone());
231 }
232 current.clear();
233 }
234 if current.len() >= 3 {
235 seen.insert(current);
236 }
237
238 seen.into_iter().take(128).collect()
239}
240
241pub fn detect_entities(title: &str, content: &str) -> Vec<String> {
242 let mut entities = BTreeSet::new();
243 for token in format!("{}\n{}", title, content)
244 .split(|ch: char| !(ch.is_ascii_alphanumeric() || ch == '-' || ch == '_' || ch == '/'))
245 {
246 let trimmed = token.trim();
247 if trimmed.len() < 3 {
248 continue;
249 }
250 let has_upper = trimmed.chars().any(|ch| ch.is_ascii_uppercase());
251 let has_separator = trimmed.contains('-') || trimmed.contains('_') || trimmed.contains('/');
252 if has_upper || has_separator {
253 entities.insert(trimmed.to_string());
254 }
255 }
256 entities.into_iter().take(64).collect()
257}
258
259pub fn sanitize_component(input: &str) -> String {
260 let trimmed = input.trim();
261 if trimmed.is_empty() {
262 return "unknown".to_string();
263 }
264
265 let mut out = String::with_capacity(trimmed.len());
266 let mut prev_dash = false;
267 for ch in trimmed.chars() {
268 let normalized = match ch {
269 'A'..='Z' => ch.to_ascii_lowercase(),
270 'a'..='z' | '0'..='9' => ch,
271 _ => '-',
272 };
273 if normalized == '-' {
274 if prev_dash {
275 continue;
276 }
277 prev_dash = true;
278 out.push('-');
279 } else {
280 prev_dash = false;
281 out.push(normalized);
282 }
283 }
284
285 let out = out.trim_matches('-').to_string();
286 if out.is_empty() {
287 "unknown".to_string()
288 } else {
289 out
290 }
291}
292
293pub fn project_key_from_path(path: &Path) -> String {
294 let canonical = std::fs::canonicalize(path).unwrap_or_else(|_| path.to_path_buf());
295
296 if let Some(root) = find_git_root(&canonical) {
297 if let Some(name) = root.file_name().and_then(|value| value.to_str()) {
298 let mut key = sanitize_component(name);
299 if let Some(hash) =
300 short_stable_hash(&bamboo_infrastructure::paths::path_to_display_string(&root))
301 {
302 key.push('-');
303 key.push_str(&hash);
304 }
305 return key;
306 }
307 }
308
309 if let Some(name) = canonical.file_name().and_then(|value| value.to_str()) {
310 let mut key = sanitize_component(name);
311 if let Some(hash) = short_stable_hash(
312 &bamboo_infrastructure::paths::path_to_display_string(&canonical),
313 ) {
314 key.push('-');
315 key.push_str(&hash);
316 }
317 return key;
318 }
319
320 let raw = bamboo_infrastructure::paths::path_to_display_string(&canonical);
321 format!(
322 "path-{}",
323 short_stable_hash(&raw).unwrap_or_else(|| "unknown".to_string())
324 )
325}
326
327pub fn find_git_root(start: &Path) -> Option<PathBuf> {
328 for ancestor in start.ancestors() {
329 let git_dir = ancestor.join(".git");
330 if git_dir.is_dir() || git_dir.is_file() {
331 return Some(ancestor.to_path_buf());
332 }
333 }
334 None
335}
336
337pub fn short_stable_hash(input: &str) -> Option<String> {
338 use std::hash::{Hash, Hasher};
339
340 let trimmed = input.trim();
341 if trimmed.is_empty() {
342 return None;
343 }
344 let mut hasher = std::collections::hash_map::DefaultHasher::new();
345 trimmed.hash(&mut hasher);
346 Some(format!("{:08x}", (hasher.finish() & 0xffff_ffff) as u32))
347}
348
349pub fn build_yaml_frontmatter(frontmatter: &DurableMemoryFrontmatter) -> io::Result<String> {
