gpt4all/gpt4all-chat/src/database.cpp
AT 861453c4d7
Fixup docx parsing (#3140)
Signed-off-by: Adam Treat <treat.adam@gmail.com>
2024-10-28 13:32:16 -04:00

2751 lines
87 KiB
C++

#include "database.h"
#include "mysettings.h"
#include "utils.h"
#include <duckx/duckx.hpp>
#include <fmt/format.h>
#include <usearch/index_plugins.hpp>
#include <QDebug>
#include <QDir>
#include <QDirIterator>
#include <QFile>
#include <QFileSystemWatcher>
#include <QIODevice>
#include <QPdfDocument>
#include <QPdfSelection>
#include <QRegularExpression>
#include <QSqlError>
#include <QSqlQuery>
#include <QTextStream>
#include <QTimer>
#include <QMap>
#include <QUtf8StringView>
#include <QVariant>
#include <Qt>
#include <QtLogging>
#include <algorithm>
#include <cmath>
#include <optional>
#include <stdexcept>
using namespace Qt::Literals::StringLiterals;
namespace ranges = std::ranges;
namespace us = unum::usearch;
//#define DEBUG
//#define DEBUG_EXAMPLE
namespace {
/* QFile that checks input for binary data. If seen, it fails the read and returns true
* for binarySeen(). */
class BinaryDetectingFile: public QFile {
public:
using QFile::QFile;
bool binarySeen() const { return m_binarySeen; }
protected:
qint64 readData(char *data, qint64 maxSize) override {
qint64 res = QFile::readData(data, maxSize);
return checkData(data, res);
}
qint64 readLineData(char *data, qint64 maxSize) override {
qint64 res = QFile::readLineData(data, maxSize);
return checkData(data, res);
}
private:
qint64 checkData(const char *data, qint64 size) {
Q_ASSERT(!isTextModeEnabled()); // We need raw bytes from the underlying QFile
if (size != -1 && !m_binarySeen) {
for (qint64 i = 0; i < size; i++) {
/* Control characters we should never see in plain text:
* 0x00 NUL - 0x06 ACK
* 0x0E SO - 0x1A SUB
* 0x1C FS - 0x1F US */
auto c = static_cast<unsigned char>(data[i]);
if (c < 0x07 || (c >= 0x0E && c < 0x1B) || (c >= 0x1C && c < 0x20)) {
m_binarySeen = true;
break;
}
}
}
return m_binarySeen ? -1 : size;
}
bool m_binarySeen = false;
};
} // namespace
static int s_batchSize = 100;
static const QString INIT_DB_SQL[] = {
// automatically free unused disk space
u"pragma auto_vacuum = FULL;"_s,
// create tables
uR"(
create table chunks(
id integer primary key autoincrement,
document_id integer not null,
chunk_text text not null,
file text not null,
title text,
author text,
subject text,
keywords text,
page integer,
line_from integer,
line_to integer,
words integer default 0 not null,
tokens integer default 0 not null,
foreign key(document_id) references documents(id)
);
)"_s, uR"(
create virtual table chunks_fts using fts5(
id unindexed,
document_id unindexed,
chunk_text,
file,
title,
author,
subject,
keywords,
content='chunks',
content_rowid='id',
tokenize='porter'
);
)"_s, uR"(
create table collections(
id integer primary key,
name text unique not null,
start_update_time integer,
last_update_time integer,
embedding_model text
);
)"_s, uR"(
create table folders(
id integer primary key autoincrement,
path text unique not null
);
)"_s, uR"(
create table collection_items(
collection_id integer not null,
folder_id integer not null,
foreign key(collection_id) references collections(id)
foreign key(folder_id) references folders(id),
unique(collection_id, folder_id)
);
)"_s, uR"(
create table documents(
id integer primary key,
folder_id integer not null,
document_time integer not null,
document_path text unique not null,
foreign key(folder_id) references folders(id)
);
)"_s, uR"(
create table embeddings(
model text not null,
folder_id integer not null,
chunk_id integer not null,
embedding blob not null,
primary key(model, folder_id, chunk_id),
foreign key(folder_id) references folders(id),
foreign key(chunk_id) references chunks(id),
unique(model, chunk_id)
);
)"_s,
};
static const QString INSERT_CHUNK_SQL = uR"(
insert into chunks(document_id, chunk_text,
file, title, author, subject, keywords, page, line_from, line_to, words)
values(?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
returning id;
)"_s;
static const QString INSERT_CHUNK_FTS_SQL = uR"(
insert into chunks_fts(document_id, chunk_text,
file, title, author, subject, keywords)
values(?, ?, ?, ?, ?, ?, ?);
)"_s;
static const QString SELECT_CHUNKED_DOCUMENTS_SQL[] = {
uR"(
select distinct document_id from chunks;
)"_s, uR"(
select distinct document_id from chunks_fts;
)"_s,
};
static const QString DELETE_CHUNKS_SQL[] = {
uR"(
delete from embeddings
where chunk_id in (
select id from chunks where document_id = ?
);
)"_s, uR"(
delete from chunks where document_id = ?;
)"_s, uR"(
delete from chunks_fts where document_id = ?;
)"_s,
};
static const QString SELECT_CHUNKS_BY_DOCUMENT_SQL = uR"(
select id from chunks WHERE document_id = ?;
)"_s;
static const QString SELECT_CHUNKS_SQL = uR"(
select c.id, d.document_time, d.document_path, c.chunk_text, c.file, c.title, c.author, c.page, c.line_from, c.line_to, co.name
from chunks c
join documents d on d.id = c.document_id
join folders f on f.id = d.folder_id
join collection_items ci on ci.folder_id = f.id
join collections co on co.id = ci.collection_id
where c.id in (%1);
)"_s;
static const QString SELECT_UNCOMPLETED_CHUNKS_SQL = uR"(
select co.name, co.embedding_model, c.id, d.folder_id, c.chunk_text
from chunks c
join documents d on d.id = c.document_id
join folders f on f.id = d.folder_id
join collection_items ci on ci.folder_id = f.id
join collections co on co.id = ci.collection_id and co.embedding_model is not null
where not exists(
select 1
from embeddings e
where e.chunk_id = c.id and e.model = co.embedding_model
);
)"_s;
static const QString SELECT_COUNT_CHUNKS_SQL = uR"(
select count(c.id)
from chunks c
join documents d on d.id = c.document_id
where d.folder_id = ?;
)"_s;
static const QString SELECT_CHUNKS_FTS_SQL = uR"(
select fts.id, bm25(chunks_fts) as score
from chunks_fts fts
join documents d on fts.document_id = d.id
join collection_items ci on d.folder_id = ci.folder_id
join collections co on ci.collection_id = co.id
where chunks_fts match ?
