vipune 0.7.0

A minimal memory layer for AI agents
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

//! `vipune doctor --embeddings` handler.
//!
//! Reports per-project: total / real / mock / unknown rows using the L2-norm classifier.

use crate::errors::Error;
use crate::output::{DoctorResponse, print_json};
use crate::sqlite::Database;
use crate::sqlite::embedding::classify_embedding;
use std::path::Path;
use std::process::ExitCode;

/// Wrap a database error, converting SQLITE_BUSY into the actionable MCP-server message.
fn wrap_busy<T>(result: Result<T, Error>) -> Result<T, Error> {
    match result {
        Ok(v) => Ok(v),
        Err(Error::SqliteModule(msg)) if msg.contains("database is locked") => {
            Err(Error::Config(
                "Database is locked. Another process (likely the MCP server) is holding a lock. Stop the MCP server and retry.".to_string()
            ))
        }
        Err(e) => Err(e),
    }
}

/// Run the embedding doctor check on the database.
///
/// # Arguments
///
/// * `db_path` - Path to the SQLite database
/// * `project_filter` - If Some, only check this project; if None, check all projects
/// * `json` - If true, output JSON; otherwise human-readable
///
/// # Errors
///
/// Returns error if the database cannot be opened or queried.
pub fn handle_doctor(
    db_path: &Path,
    project_filter: Option<&str>,
    json: bool,
) -> Result<ExitCode, Error> {
    // Open database
    let db = Database::open(db_path).map_err(|e| {
        let err_msg = e.to_string();
        if err_msg.contains("database is locked") {
            return Error::Config(
                "Database is locked. Another process (likely the MCP server) is holding a lock. Stop the MCP server and retry.".to_string()
            );
        }
        Error::Config(err_msg)
    })?;

    // Determine which projects to audit
    let projects: Vec<String> = if let Some(filter) = project_filter {
        vec![filter.to_string()]
    } else {
        wrap_busy(db.list_all_project_ids().map_err(Error::from))?
    };

    if projects.is_empty() {
        if json {
            print_json(&[DoctorResponse {
                project_id: "(none)".to_string(),
                total_rows: 0,
                real_rows: 0,
                mock_rows: 0,
                unknown_rows: 0,
            }]);
        } else {
            println!("No projects found in database.");
        }
        return Ok(ExitCode::SUCCESS);
    }

    let mut responses: Vec<DoctorResponse> = vec![];

    for project_id in &projects {
        let result = wrap_busy(audit_project(&db, project_id))?;
        responses.push(DoctorResponse {
            project_id: project_id.clone(),
            total_rows: result.total,
            real_rows: result.real_count,
            mock_rows: result.mock_count,
            unknown_rows: result.unknown_count,
        });

        if !json {
            println!("Project: {}", project_id);
            println!("  Total rows: {}", result.total);
            println!("  Real:     {}", result.real_count);
            println!("  Mock:     {}", result.mock_count);
            println!("  Unknown:  {}", result.unknown_count);
            if result.mock_count > 0 || result.unknown_count > 0 {
                println!(
                    "  → Run 'vipune reindex' to repair mock rows ({} candidates)",
                    result.mock_count
                );
            }
        }
    }

    // Print JSON: single array of all project responses
    if json {
        print_json(&responses);
    }

    Ok(ExitCode::SUCCESS)
}

struct AuditResult {
    total: usize,
    real_count: usize,
    mock_count: usize,
    unknown_count: usize,
}

fn audit_project(db: &Database, project_id: &str) -> Result<AuditResult, Error> {
    let rows = db.list_all_rows_for_project(project_id)?;

    let mut result = AuditResult {
        total: 0,
        real_count: 0,
        mock_count: 0,
        unknown_count: 0,
    };

    for (_id, _content, embedding) in rows {
        result.total += 1;
        match classify_embedding(&embedding) {
            crate::sqlite::embedding::EmbeddingClass::Real => result.real_count += 1,
            crate::sqlite::embedding::EmbeddingClass::Mock => result.mock_count += 1,
            crate::sqlite::embedding::EmbeddingClass::Unknown => result.unknown_count += 1,
        }
    }

    Ok(result)
}

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

    /// Helper: create a test database with known embeddings.
    fn setup_test_db() -> (TempDir, Database) {
        let dir = TempDir::new().unwrap();
        let path = dir.path().join("test.db");
        let db = Database::open(&path).unwrap();
        (dir, db)
    }

    #[test]
    fn test_doctor_empty_database() {
        let (_dir, db) = setup_test_db();
        let result = audit_project(&db, "nonexistent-project").unwrap();
        assert_eq!(result.total, 0);
        assert_eq!(result.real_count, 0);
        assert_eq!(result.mock_count, 0);
        assert_eq!(result.unknown_count, 0);
    }

    #[test]
    fn test_doctor_classifies_real_embeddings() {
        let (_dir, db) = setup_test_db();
        // Insert a real (L2-normalised) embedding
        let mut vec = vec![0.0f32; 384];
        vec[0] = 1.0; // norm = 1.0 → Real
        db.insert("test-proj", "real memory", &vec, None, "fact", "active")
            .unwrap();

        let result = audit_project(&db, "test-proj").unwrap();
        assert_eq!(result.total, 1);
        assert_eq!(result.real_count, 1);
        assert_eq!(result.mock_count, 0);
        assert_eq!(result.unknown_count, 0);
    }

    #[test]
    fn test_doctor_classifies_mock_embeddings() {
        let (_dir, db) = setup_test_db();
        // Insert a mock-like embedding (uniform ones, norm ≈ 19.6)
        let vec = vec![1.0f32; 384];
        db.insert("test-proj", "mock memory", &vec, None, "fact", "active")
            .unwrap();

        let result = audit_project(&db, "test-proj").unwrap();
        assert_eq!(result.total, 1);
        assert_eq!(result.real_count, 0);
        assert_eq!(result.mock_count, 1);
        assert_eq!(result.unknown_count, 0);
    }

    #[test]
    fn test_doctor_mixed_classifications() {
        let (_dir, db) = setup_test_db();
        // Real: norm = 1.0
        let mut real_vec = vec![0.0f32; 384];
        real_vec[0] = 1.0;
        db.insert("proj", "real", &real_vec, None, "fact", "active")
            .unwrap();
        // Mock: norm ≈ 19.6
        let mock_vec = vec![1.0f32; 384];
        db.insert("proj", "mock", &mock_vec, None, "fact", "active")
            .unwrap();
        // Unknown: norm = 0
        let unknown_vec = vec![0.0f32; 384];
        db.insert("proj", "unknown", &unknown_vec, None, "fact", "active")
            .unwrap();

        let result = audit_project(&db, "proj").unwrap();
        assert_eq!(result.total, 3);
        assert_eq!(result.real_count, 1);
        assert_eq!(result.mock_count, 1);
        assert_eq!(result.unknown_count, 1);
    }
}