350 serde_yaml::to_string(frontmatter).map_err(|error| {
351 io::Error::new(
352 io::ErrorKind::InvalidData,
353 format!("failed to serialize memory frontmatter: {error}"),
354 )
355 })
356}
357
358pub fn parse_markdown_document(content: &str) -> io::Result<(DurableMemoryFrontmatter, String)> {
359 let trimmed = content.trim_start_matches('\u{feff}');
360 let Some(rest) = trimmed.strip_prefix("---\n") else {
361 return Err(io::Error::new(
362 io::ErrorKind::InvalidData,
363 "missing frontmatter start marker",
364 ));
365 };
366 let Some(end_idx) = rest.find("\n---\n") else {
367 return Err(io::Error::new(
368 io::ErrorKind::InvalidData,
369 "missing frontmatter end marker",
370 ));
371 };
372 let yaml = &rest[..end_idx];
373 let body = &rest[end_idx + "\n---\n".len()..];
374 let frontmatter: DurableMemoryFrontmatter = serde_yaml::from_str(yaml).map_err(|error| {
375 io::Error::new(
376 io::ErrorKind::InvalidData,
377 format!("failed to parse memory frontmatter: {error}"),
378 )
379 })?;
380 Ok((frontmatter, body.trim().to_string()))
381}
382
383pub fn render_markdown_document(
384 frontmatter: &DurableMemoryFrontmatter,
385 body: &str,
386) -> io::Result<String> {
387 let yaml = build_yaml_frontmatter(frontmatter)?;
388 Ok(format!("---\n{}---\n\n{}\n", yaml, body.trim()))
389}
390
391#[derive(Debug, Clone, Serialize, Deserialize, Default)]
392pub struct LexicalIndex {
393 pub generated_at: String,
394 pub items: Vec<LexicalIndexItem>,
395}
396
397#[derive(Debug, Clone, Serialize, Deserialize)]
398pub struct LexicalIndexItem {
399 pub id: String,
400 pub title: String,
401 pub scope: MemoryScope,
402 pub project_key: Option<String>,
403 pub r#type: DurableMemoryType,
404 pub status: DurableMemoryStatus,
405 pub tags: Vec<String>,
406 pub keywords: Vec<String>,
407 pub entities: Vec<String>,
408 pub updated_at: String,
409 pub created_at: String,
410 pub summary: String,
411}
412
413#[derive(Debug, Clone, Serialize, Deserialize, Default)]
414pub struct RecentIndex {
415 pub generated_at: String,
416 pub items: Vec<RecentIndexItem>,
417}
418
419#[derive(Debug, Clone, Serialize, Deserialize)]
420pub struct RecentIndexItem {
421 pub id: String,
422 pub title: String,
423 pub updated_at: String,
424 pub last_accessed_at: Option<String>,
425 pub status: DurableMemoryStatus,
426}
427
428#[derive(Debug, Clone, Serialize, Deserialize, Default)]
429pub struct GraphIndex {
430 pub generated_at: String,
431 pub items: Vec<GraphIndexItem>,
432}
433
434#[derive(Debug, Clone, Serialize, Deserialize)]
435pub struct GraphIndexItem {
436 pub id: String,
437 pub related: Vec<String>,
438 pub supersedes: Vec<String>,
439 pub contradicted_by: Vec<String>,
440}
441
442#[derive(Debug, Clone, Serialize, Deserialize, Default)]
443pub struct StaleCandidatesIndex {
444 pub generated_at: String,
445 pub items: Vec<StaleCandidateItem>,
446}
447
448#[derive(Debug, Clone, Serialize, Deserialize)]
449pub struct StaleCandidateItem {
450 pub id: String,
451 pub title: String,
452 pub status: DurableMemoryStatus,
453 pub updated_at: String,
454 pub reason: String,
455}
456
457#[derive(Debug, Clone, Serialize, Deserialize, Default)]
458pub struct TaxonomyIndex {