and co.name in ('%1')
order by score limit %2;
)"_s;
#define NAMED_PAIR(name, typea, a, typeb, b) \
struct name { typea a; typeb b; }; \
static bool operator==(const name &x, const name &y) { return x.a == y.a && x.b == y.b; } \
static size_t qHash(const name &x, size_t seed) { return qHashMulti(seed, x.a, x.b); }
// struct compared by embedding key, can be extended with additional unique data
NAMED_PAIR(EmbeddingKey, QString, embedding_model, int, chunk_id)
namespace {
struct IncompleteChunk: EmbeddingKey { int folder_id; QString text; };
} // namespace
static bool selectAllUncompletedChunks(QSqlQuery &q, QHash<IncompleteChunk, QStringList> &chunks)
{
if (!q.exec(SELECT_UNCOMPLETED_CHUNKS_SQL))
return false;
while (q.next()) {
QString collection = q.value(0).toString();
IncompleteChunk ic {
/*EmbeddingKey*/ {
.embedding_model = q.value(1).toString(),
.chunk_id = q.value(2).toInt(),
},
/*folder_id =*/ q.value(3).toInt(),
/*text =*/ q.value(4).toString(),
};
chunks[ic] << collection;
}
return true;
}
static bool selectCountChunks(QSqlQuery &q, int folder_id, int &count)
{
if (!q.prepare(SELECT_COUNT_CHUNKS_SQL))
return false;
q.addBindValue(folder_id);
if (!q.exec())
return false;
if (!q.next()) {
count = 0;
return false;
}
count = q.value(0).toInt();
return true;
}
static bool selectChunk(QSqlQuery &q, const QList<int> &chunk_ids)
{
QString chunk_ids_str = QString::number(chunk_ids[0]);
for (size_t i = 1; i < chunk_ids.size(); ++i)
chunk_ids_str += "," + QString::number(chunk_ids[i]);
const QString formatted_query = SELECT_CHUNKS_SQL.arg(chunk_ids_str);
if (!q.prepare(formatted_query))
return false;
return q.exec();
}
static const QString INSERT_COLLECTION_SQL = uR"(
insert into collections(name, start_update_time, last_update_time, embedding_model)
values(?, ?, ?, ?)
returning id;
)"_s;
static const QString SELECT_FOLDERS_FROM_COLLECTIONS_SQL = uR"(
select f.id, f.path
from collections c
join collection_items ci on ci.collection_id = c.id
join folders f on ci.folder_id = f.id
where c.name = ?;
)"_s;
static const QString SELECT_COLLECTIONS_SQL_V1 = uR"(
select c.collection_name, f.folder_path, f.id
from collections c
join folders f on c.folder_id = f.id
order by c.collection_name asc, f.folder_path asc;
)"_s;
static const QString SELECT_COLLECTIONS_SQL_V2 = uR"(
select c.id, c.name, f.path, f.id, c.start_update_time, c.last_update_time, c.embedding_model
from collections c
join collection_items ci on ci.collection_id = c.id
join folders f on ci.folder_id = f.id
order by c.name asc, f.path asc;
)"_s;
static const QString SELECT_COLLECTION_BY_NAME_SQL = uR"(
select id, name, start_update_time, last_update_time, embedding_model
from collections c
where name = ?;
)"_s;
static const QString SET_COLLECTION_EMBEDDING_MODEL_SQL = uR"(
update collections
set embedding_model = ?
where name = ?;
)"_s;
static const QString UPDATE_START_UPDATE_TIME_SQL = uR"(
update collections set start_update_time = ? where id = ?;
)"_s;
static const QString UPDATE_LAST_UPDATE_TIME_SQL = uR"(
update collections set last_update_time = ? where id = ?;
)"_s;
static const QString FTS_INTEGRITY_SQL = uR"(
insert into chunks_fts(chunks_fts, rank) values('integrity-check', 1);
)"_s;
static const QString FTS_REBUILD_SQL = uR"(
insert into chunks_fts(chunks_fts) values('rebuild');
)"_s;
static bool addCollection(QSqlQuery &q, const QString &collection_name, const QDateTime &start_update,
const QDateTime &last_update, const QString &embedding_model, CollectionItem &item)
{
if (!q.prepare(INSERT_COLLECTION_SQL))
return false;
q.addBindValue(collection_name);
q.addBindValue(start_update);
q.addBindValue(last_update);
q.addBindValue(embedding_model);
if (!q.exec() || !q.next())
return false;
item.collection_id = q.value(0).toInt();
item.collection = collection_name;
item.embeddingModel = embedding_model;
return true;
}
static bool selectFoldersFromCollection(QSqlQuery &q, const QString &collection_name, QList<QPair<int, QString>> *folders)
{
if (!q.prepare(SELECT_FOLDERS_FROM_COLLECTIONS_SQL))
return false;
q.addBindValue(collection_name);
if (!q.exec())
return false;
while (q.next())
folders->append({q.value(0).toInt(), q.value(1).toString()});
return true;
}
static QList<CollectionItem> sqlExtractCollections(QSqlQuery &q, bool with_folder = false, int version = LOCALDOCS_VERSION)
{
QList<CollectionItem> collections;
while (q.next()) {
CollectionItem i;
int idx = 0;
