hermes_core/index/
metadata.rs

1//! Unified index metadata - segments list + vector index state
2//!
3//! This module manages all index-level metadata in a single `metadata.json` file:
4//! - List of committed segments
5//! - Vector index state per field (Flat/Built)
6//! - Trained centroids/codebooks paths
7//!
8//! The workflow is:
9//! 1. During accumulation: segments store Flat vectors, state is Flat
10//! 2. When threshold crossed: train ONCE, update state to Built
11//! 3. On index open: load metadata, skip re-training if already built
12
13use serde::{Deserialize, Serialize};
14use std::collections::HashMap;
15use std::path::Path;
16
17use crate::dsl::VectorIndexType;
18use crate::error::{Error, Result};
19use crate::schema::Schema;
20
21/// Metadata file name at index level
22pub const INDEX_META_FILENAME: &str = "metadata.json";
23
24/// State of vector index for a field
25#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Default)]
26pub enum VectorIndexState {
27    /// Accumulating vectors - using Flat (brute-force) search
28    #[default]
29    Flat,
30    /// Index structures built - using ANN search
31    Built {
32        /// Total vector count when training happened
33        vector_count: usize,
34        /// Number of clusters used
35        num_clusters: usize,
36    },
37}
38
39/// Per-segment metadata stored in index metadata
40/// This allows merge decisions without loading segment files
41#[derive(Debug, Clone, Serialize, Deserialize)]
42pub struct SegmentMetaInfo {
43    /// Number of documents in this segment
44    pub num_docs: u32,
45}
46
47/// Per-field vector index metadata
48#[derive(Debug, Clone, Serialize, Deserialize)]
49pub struct FieldVectorMeta {
50    /// Field ID
51    pub field_id: u32,
52    /// Configured index type (target type when built)
53    pub index_type: VectorIndexType,
54    /// Current state
55    pub state: VectorIndexState,
56    /// Path to centroids file (relative to index dir)
57    #[serde(skip_serializing_if = "Option::is_none")]
58    pub centroids_file: Option<String>,
59    /// Path to codebook file (relative to index dir, for ScaNN)
60    #[serde(skip_serializing_if = "Option::is_none")]
61    pub codebook_file: Option<String>,
62}
63
64/// Unified index metadata - single source of truth for index state
65#[derive(Debug, Clone, Serialize, Deserialize)]
66pub struct IndexMetadata {
67    /// Version for compatibility
68    pub version: u32,
69    /// Index schema
70    pub schema: Schema,
71    /// Segment metadata: segment_id -> info (doc count, etc.)
72    /// Using HashMap allows O(1) lookup and stores doc counts for merge decisions
73    #[serde(default)]
74    pub segment_metas: HashMap<String, SegmentMetaInfo>,
75    /// Per-field vector index metadata
76    #[serde(default)]
77    pub vector_fields: HashMap<u32, FieldVectorMeta>,
78    /// Total vectors across all segments (updated on commit)
79    #[serde(default)]
80    pub total_vectors: usize,
81}
82
83impl IndexMetadata {
84    /// Create new metadata with schema
85    pub fn new(schema: Schema) -> Self {
86        Self {
87            version: 1,
88            schema,
89            segment_metas: HashMap::new(),
90            vector_fields: HashMap::new(),
91            total_vectors: 0,
92        }
93    }
94
95    /// Get segment IDs as a Vec (for compatibility)
96    pub fn segment_ids(&self) -> Vec<String> {
97        self.segment_metas.keys().cloned().collect()
98    }
99
100    /// Add or update a segment with its doc count
101    pub fn add_segment(&mut self, segment_id: String, num_docs: u32) {
102        self.segment_metas
103            .insert(segment_id, SegmentMetaInfo { num_docs });
104    }
105
106    /// Remove a segment
107    pub fn remove_segment(&mut self, segment_id: &str) {
108        self.segment_metas.remove(segment_id);
109    }
110
111    /// Check if segment exists
112    pub fn has_segment(&self, segment_id: &str) -> bool {
113        self.segment_metas.contains_key(segment_id)
114    }
115
116    /// Get segment doc count