459 pub generated_at: String,
460 pub by_type: BTreeMap<String, usize>,
461 pub by_status: BTreeMap<String, usize>,
462 pub by_scope: BTreeMap<String, usize>,
463 pub total: usize,
464}
465
466#[derive(Debug, Clone, Serialize, Deserialize)]
467pub struct AuditLogEntry {
468 pub timestamp: String,
469 pub action: String,
470 pub scope: MemoryScope,
471 pub memory_id: Option<String>,
472 pub session_id: Option<String>,
473 pub topic: Option<String>,
474 pub summary: String,
475 #[serde(default, skip_serializing_if = "Option::is_none")]
476 pub metadata: Option<serde_json::Value>,
477}
478
479pub fn parse_rfc3339(value: &str) -> Option<DateTime<Utc>> {
480 chrono::DateTime::parse_from_rfc3339(value)
481 .ok()
482 .map(|dt| dt.with_timezone(&Utc))
483}
484
485pub fn sort_memories_desc(memories: &mut [DurableMemoryDocument]) {
486 memories.sort_by(|left, right| {
487 let left_dt =
488 parse_rfc3339(&left.frontmatter.updated_at).unwrap_or(DateTime::<Utc>::MIN_UTC);
489 let right_dt =
490 parse_rfc3339(&right.frontmatter.updated_at).unwrap_or(DateTime::<Utc>::MIN_UTC);
491 right_dt
492 .cmp(&left_dt)
493 .then_with(|| left.frontmatter.id.cmp(&right.frontmatter.id))
494 });
495}
496
497pub fn match_memory_query(
498 doc: &DurableMemoryDocument,
499 query: Option<&str>,
500 filter_types: Option<&HashSet<DurableMemoryType>>,
501 filter_statuses: Option<&HashSet<DurableMemoryStatus>>,
502) -> Option<f64> {
503 if let Some(types) = filter_types {
504 if !types.contains(&doc.frontmatter.r#type) {
505 return None;
506 }
507 }
508 if let Some(statuses) = filter_statuses {
509 if !statuses.contains(&doc.frontmatter.status) {
510 return None;
511 }
512 }
513
514 let Some(query) = query.map(str::trim).filter(|value| !value.is_empty()) else {
515 return Some(1.0);
516 };
517
518 let query_tokens = extract_keywords(query, "", &[]);
519 if query_tokens.is_empty() {
520 return Some(1.0);
521 }
522
523 let title = doc.frontmatter.title.to_ascii_lowercase();
524 let body = doc.body.to_ascii_lowercase();
525 let keywords: HashSet<String> = doc
526 .frontmatter
527 .retrieval
528 .keywords
529 .iter()
530 .map(|value| value.to_ascii_lowercase())
531 .collect();
532 let tags: HashSet<String> = doc
533 .frontmatter
534 .tags
535 .iter()
536 .map(|value| value.to_ascii_lowercase())
537 .collect();
538 let entities: HashSet<String> = doc
539 .frontmatter
540 .retrieval
541 .entities
542 .iter()
543 .map(|value| value.to_ascii_lowercase())
544 .collect();
545
546 let mut score = 0.0;
547 let mut matched_any = false;
548 for token in &query_tokens {
549 let mut token_score = 0.0;
550 if title.contains(token) {
551 token_score += 3.0;
552 }
553 if keywords.contains(token) {
554 token_score += 2.5;
555 }
556 if tags.contains(token) {
557 token_score += 2.0;
558 }
559 if entities.contains(token) {
560 token_score += 1.5;
561 }
562 if body.contains(token) {
563 token_score += 1.0;
564 }
565 if token_score > 0.0 {
566 matched_any = true;
567 score += token_score;
568 }
569 }
570
571 matched_any.then_some(score / query_tokens.len() as f64)
572}
573
574pub fn build_memory_markdown_view(
575 scope: MemoryScope,
576 project_key: Option<&str>,
577 docs: &[DurableMemoryDocument],
578) -> String {
579 let title = match scope {