if (version >= 2)
i.collection_id = q.value(idx++).toInt();
i.collection = q.value(idx++).toString();
if (with_folder) {
i.folder_path = q.value(idx++).toString();
i.folder_id = q.value(idx++).toInt();
}
i.indexing = false;
i.installed = true;
if (version >= 2) {
bool ok;
const qint64 start_update = q.value(idx++).toLongLong(&ok);
if (ok) i.startUpdate = QDateTime::fromMSecsSinceEpoch(start_update);
const qint64 last_update = q.value(idx++).toLongLong(&ok);
if (ok) i.lastUpdate = QDateTime::fromMSecsSinceEpoch(last_update);
i.embeddingModel = q.value(idx++).toString();
}
if (i.embeddingModel.isNull()) {
// unknown embedding model -> need to re-index
i.forceIndexing = true;
}
collections << i;
}
return collections;
}
static bool selectAllFromCollections(QSqlQuery &q, QList<CollectionItem> *collections, int version = LOCALDOCS_VERSION)
{
switch (version) {
case 1:
if (!q.prepare(SELECT_COLLECTIONS_SQL_V1))
return false;
break;
case 2:
case 3:
if (!q.prepare(SELECT_COLLECTIONS_SQL_V2))
return false;
break;
default:
Q_UNREACHABLE();
return false;
}
if (!q.exec())
return false;
*collections = sqlExtractCollections(q, true, version);
return true;
}
static bool selectCollectionByName(QSqlQuery &q, const QString &name, std::optional<CollectionItem> &collection)
{
if (!q.prepare(SELECT_COLLECTION_BY_NAME_SQL))
return false;
q.addBindValue(name);
if (!q.exec())
return false;
QList<CollectionItem> collections = sqlExtractCollections(q);
Q_ASSERT(collections.count() <= 1);
collection.reset();
if (!collections.isEmpty())
collection = collections.first();
return true;
}
static bool setCollectionEmbeddingModel(QSqlQuery &q, const QString &collection_name, const QString &embedding_model)
{
if (!q.prepare(SET_COLLECTION_EMBEDDING_MODEL_SQL))
return false;
q.addBindValue(embedding_model);
q.addBindValue(collection_name);
return q.exec();
}
static bool updateStartUpdateTime(QSqlQuery &q, int id, qint64 update_time)
{
if (!q.prepare(UPDATE_START_UPDATE_TIME_SQL))
return false;
q.addBindValue(update_time);
q.addBindValue(id);
return q.exec();
}
static bool updateLastUpdateTime(QSqlQuery &q, int id, qint64 update_time)
{
if (!q.prepare(UPDATE_LAST_UPDATE_TIME_SQL))
return false;
q.addBindValue(update_time);
q.addBindValue(id);
return q.exec();
}
static const QString INSERT_FOLDERS_SQL = uR"(
insert into folders(path) values(?);
)"_s;
static const QString DELETE_FOLDERS_SQL = uR"(
delete from folders where id = ?;
)"_s;
static const QString SELECT_FOLDERS_FROM_PATH_SQL = uR"(
select id from folders where path = ?;
)"_s;
static const QString GET_FOLDER_EMBEDDING_MODEL_SQL = uR"(
select co.embedding_model
from collections co
join collection_items ci on ci.collection_id = co.id
where ci.folder_id = ?;
)"_s;
static const QString FOLDER_REMOVE_ALL_DOCS_SQL[] = {
uR"(
delete from embeddings
where chunk_id in (
select c.id
from chunks c
join documents d on d.id = c.document_id
join folders f on f.id = d.folder_id
where f.path = ?
);
)"_s, uR"(
delete from chunks
where document_id in (
select d.id
from documents d
join folders f on f.id = d.folder_id
where f.path = ?
);
)"_s, uR"(
delete from documents
where id in (
select d.id
from documents d
join folders f on f.id = d.folder_id
where f.path = ?
);
)"_s,
};
static bool addFolderToDB(QSqlQuery &q, const QString &folder_path, int *folder_id)
{
if (!q.prepare(INSERT_FOLDERS_SQL))
return false;
q.addBindValue(folder_path);
if (!q.exec())
return false;
*folder_id = q.lastInsertId().toInt();
return true;
}
static bool removeFolderFromDB(QSqlQuery &q, int folder_id)
{
if (!q.prepare(DELETE_FOLDERS_SQL))
return false;
q.addBindValue(folder_id);
return q.exec();
}
static bool selectFolder(QSqlQuery &q, const QString &folder_path, int *id)
{
if (!q.prepare(SELECT_FOLDERS_FROM_PATH_SQL))
return false;
q.addBindValue(folder_path);
if (!q.exec())
return false;
Q_ASSERT(q.size() < 2);
if (q.next())
*id = q.value(0).toInt();
return true;
}
static bool sqlGetFolderEmbeddingModel(QSqlQuery &q, int id, QString &embedding_model)
{
if (!q.prepare(GET_FOLDER_EMBEDDING_MODEL_SQL))
return false;
q.addBindValue(id);
if (!q.exec() || !q.next())
return false;
// FIXME(jared): there may be more than one if a folder is shared between collections
Q_ASSERT(q.size() < 2);
embedding_model = q.value(0).toString();
return true;
}
static const QString INSERT_COLLECTION_ITEM_SQL = uR"(
insert into collection_items(collection_id, folder_id)
values(?, ?)