117    pub fn segment_doc_count(&self, segment_id: &str) -> Option<u32> {
118        self.segment_metas.get(segment_id).map(|m| m.num_docs)
119    }
120
121    /// Check if a field has been built
122    pub fn is_field_built(&self, field_id: u32) -> bool {
123        self.vector_fields
124            .get(&field_id)
125            .map(|f| matches!(f.state, VectorIndexState::Built { .. }))
126            .unwrap_or(false)
127    }
128
129    /// Get field metadata
130    pub fn get_field_meta(&self, field_id: u32) -> Option<&FieldVectorMeta> {
131        self.vector_fields.get(&field_id)
132    }
133
134    /// Initialize field metadata (called when field is first seen)
135    pub fn init_field(&mut self, field_id: u32, index_type: VectorIndexType) {
136        self.vector_fields
137            .entry(field_id)
138            .or_insert(FieldVectorMeta {
139                field_id,
140                index_type,
141                state: VectorIndexState::Flat,
142                centroids_file: None,
143                codebook_file: None,
144            });
145    }
146
147    /// Mark field as built with trained structures
148    pub fn mark_field_built(
149        &mut self,
150        field_id: u32,
151        vector_count: usize,
152        num_clusters: usize,
153        centroids_file: String,
154        codebook_file: Option<String>,
155    ) {
156        if let Some(field) = self.vector_fields.get_mut(&field_id) {
157            field.state = VectorIndexState::Built {
158                vector_count,
159                num_clusters,
160            };
161            field.centroids_file = Some(centroids_file);
162            field.codebook_file = codebook_file;
163        }
164    }
165
166    /// Check if field should be built based on threshold
167    pub fn should_build_field(&self, field_id: u32, threshold: usize) -> bool {
168        // Don't build if already built
169        if self.is_field_built(field_id) {
170            return false;
171        }
172        // Build if we have enough vectors
173        self.total_vectors >= threshold
174    }
175
176    /// Load from directory
177    pub async fn load<D: crate::directories::Directory>(dir: &D) -> Result<Self> {
178        let path = Path::new(INDEX_META_FILENAME);
179        let slice = dir.open_read(path).await?;
180        let bytes = slice.read_bytes().await?;
181        serde_json::from_slice(bytes.as_slice()).map_err(|e| Error::Serialization(e.to_string()))
182    }
183
184    /// Save to directory
185    pub async fn save<D: crate::directories::DirectoryWriter>(&self, dir: &D) -> Result<()> {
186        let path = Path::new(INDEX_META_FILENAME);
187        let bytes =
188            serde_json::to_vec_pretty(self).map_err(|e| Error::Serialization(e.to_string()))?;
189        dir.write(path, &bytes).await.map_err(Error::Io)
190    }
191
192    /// Load trained centroids and codebooks from index-level files
193    ///
194    /// Returns (centroids_map, codebooks_map) for fields that are Built
195    pub async fn load_trained_structures<D: crate::directories::Directory>(
196        &self,
197        dir: &D,
198    ) -> (
199        rustc_hash::FxHashMap<u32, std::sync::Arc<crate::structures::CoarseCentroids>>,
200        rustc_hash::FxHashMap<u32, std::sync::Arc<crate::structures::PQCodebook>>,
201    ) {
202        use std::sync::Arc;
203
204        let mut centroids = rustc_hash::FxHashMap::default();
205        let mut codebooks = rustc_hash::FxHashMap::default();
206
207        for (field_id, field_meta) in &self.vector_fields {
208            if !matches!(field_meta.state, VectorIndexState::Built { .. }) {
209                continue;
210            }
211
212            // Load centroids
213            if let Some(ref file) = field_meta.centroids_file
214                && let Ok(slice) = dir.open_read(Path::new(file)).await
215                && let Ok(bytes) = slice.read_bytes().await
216                && let Ok(c) =
217                    serde_json::from_slice::<crate::structures::CoarseCentroids>(bytes.as_slice())