580 MemoryScope::Global => "# Bamboo Memory Index (Global)".to_string(),
581 MemoryScope::Project => format!(
582 "# Bamboo Memory Index (Project: {})",
583 project_key.unwrap_or("unknown")
584 ),
585 MemoryScope::Session => "# Bamboo Memory Index (Session)".to_string(),
586 };
587 let mut out = String::new();
588 out.push_str(&title);
589 out.push_str("\n\n");
590 if docs.is_empty() {
591 out.push_str("_(empty)_\n");
592 return out;
593 }
594
595 for doc in docs {
596 out.push_str(&format!(
597 "- `{}` {} [{} / {}] updated {}\n",
598 doc.frontmatter.id,
599 doc.frontmatter.title,
600 doc.frontmatter.r#type.as_str(),
601 doc.frontmatter.status.as_str(),
602 doc.frontmatter.updated_at,
603 ));
604 let summary = derive_summary(&doc.body, 160);
605 if !summary.is_empty() {
606 out.push_str(&format!(" - {}\n", summary));
607 }
608 }
609 out
610}
611
612pub fn build_recent_markdown_view(docs: &[DurableMemoryDocument]) -> String {
613 let mut out = String::from("# Recent Memory Updates\n\n");
614 if docs.is_empty() {
615 out.push_str("_(empty)_\n");
616 return out;
617 }
618 for doc in docs.iter().take(20) {
619 out.push_str(&format!(
620 "- `{}` {} — {}\n",
621 doc.frontmatter.id, doc.frontmatter.title, doc.frontmatter.updated_at
622 ));
623 }
624 out
625}
626
627pub fn build_stale_markdown_view(docs: &[DurableMemoryDocument]) -> String {
628 let mut out = String::from("# Stale Memory Candidates\n\n");
629 let stale: Vec<_> = docs
630 .iter()
631 .filter(|doc| doc.frontmatter.status != DurableMemoryStatus::Active)
632 .collect();
633 if stale.is_empty() {
634 out.push_str("_(no stale items)_\n");
635 return out;
636 }
637 for doc in stale {
638 out.push_str(&format!(
639 "- `{}` {} [{}]\n",
640 doc.frontmatter.id,
641 doc.frontmatter.title,
642 doc.frontmatter.status.as_str()
643 ));
644 }
645 out
646}
647
648pub fn build_dream_view(existing: Option<&str>) -> String {
649 match existing.map(str::trim).filter(|value| !value.is_empty()) {
650 Some(value) => value.to_string(),
651 None => "# Bamboo Dream Notebook\n\n_(empty)_\n".to_string(),
652 }
653}
654
655pub fn parse_query_cursor(cursor: Option<&str>) -> usize {
656 cursor
657 .and_then(|raw| raw.rsplit(':').next())
658 .and_then(|raw| raw.parse::<usize>().ok())
659 .unwrap_or(0)
660}
661
662pub fn make_query_cursor(scope: MemoryScope, offset: usize) -> String {
663 format!("{}:{}", scope.as_str(), offset)
664}
665
666pub fn summary_json(items: usize, total: usize) -> String {
667 if total == 0 {
668 "No matching memories found.".to_string()
669 } else {
670 format!("Returned top {} of {} matching memories.", items, total)
671 }
672}
673
674#[cfg(test)]
675mod tests {
676 use super::*;
677
678 #[test]
679 fn normalize_tags_dedupes_and_sanitizes() {
680 let tags = normalize_tags(["User Preference", "user-preference", "release/freeze"]);
681 assert_eq!(tags, vec!["release-freeze", "user-preference"]);
682 }
683
684 #[test]
685 fn project_key_from_path_is_stable() {
686 let key = project_key_from_path(Path::new("/tmp/My Project"));
687 assert!(key.starts_with("my-project-"));
688 }
689
690 #[test]
691 fn parse_markdown_document_requires_frontmatter() {
692 let result = parse_markdown_document("plain body");
693 assert!(result.is_err());
694 }
695}