on conflict do nothing;
)"_s;
static const QString DELETE_COLLECTION_FOLDER_SQL = uR"(
delete from collection_items
where collection_id = (select id from collections where name = :name) and folder_id = :folder_id
returning (select count(*) from collection_items where folder_id = :folder_id);
)"_s;
static const QString PRUNE_COLLECTIONS_SQL = uR"(
delete from collections
where id not in (select collection_id from collection_items);
)"_s;
// 0 = already exists, 1 = added, -1 = error
static int addCollectionItem(QSqlQuery &q, int collection_id, int folder_id)
{
if (!q.prepare(INSERT_COLLECTION_ITEM_SQL))
return -1;
q.addBindValue(collection_id);
q.addBindValue(folder_id);
if (q.exec())
return q.numRowsAffected();
return -1;
}
// returns the number of remaining references to the folder, or -1 on error
static int removeCollectionFolder(QSqlQuery &q, const QString &collection_name, int folder_id)
{
if (!q.prepare(DELETE_COLLECTION_FOLDER_SQL))
return -1;
q.bindValue(":name", collection_name);
q.bindValue(":folder_id", folder_id);
if (!q.exec() || !q.next())
return -1;
return q.value(0).toInt();
}
static bool sqlPruneCollections(QSqlQuery &q)
{
return q.exec(PRUNE_COLLECTIONS_SQL);
}
static const QString INSERT_DOCUMENTS_SQL = uR"(
insert into documents(folder_id, document_time, document_path) values(?, ?, ?);
)"_s;
static const QString UPDATE_DOCUMENT_TIME_SQL = uR"(
update documents set document_time = ? where id = ?;
)"_s;
static const QString DELETE_DOCUMENTS_SQL = uR"(
delete from documents where id = ?;
)"_s;
static const QString SELECT_DOCUMENT_SQL = uR"(
select id, document_time from documents where document_path = ?;
)"_s;
static const QString SELECT_DOCUMENTS_SQL = uR"(
select id from documents where folder_id = ?;
)"_s;
static const QString SELECT_ALL_DOCUMENTS_SQL = uR"(
select id, document_path from documents;
)"_s;
static const QString SELECT_COUNT_STATISTICS_SQL = uR"(
select count(distinct d.id), sum(c.words), sum(c.tokens)
from documents d
left join chunks c on d.id = c.document_id
where d.folder_id = ?;
)"_s;
static bool addDocument(QSqlQuery &q, int folder_id, qint64 document_time, const QString &document_path, int *document_id)
{
if (!q.prepare(INSERT_DOCUMENTS_SQL))
return false;
q.addBindValue(folder_id);
q.addBindValue(document_time);
q.addBindValue(document_path);
if (!q.exec())
return false;
*document_id = q.lastInsertId().toInt();
return true;
}
static bool removeDocument(QSqlQuery &q, int document_id)
{
if (!q.prepare(DELETE_DOCUMENTS_SQL))
return false;
q.addBindValue(document_id);
return q.exec();
}
static bool updateDocument(QSqlQuery &q, int id, qint64 document_time)
{
if (!q.prepare(UPDATE_DOCUMENT_TIME_SQL))
return false;
q.addBindValue(document_time);
q.addBindValue(id);
return q.exec();
}
static bool selectDocument(QSqlQuery &q, const QString &document_path, int *id, qint64 *document_time)
{
if (!q.prepare(SELECT_DOCUMENT_SQL))
return false;
q.addBindValue(document_path);
if (!q.exec())
return false;
Q_ASSERT(q.size() < 2);
if (q.next()) {
*id = q.value(0).toInt();
*document_time = q.value(1).toLongLong();
}
return true;
}
static bool selectDocuments(QSqlQuery &q, int folder_id, QList<int> *documentIds)
{
if (!q.prepare(SELECT_DOCUMENTS_SQL))
return false;
q.addBindValue(folder_id);
if (!q.exec())
return false;
while (q.next())
documentIds->append(q.value(0).toInt());
return true;
}
static bool selectCountStatistics(QSqlQuery &q, int folder_id, int *total_docs, int *total_words, int *total_tokens)
{
if (!q.prepare(SELECT_COUNT_STATISTICS_SQL))
return false;
q.addBindValue(folder_id);
if (!q.exec())
return false;
if (q.next()) {
*total_docs = q.value(0).toInt();
*total_words = q.value(1).toInt();
*total_tokens = q.value(2).toInt();
}
return true;
}
// insert embedding only if still needed
static const QString INSERT_EMBEDDING_SQL = uR"(
insert into embeddings(model, folder_id, chunk_id, embedding)
select :model, d.folder_id, :chunk_id, :embedding
from chunks c
join documents d on d.id = c.document_id
join folders f on f.id = d.folder_id
join collection_items ci on ci.folder_id = f.id
join collections co on co.id = ci.collection_id
where co.embedding_model = :model and c.id = :chunk_id
limit 1;
)"_s;
static const QString GET_COLLECTION_EMBEDDINGS_SQL = uR"(
select e.chunk_id, e.embedding
from embeddings e
join collections co on co.embedding_model = e.model
join collection_items ci on ci.folder_id = e.folder_id and ci.collection_id = co.id
where co.name in ('%1');
)"_s;
static const QString GET_CHUNK_EMBEDDINGS_SQL = uR"(
select e.chunk_id, e.embedding
from embeddings e
where e.chunk_id in (%1);
)"_s;
static const QString GET_CHUNK_FILE_SQL = uR"(
select file from chunks where id = ?;
)"_s;
namespace {
struct Embedding { QString model; int folder_id; int chunk_id; QByteArray data; };
struct EmbeddingStat { QString lastFile; int nAdded; int nSkipped; };
} // namespace
NAMED_PAIR(EmbeddingFolder, QString, embedding_model, int, folder_id)
static bool sqlAddEmbeddings(QSqlQuery &q, const QList<Embedding> &embeddings, QHash<EmbeddingFolder, EmbeddingStat> &embeddingStats)
{
if (!q.prepare(INSERT_EMBEDDING_SQL))
return false;
// insert embedding if needed
for (const auto &e: embeddings) {
q.bindValue(":model", e.model);
q.bindValue(":chunk_id", e.chunk_id);
q.bindValue(":embedding", e.data);
if (!q.exec())
return false;
auto &stat = embeddingStats[{ e.model, e.folder_id }];