218            {
219                centroids.insert(*field_id, Arc::new(c));
220            }
221
222            // Load codebook (for ScaNN)
223            if let Some(ref file) = field_meta.codebook_file
224                && let Ok(slice) = dir.open_read(Path::new(file)).await
225                && let Ok(bytes) = slice.read_bytes().await
226                && let Ok(c) =
227                    serde_json::from_slice::<crate::structures::PQCodebook>(bytes.as_slice())
228            {
229                codebooks.insert(*field_id, Arc::new(c));
230            }
231        }
232
233        (centroids, codebooks)
234    }
235}
236
237#[cfg(test)]
238mod tests {
239    use super::*;
240
241    fn test_schema() -> Schema {
242        Schema::default()
243    }
244
245    #[test]
246    fn test_metadata_init() {
247        let mut meta = IndexMetadata::new(test_schema());
248        assert_eq!(meta.total_vectors, 0);
249        assert!(meta.segment_metas.is_empty());
250        assert!(!meta.is_field_built(0));
251
252        meta.init_field(0, VectorIndexType::IvfRaBitQ);
253        assert!(!meta.is_field_built(0));
254        assert!(meta.vector_fields.contains_key(&0));
255    }
256
257    #[test]
258    fn test_metadata_segments() {
259        let mut meta = IndexMetadata::new(test_schema());
260        meta.add_segment("abc123".to_string(), 50);
261        meta.add_segment("def456".to_string(), 100);
262        assert_eq!(meta.segment_metas.len(), 2);
263        assert_eq!(meta.segment_doc_count("abc123"), Some(50));
264        assert_eq!(meta.segment_doc_count("def456"), Some(100));
265
266        // Overwrites existing
267        meta.add_segment("abc123".to_string(), 75);
268        assert_eq!(meta.segment_metas.len(), 2);
269        assert_eq!(meta.segment_doc_count("abc123"), Some(75));
270
271        meta.remove_segment("abc123");
272        assert_eq!(meta.segment_metas.len(), 1);
273        assert!(meta.has_segment("def456"));
274        assert!(!meta.has_segment("abc123"));
275    }
276
277    #[test]
278    fn test_mark_field_built() {
279        let mut meta = IndexMetadata::new(test_schema());
280        meta.init_field(0, VectorIndexType::IvfRaBitQ);
281        meta.total_vectors = 10000;
282
283        assert!(!meta.is_field_built(0));
284
285        meta.mark_field_built(0, 10000, 256, "field_0_centroids.bin".to_string(), None);
286
287        assert!(meta.is_field_built(0));
288        let field = meta.get_field_meta(0).unwrap();
289        assert_eq!(
290            field.centroids_file.as_deref(),
291            Some("field_0_centroids.bin")
292        );
293    }
294
295    #[test]
296    fn test_should_build_field() {
297        let mut meta = IndexMetadata::new(test_schema());
298        meta.init_field(0, VectorIndexType::IvfRaBitQ);
299
300        // Below threshold
301        meta.total_vectors = 500;
302        assert!(!meta.should_build_field(0, 1000));
303
304        // Above threshold
305        meta.total_vectors = 1500;
306        assert!(meta.should_build_field(0, 1000));
307
308        // Already built - should not build again
309        meta.mark_field_built(0, 1500, 256, "centroids.bin".to_string(), None);
310        assert!(!meta.should_build_field(0, 1000));
311    }
312
313    #[test]
314    fn test_serialization() {
315        let mut meta = IndexMetadata::new(test_schema());
316        meta.add_segment("seg1".to_string(), 100);
317        meta.init_field(0, VectorIndexType::IvfRaBitQ);
318        meta.total_vectors = 5000;
319
320        let json = serde_json::to_string_pretty(&meta).unwrap();
321        let loaded: IndexMetadata = serde_json::from_str(&json).unwrap();
322
323        assert_eq!(loaded.segment_ids().len(), meta.segment_ids().len());
324        assert_eq!(loaded.segment_doc_count("seg1"), Some(100));
325        assert_eq!(loaded.total_vectors, meta.total_vectors);
326        assert!(loaded.vector_fields.contains_key(&0));
327    }
328}
hermes_core/index/metadata.rs

hermes_core/index/
metadata.rs
