feat(memory): tantivy-jieba TantivyStore (CC-Adv alternative)

crates/kestrel-memory/Cargo.toml

@@ -18,11 +18,10 @@ tracing = { workspace = true }
 toml = { workspace = true }
 dirs = { workspace = true }
 lru = { workspace = true }
-lancedb = { workspace = true }
-arrow-array = { workspace = true }
-arrow-schema = { workspace = true }
-futures = { workspace = true }
 fs4 = { workspace = true }
+tantivy = "0.26"
+tantivy-jieba = "0.19"
 
 [dev-dependencies]
 tempfile = { workspace = true }
+futures = { workspace = true }

crates/kestrel-memory/src/config.rs

@@ -12,8 +12,7 @@ use std::path::PathBuf;
 /// ```toml
 /// max_entries = 1000
 /// hot_store_path = "/home/user/.kestrel/memory/hot.jsonl"
-/// warm_store_path = "/home/user/.kestrel/memory/warm"
-/// embedding_dim = 1536
+/// tantivy_index_path = "/home/user/.kestrel/memory/tantivy"
 /// ```
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct MemoryConfig {
@@ -25,13 +24,9 @@ pub struct MemoryConfig {
     #[serde(default = "default_hot_store_path")]
     pub hot_store_path: PathBuf,
 
-    /// Path to the warm store data directory.
-    #[serde(default = "default_warm_store_path")]
-    pub warm_store_path: PathBuf,
-
-    /// Dimension of embedding vectors for semantic search.
-    #[serde(default = "default_embedding_dim")]
-    pub embedding_dim: usize,
+    /// Path to the tantivy full-text search index directory.
+    #[serde(default = "default_tantivy_index_path")]
+    pub tantivy_index_path: PathBuf,
 
     /// Character budget for recalled memory content injected into prompts.
     #[serde(default = "default_memory_char_budget")]
@@ -54,16 +49,12 @@ fn default_hot_store_path() -> PathBuf {
         .join("hot.jsonl")
 }
 
-fn default_warm_store_path() -> PathBuf {
+fn default_tantivy_index_path() -> PathBuf {
     dirs::home_dir()
         .unwrap_or_else(|| PathBuf::from("."))
         .join(".kestrel")
         .join("memory")
-        .join("warm")
-}
-
-fn default_embedding_dim() -> usize {
-    1536
+        .join("tantivy")
 }
 
 fn default_memory_char_budget() -> usize {
@@ -79,8 +70,7 @@ impl Default for MemoryConfig {
         Self {
             max_entries: default_max_entries(),
             hot_store_path: default_hot_store_path(),
-            warm_store_path: default_warm_store_path(),
-            embedding_dim: default_embedding_dim(),
+            tantivy_index_path: default_tantivy_index_path(),
             memory_char_budget: default_memory_char_budget(),
             memory_char_budget_overflow: default_memory_char_budget_overflow(),
         }
@@ -93,8 +83,7 @@ impl MemoryConfig {
         Self {
             max_entries: 100,
             hot_store_path: temp_dir.join("hot.jsonl"),
-            warm_store_path: temp_dir.join("warm"),
-            embedding_dim: 8,
+            tantivy_index_path: temp_dir.join("tantivy"),
             memory_char_budget: default_memory_char_budget(),
             memory_char_budget_overflow: default_memory_char_budget_overflow(),
         }
@@ -119,12 +108,11 @@ mod tests {
     fn test_default_config() {
         let config = MemoryConfig::default();
         assert_eq!(config.max_entries, 1000);
-        assert_eq!(config.embedding_dim, 1536);
         assert_eq!(config.memory_char_budget, 2200);
         assert_eq!(config.memory_char_budget_overflow, 1375);
         assert!(config.hot_store_path.to_string_lossy().contains(".kestrel"));
         assert!(config
-            .warm_store_path
+            .tantivy_index_path
             .to_string_lossy()
             .contains(".kestrel"));
     }
@@ -134,38 +122,33 @@ mod tests {
         let temp = std::env::temp_dir();
         let config = MemoryConfig::for_test(&temp);
         assert_eq!(config.max_entries, 100);
-        assert_eq!(config.embedding_dim, 8);
         assert!(config.hot_store_path.starts_with(&temp));
-        assert!(config.warm_store_path.starts_with(&temp));
+        assert!(config.tantivy_index_path.starts_with(&temp));
     }
 
     #[test]
     fn test_toml_roundtrip() {
         let config = MemoryConfig {
             max_entries: 500,
             hot_store_path: PathBuf::from("/tmp/hot.jsonl"),
-            warm_store_path: PathBuf::from("/tmp/warm"),
-            embedding_dim: 768,
+            tantivy_index_path: PathBuf::from("/tmp/tantivy"),
             memory_char_budget: 3000,
             memory_char_budget_overflow: 1500,
         };
         let toml_str = config.to_toml().unwrap();
         let parsed = MemoryConfig::from_toml(&toml_str).unwrap();
         assert_eq!(parsed.max_entries, 500);
-        assert_eq!(parsed.embedding_dim, 768);
         assert_eq!(parsed.memory_char_budget, 3000);
         assert_eq!(parsed.memory_char_budget_overflow, 1500);
         assert_eq!(parsed.hot_store_path, PathBuf::from("/tmp/hot.jsonl"));
-        assert_eq!(parsed.warm_store_path, PathBuf::from("/tmp/warm"));
+        assert_eq!(parsed.tantivy_index_path, PathBuf::from("/tmp/tantivy"));
     }
 
     #[test]
     fn test_toml_parse_partial() {
         let toml_str = "max_entries = 42";
         let config = MemoryConfig::from_toml(toml_str).unwrap();
         assert_eq!(config.max_entries, 42);
-        // Other fields get defaults
-        assert_eq!(config.embedding_dim, 1536);
         assert_eq!(config.memory_char_budget, 2200);
         assert_eq!(config.memory_char_budget_overflow, 1375);
     }

crates/kestrel-memory/src/error.rs

@@ -26,22 +26,13 @@ pub enum MemoryError {
         current: usize,
     },
 
-    /// An invalid embedding vector was provided.
-    #[error("Invalid embedding: expected dimension {expected}, got {actual}")]
-    InvalidEmbedding {
-        /// Expected embedding dimension.
-        expected: usize,
-        /// Actual embedding dimension provided.
-        actual: usize,
-    },
-
     /// A configuration error occurred.
     #[error("Configuration error: {0}")]
     Config(String),
 
-    /// A LanceDB error occurred.
-    #[error("LanceDB error: {0}")]
-    LanceDb(String),
+    /// A search engine (tantivy) error occurred.
+    #[error("Search engine error: {0}")]
+    SearchEngine(String),
 
     /// A security violation was detected in a memory entry.
     #[error("Security violation: {0}")]
@@ -70,18 +61,11 @@ mod tests {
         };
         assert!(err.to_string().contains("100"));
 
-        let err = MemoryError::InvalidEmbedding {
-            expected: 1536,
-            actual: 512,
-        };
-        assert!(err.to_string().contains("1536"));
-        assert!(err.to_string().contains("512"));
-
         let err = MemoryError::Config("bad config".to_string());
         assert!(err.to_string().contains("bad config"));
 
-        let err = MemoryError::LanceDb("table not found".to_string());
-        assert!(err.to_string().contains("table not found"));
+        let err = MemoryError::SearchEngine("index corrupted".to_string());
+        assert!(err.to_string().contains("index corrupted"));
     }
 
     #[test]