if (q.numRowsAffected()) {
stat.nAdded++; // embedding added
} else {
stat.nSkipped++; // embedding no longer needed
}
}
if (!q.prepare(GET_CHUNK_FILE_SQL))
return false;
// populate statistics for each collection item
for (const auto &e: embeddings) {
auto &stat = embeddingStats[{ e.model, e.folder_id }];
if (stat.nAdded && stat.lastFile.isNull()) {
q.addBindValue(e.chunk_id);
if (!q.exec() || !q.next())
return false;
stat.lastFile = q.value(0).toString();
}
}
return true;
}
void Database::transaction()
{
bool ok = m_db.transaction();
Q_ASSERT(ok);
}
void Database::commit()
{
bool ok = m_db.commit();
Q_ASSERT(ok);
}
void Database::rollback()
{
bool ok = m_db.rollback();
Q_ASSERT(ok);
}
bool Database::refreshDocumentIdCache(QSqlQuery &q)
{
m_documentIdCache.clear();
for (const auto &cmd: SELECT_CHUNKED_DOCUMENTS_SQL) {
if (!q.exec(cmd))
return false;
while (q.next())
m_documentIdCache << q.value(0).toInt();
}
return true;
}
bool Database::addChunk(QSqlQuery &q, int document_id, const QString &chunk_text, const QString &file,
const QString &title, const QString &author, const QString &subject, const QString &keywords,
int page, int from, int to, int words, int *chunk_id)
{
if (!q.prepare(INSERT_CHUNK_SQL))
return false;
q.addBindValue(document_id);
q.addBindValue(chunk_text);
q.addBindValue(file);
q.addBindValue(title);
q.addBindValue(author);
q.addBindValue(subject);
q.addBindValue(keywords);
q.addBindValue(page);
q.addBindValue(from);
q.addBindValue(to);
q.addBindValue(words);
if (!q.exec() || !q.next())
return false;
*chunk_id = q.value(0).toInt();
if (!q.prepare(INSERT_CHUNK_FTS_SQL))
return false;
q.addBindValue(document_id);
q.addBindValue(chunk_text);
q.addBindValue(file);
q.addBindValue(title);
q.addBindValue(author);
q.addBindValue(subject);
q.addBindValue(keywords);
if (!q.exec())
return false;
m_documentIdCache << document_id;
return true;
}
bool Database::removeChunksByDocumentId(QSqlQuery &q, int document_id)
{
for (const auto &cmd: DELETE_CHUNKS_SQL) {
if (!q.prepare(cmd))
return false;
q.addBindValue(document_id);
if (!q.exec())
return false;
}
m_documentIdCache.remove(document_id);
return true;
}
bool Database::sqlRemoveDocsByFolderPath(QSqlQuery &q, const QString &path)
{
for (const auto &cmd: FOLDER_REMOVE_ALL_DOCS_SQL) {
if (!q.prepare(cmd))
return false;
q.addBindValue(path);
if (!q.exec())
return false;
}
return refreshDocumentIdCache(q);
}
bool Database::hasContent()
{
return m_db.tables().contains("chunks", Qt::CaseInsensitive);
}
int Database::openDatabase(const QString &modelPath, bool create, int ver)
{
if (!QFileInfo(modelPath).isDir()) {
qWarning() << "ERROR: invalid download path" << modelPath;
return -1;
}
if (m_db.isOpen())
m_db.close();
auto dbPath = u"%1/localdocs_v%2.db"_s.arg(modelPath).arg(ver);
if (!create && !QFileInfo::exists(dbPath))
return 0;
m_db.setDatabaseName(dbPath);
if (!m_db.open()) {
qWarning() << "ERROR: opening db" << dbPath << m_db.lastError();
return -1;
}
return hasContent();
}
bool Database::openLatestDb(const QString &modelPath, QList<CollectionItem> &oldCollections)
{
/*
* Support upgrade path from older versions:
*
* 1. Detect and load dbPath with older versions
* 2. Provide versioned SQL select statements
* 3. Upgrade the tables to the new version
* 4. By default mark all collections of older versions as force indexing and present to the user
* the an 'update' button letting them know a breaking change happened and that the collection
* will need to be indexed again
*/
int dbVer;
for (dbVer = LOCALDOCS_VERSION;; dbVer--) {
if (dbVer < LOCALDOCS_MIN_VER) return true; // create a new db
int res = openDatabase(modelPath, false, dbVer);
if (res == 1) break; // found one with content
if (res == -1) return false; // error
}
if (dbVer == LOCALDOCS_VERSION) return true; // already up-to-date
// If we're upgrading, then we need to do a select on the current version of the collections table,
// then create the new one and populate the collections table and mark them as needing forced
// indexing
#if defined(DEBUG)
qDebug() << "Older localdocs version found" << dbVer << "upgrade to" << LOCALDOCS_VERSION;
#endif
// Select the current collections which will be marked to force indexing
QSqlQuery q(m_db);
if (!selectAllFromCollections(q, &oldCollections, dbVer)) {
qWarning() << "ERROR: Could not open select old collections" << q.lastError();
return false;
}
m_db.close();
return true;
}
bool Database::initDb(const QString &modelPath, const QList<CollectionItem> &oldCollections)
{
if (!m_db.isOpen()) {
int res = openDatabase(modelPath);
if (res == 1) return true; // already populated
if (res == -1) return false; // error
} else if (hasContent()) {
return true; // already populated
}
transaction();
QSqlQuery q(m_db);
for (const auto &cmd: INIT_DB_SQL) {
if (!q.exec(cmd)) {
qWarning() << "ERROR: failed to create tables" << q.lastError();
rollback();
return false;
}
}
/* These are collection items that came from an older version of localdocs which
* require forced indexing that should only be done when the user has explicitly asked
* for them to be indexed again */
for (const CollectionItem &item : oldCollections) {
if (!addFolder(item.collection, item.folder_path, QString())) {
qWarning() << "ERROR: failed to add previous collections to new database";
rollback();
return false;
}
}
commit();
return true;
}
Database::Database(int chunkSize, QStringList extensions)
: QObject(nullptr)
, m_chunkSize(chunkSize)
, m_scannedFileExtensions(std::move(extensions))
, m_scanIntervalTimer(new QTimer(this))
, m_watcher(new QFileSystemWatcher(this))
, m_embLLM(new EmbeddingLLM)
, m_databaseValid(true)
, m_chunkStreamer(this)
{
m_db = QSqlDatabase::database(QSqlDatabase::defaultConnection, false);
if (!m_db.isValid())
m_db = QSqlDatabase::addDatabase("QSQLITE");
Q_ASSERT(m_db.isValid());
moveToThread(&m_dbThread);
m_dbThread.setObjectName("database");
m_dbThread.start();
}
Database::~Database()
{
m_dbThread.quit();
m_dbThread.wait();
delete m_embLLM;
}
void Database::setStartUpdateTime(CollectionItem &item)
{
QSqlQuery q(m_db);
const qint64 update_time = QDateTime::currentMSecsSinceEpoch();
if (!updateStartUpdateTime(q, item.collection_id, update_time))
qWarning() << "Database ERROR: failed to set start update time:" << q.lastError();
else
item.startUpdate = QDateTime::fromMSecsSinceEpoch(update_time);
}
void Database::setLastUpdateTime(CollectionItem &item)
{
QSqlQuery q(m_db);
const qint64 update_time = QDateTime::currentMSecsSinceEpoch();
if (!updateLastUpdateTime(q, item.collection_id, update_time))
qWarning() << "Database ERROR: failed to set last update time:" << q.lastError();
else
item.lastUpdate = QDateTime::fromMSecsSinceEpoch(update_time);
}
CollectionItem Database::guiCollectionItem(int folder_id) const
{
Q_ASSERT(m_collectionMap.contains(folder_id));
return m_collectionMap.value(folder_id);
}
void Database::updateGuiForCollectionItem(const CollectionItem &item)
{
m_collectionMap.insert(item.folder_id, item);
emit requestUpdateGuiForCollectionItem(item);
}
void Database::addGuiCollectionItem(const CollectionItem &item)
{
m_collectionMap.insert(item.folder_id, item);
emit requestAddGuiCollectionItem(item);
}
void Database::removeGuiFolderById(const QString &collection, int folder_id)
{
emit requestRemoveGuiFolderById(collection, folder_id);
}
void Database::guiCollectionListUpdated(const QList<CollectionItem> &collectionList)
{
for (const auto &i : collectionList)
m_collectionMap.insert(i.folder_id, i);
emit requestGuiCollectionListUpdated(collectionList);
}
void Database::updateFolderToIndex(int folder_id, size_t countForFolder, bool sendChunks)
{
CollectionItem item = guiCollectionItem(folder_id);
item.currentDocsToIndex = countForFolder;
if (!countForFolder) {
if (sendChunks && !m_chunkList.isEmpty())
sendChunkList(); // send any remaining embedding chunks to llm
item.indexing = false;
item.installed = true;
// Set the last update if we are done
if (item.startUpdate > item.lastUpdate && item.currentEmbeddingsToIndex == 0)
setLastUpdateTime(item);
}
updateGuiForCollectionItem(item);
}
static void handleDocumentError(const QString &errorMessage, int document_id, const QString &document_path,
const QSqlError &error)
{
qWarning() << errorMessage << document_id << document_path << error;
}
class DocumentReader {
public:
struct Metadata { QString title, author, subject, keywords; };
static std::unique_ptr<DocumentReader> fromDocument(const DocumentInfo &info);
const DocumentInfo &doc () const { return *m_info; }
const Metadata &metadata() const { return m_metadata; }
const std::optional<QString> &word () const { return m_word; }
const std::optional<QString> &nextWord() { m_word = advance(); return m_word; }
virtual std::optional<ChunkStreamer::Status> getError() const { return std::nullopt; }
virtual int page() const { return -1; }
virtual ~DocumentReader() = default;
protected:
explicit DocumentReader(const DocumentInfo &info)
: m_info(&info) {}
void postInit(Metadata &&metadata = {})
{
m_metadata = std::move(metadata);
m_word = advance();
}
virtual std::optional<QString> advance() = 0;
const DocumentInfo *m_info;
Metadata m_metadata;
std::optional<QString> m_word;
};
namespace {
class PdfDocumentReader final : public DocumentReader {
public:
explicit PdfDocumentReader(const DocumentInfo &info)
: DocumentReader(info)
{
QString path = info.file.canonicalFilePath();
if (m_doc.load(path) != QPdfDocument::Error::None)
throw std::runtime_error(fmt::format("Failed to load PDF: {}", path));
Metadata metadata {
.title = m_doc.metaData(QPdfDocument::MetaDataField::Title ).toString(),
.author = m_doc.metaData(QPdfDocument::MetaDataField::Author ).toString(),
.subject = m_doc.metaData(QPdfDocument::MetaDataField::Subject ).toString(),
.keywords = m_doc.metaData(QPdfDocument::MetaDataField::Keywords).toString(),
};
postInit(std::move(metadata));
}
int page() const override { return m_currentPage; }
private:
std::optional<QString> advance() override
{
QString word;
do {
while (!m_stream || m_stream->atEnd()) {
if (m_currentPage >= m_doc.pageCount())
return std::nullopt;
m_pageText = m_doc.getAllText(m_currentPage++).text();
m_stream.emplace(&m_pageText);
}
*m_stream >> word;
} while (word.isEmpty());
return word;
}
QPdfDocument m_doc;
int m_currentPage = 0;
QString m_pageText;
std::optional<QTextStream> m_stream;
};
class WordDocumentReader final : public DocumentReader {
public:
explicit WordDocumentReader(const DocumentInfo &info)
: DocumentReader(info)
, m_doc(info.file.canonicalFilePath().toStdString())
{
m_doc.open();
if (!m_doc.is_open())
throw std::runtime_error(fmt::format("Failed to open DOCX: {}", info.file.canonicalFilePath()));
m_paragraph = &m_doc.paragraphs();
m_run = &m_paragraph->runs();
// TODO(jared): metadata for Word documents?
postInit();
}
protected:
std::optional<QString> advance() override
{
// find non-space char
qsizetype wordStart = 0;
while (m_buffer.isEmpty() || m_buffer[wordStart].isSpace()) {
if (m_buffer.isEmpty() && !fillBuffer())
return std::nullopt;
if (m_buffer[wordStart].isSpace() && ++wordStart >= m_buffer.size()) {
m_buffer.clear();
wordStart = 0;
}
}
// find space char
qsizetype wordEnd = wordStart + 1;
while (wordEnd >= m_buffer.size() || !m_buffer[wordEnd].isSpace()) {
if (wordEnd >= m_buffer.size() && !fillBuffer())
break;
if (!m_buffer[wordEnd].isSpace())
++wordEnd;
}
if (wordStart == wordEnd)
return std::nullopt;
auto size = wordEnd - wordStart;
QString word = std::move(m_buffer);
m_buffer = word.sliced(wordStart + size);
if (wordStart == 0)
word.resize(size);
else
word = word.sliced(wordStart, size);
return word;
}
bool fillBuffer()
{
for (;;) {
// get a run
while (!m_run->has_next()) {
// try next paragraph
if (!m_paragraph->has_next())
return false;
m_paragraph->next();
m_buffer += u'\n';
}
bool foundText = false;
auto &run = m_run->get_node();
for (auto node = run.first_child(); node; node = node.next_sibling()) {
std::string node_name = node.name();
if (node_name == "w:t") {
const char *text = node.text().get();
if (*text) {
foundText = true;
m_buffer += QUtf8StringView(text);
}
} else if (node_name == "w:br") {
m_buffer += u'\n';
} else if (node_name == "w:tab") {
m_buffer += u'\t';
}
}
m_run->next();
if (foundText) return true;
}
}
duckx::Document m_doc;
duckx::Paragraph *m_paragraph;
duckx::Run *m_run;
QString m_buffer;
};
class TxtDocumentReader final : public DocumentReader {
public:
explicit TxtDocumentReader(const DocumentInfo &info)
: DocumentReader(info)
, m_file(info.file.canonicalFilePath())
{
if (!m_file.open(QIODevice::ReadOnly))
throw std::runtime_error(fmt::format("Failed to open text file: {}", m_file.fileName()));
m_stream.setDevice(&m_file);
postInit();
}
protected:
std::optional<QString> advance() override
{
if (getError())
return std::nullopt;
while (!m_stream.atEnd()) {
QString word;
m_stream >> word;
if (getError())
return std::nullopt;
if (!word.isEmpty())
return word;
}
return std::nullopt;
}
std::optional<ChunkStreamer::Status> getError() const override
{
if (m_file.binarySeen())
return ChunkStreamer::Status::BINARY_SEEN;
if (m_file.error())
return ChunkStreamer::Status::ERROR;
return std::nullopt;
}
BinaryDetectingFile m_file;
QTextStream m_stream;
};
} // namespace
std::unique_ptr<DocumentReader> DocumentReader::fromDocument(const DocumentInfo &doc)
{
if (doc.isPdf())
return std::make_unique<PdfDocumentReader>(doc);
if (doc.isDocx())
return std::make_unique<WordDocumentReader>(doc);
return std::make_unique<TxtDocumentReader>(doc);
}
ChunkStreamer::ChunkStreamer(Database *database)
: m_database(database) {}
ChunkStreamer::~ChunkStreamer() = default;
void ChunkStreamer::setDocument(const DocumentInfo &doc, int documentId, const QString &embeddingModel)
{
auto docKey = doc.key();
if (!m_docKey || *m_docKey != docKey) {
m_docKey = docKey;
m_reader = DocumentReader::fromDocument(doc);
m_documentId = documentId;
m_embeddingModel = embeddingModel;
m_chunk.clear();
m_page = 0;
// make sure the document doesn't already have any chunks
if (m_database->m_documentIdCache.contains(documentId)) {
QSqlQuery q(m_database->m_db);
if (!m_database->removeChunksByDocumentId(q, documentId))
handleDocumentError("ERROR: Cannot remove chunks of document", documentId, doc.file.canonicalPath(), q.lastError());
}
}
}
std::optional<DocumentInfo::key_type> ChunkStreamer::currentDocKey() const
{
return m_docKey;
}
void ChunkStreamer::reset()
{
m_docKey.reset();
}
ChunkStreamer::Status ChunkStreamer::step()
{
// TODO: implement line_from/line_to
constexpr int line_from = -1;
constexpr int line_to = -1;
const int folderId = m_reader->doc().folder;
const int maxChunkSize = m_database->m_chunkSize;
int nChunks = 0;
int nAddedWords = 0;
Status retval;
for (;;) {
if (auto error = m_reader->getError()) {
m_docKey.reset(); // done processing
retval = *error;
break;
}
// get a word, if needed
std::optional<QString> word = QString(); // empty string to disable EOF logic
if (m_chunk.length() < maxChunkSize + 1) {
word = m_reader->word();
if (m_chunk.isEmpty())
m_page = m_reader->page(); // page number of first word
if (word) {
m_chunk += *word;
m_chunk += u' ';
m_reader->nextWord();
m_nChunkWords++;
}
}
if (!word || m_chunk.length() >= maxChunkSize + 1) { // +1 for trailing space
if (!m_chunk.isEmpty()) {
int nThisChunkWords = 0;
auto chunk = m_chunk; // copy
// handle overlength chunks
if (m_chunk.length() > maxChunkSize + 1) {
// find the final space
qsizetype chunkEnd = chunk.lastIndexOf(u' ', -2);
qsizetype spaceSize;
if (chunkEnd >= 0) {
// slice off the last word
spaceSize = 1;
Q_ASSERT(m_nChunkWords >= 1);
// one word left
nThisChunkWords = m_nChunkWords - 1;
m_nChunkWords = 1;
} else {
// slice the overlong word
spaceSize = 0;
chunkEnd = maxChunkSize;
// partial word left, don't count it
nThisChunkWords = m_nChunkWords;
m_nChunkWords = 0;
}
// save the second part, excluding space if any
m_chunk = chunk.sliced(chunkEnd + spaceSize);
// consume the first part
chunk.truncate(chunkEnd);
} else {
nThisChunkWords = m_nChunkWords;
m_nChunkWords = 0;
// there is no second part
m_chunk.clear();
// consume the whole chunk, excluding space
chunk.chop(1);
}
Q_ASSERT(chunk.length() <= maxChunkSize);
QSqlQuery q(m_database->m_db);
int chunkId = 0;
auto &metadata = m_reader->metadata();
if (!m_database->addChunk(q,
m_documentId,
chunk,
m_reader->doc().file.fileName(), // basename
metadata.title,
metadata.author,
metadata.subject,
metadata.keywords,
m_page,
line_from,
line_to,
nThisChunkWords,
&chunkId
)) {
qWarning() << "ERROR: Could not insert chunk into db" << q.lastError();
}
nAddedWords += nThisChunkWords;
EmbeddingChunk toEmbed;
toEmbed.model = m_embeddingModel;
toEmbed.folder_id = folderId;
toEmbed.chunk_id = chunkId;
toEmbed.chunk = chunk;
m_database->appendChunk(toEmbed);
++nChunks;
}
if (!word) {
retval = Status::DOC_COMPLETE;
m_docKey.reset(); // done processing
break;
}
}
if (m_database->scanQueueInterrupted()) {
retval = Status::INTERRUPTED;
break;
}
}
if (nChunks) {
CollectionItem item = m_database->guiCollectionItem(folderId);
// Set the start update if we haven't done so already
if (item.startUpdate <= item.lastUpdate && item.currentEmbeddingsToIndex == 0)
m_database->setStartUpdateTime(item);
item.currentEmbeddingsToIndex += nChunks;
item.totalEmbeddingsToIndex += nChunks;
item.totalWords += nAddedWords;
m_database->updateGuiForCollectionItem(item);
}
return retval;
}
void Database::appendChunk(const EmbeddingChunk &chunk)
{
m_chunkList.reserve(s_batchSize);
m_chunkList.append(chunk);
if (m_chunkList.size() >= s_batchSize)
sendChunkList();
}
void Database::sendChunkList()
{
m_embLLM->generateDocEmbeddingsAsync(m_chunkList);
m_chunkList.clear();
}
void Database::handleEmbeddingsGenerated(const QVector<EmbeddingResult> &embeddings)
{
Q_ASSERT(!embeddings.isEmpty());
QList<Embedding> sqlEmbeddings;
for (const auto &e: embeddings) {
auto data = QByteArray::fromRawData(
reinterpret_cast<const char *>(e.embedding.data()),
e.embedding.size() * sizeof(e.embedding.front())
);
sqlEmbeddings.append({e.model, e.folder_id, e.chunk_id, std::move(data)});
}
transaction();
QSqlQuery q(m_db);
QHash<EmbeddingFolder, EmbeddingStat> stats;
if (!sqlAddEmbeddings(q, sqlEmbeddings, stats)) {
qWarning() << "Database ERROR: failed to add embeddings:" << q.lastError();
return rollback();
}
commit();
// FIXME(jared): embedding counts are per-collectionitem, not per-folder
for (const auto &[key, stat]: std::as_const(stats).asKeyValueRange()) {
if (!m_collectionMap.contains(key.folder_id)) continue;
CollectionItem item = guiCollectionItem(key.folder_id);
Q_ASSERT(item.currentEmbeddingsToIndex >= stat.nAdded + stat.nSkipped);
if (item.currentEmbeddingsToIndex < stat.nAdded + stat.nSkipped) {
qWarning() << "Database ERROR: underflow in current embeddings to index statistics";
item.currentEmbeddingsToIndex = 0;
} else {
item.currentEmbeddingsToIndex -= stat.nAdded + stat.nSkipped;
}
Q_ASSERT(item.totalEmbeddingsToIndex >= stat.nSkipped);
if (item.totalEmbeddingsToIndex < stat.nSkipped) {
qWarning() << "Database ERROR: underflow in total embeddings to index statistics";
item.totalEmbeddingsToIndex = 0;
} else {
item.totalEmbeddingsToIndex -= stat.nSkipped;
}
if (!stat.lastFile.isNull())
item.fileCurrentlyProcessing = stat.lastFile;
// Set the last update if we are done
Q_ASSERT(item.startUpdate > item.lastUpdate);
if (!item.indexing && item.currentEmbeddingsToIndex == 0)
setLastUpdateTime(item);
updateGuiForCollectionItem(item);
}
}
void Database::handleErrorGenerated(const QVector<EmbeddingChunk> &chunks, const QString &error)
{
/* FIXME(jared): errors are actually collection-specific because they are conditioned
* on the embedding model, but this sets the error on all collections for a given
* folder */
QSet<int> folder_ids;
for (const auto &c: chunks) { folder_ids << c.folder_id; }
for (int fid: folder_ids) {
if (!m_collectionMap.contains(fid)) continue;
CollectionItem item = guiCollectionItem(fid);
item.error = error;
updateGuiForCollectionItem(item);
}
}
size_t Database::countOfDocuments(int folder_id) const
{
if (auto it = m_docsToScan.find(folder_id); it != m_docsToScan.end())
return it->second.size();
return 0;
}
size_t Database::countOfBytes(int folder_id) const
{
if (auto it = m_docsToScan.find(folder_id); it != m_docsToScan.end()) {
size_t totalBytes = 0;
for (const DocumentInfo &f : it->second)
totalBytes += f.file.size();
return totalBytes;
}
return 0;
}
DocumentInfo Database::dequeueDocument()
{
Q_ASSERT(!m_docsToScan.empty());
auto firstEntry = m_docsToScan.begin();
auto &[firstKey, queue] = *firstEntry;
Q_ASSERT(!queue.empty());
DocumentInfo result = std::move(queue.front());
queue.pop_front();
if (queue.empty())
m_docsToScan.erase(firstEntry);
return result;
}
void Database::removeFolderFromDocumentQueue(int folder_id)
{
if (auto queueIt = m_docsToScan.find(folder_id); queueIt != m_docsToScan.end()) {
if (auto key = m_chunkStreamer.currentDocKey()) {
if (ranges::any_of(queueIt->second, [&key](auto &d) { return d.key() == key; }))
m_chunkStreamer.reset(); // done with this document
}
// remove folder from queue
m_docsToScan.erase(queueIt);
}
}
void Database::enqueueDocumentInternal(DocumentInfo &&info, bool prepend)
{
auto &queue = m_docsToScan[info.folder];
queue.insert(prepend ? queue.begin() : queue.end(), std::move(info));
}
void Database::enqueueDocuments(int folder_id, std::list<DocumentInfo> &&infos)
{
// enqueue all documents
auto &queue = m_docsToScan[folder_id];
queue.splice(queue.end(), std::move(infos));
CollectionItem item = guiCollectionItem(folder_id);
item.currentDocsToIndex = queue.size();
item.totalDocsToIndex = queue.size();
const size_t bytes = countOfBytes(folder_id);
item.currentBytesToIndex = bytes;
item.totalBytesToIndex = bytes;
updateGuiForCollectionItem(item);
m_scanIntervalTimer->start();
}
bool Database::scanQueueInterrupted() const
{
return m_scanDurationTimer.elapsed() >= 100;
}
void Database::scanQueueBatch()
{
transaction();
m_scanDurationTimer.start();
// scan for up to the maximum scan duration or until we run out of documents
while (!m_docsToScan.empty()) {
scanQueue();
if (scanQueueInterrupted())
break;
}
commit();
if (m_docsToScan.empty())
m_scanIntervalTimer->stop();
}
void Database::scanQueue()
{
DocumentInfo info = dequeueDocument();
const size_t countForFolder = countOfDocuments(info.folder);
const int folder_id = info.folder;
// Update info
info.file.stat();
// If the doc has since been deleted or no longer readable, then we schedule more work and return
// leaving the cleanup for the cleanup handler
if (!info.file.exists() || !info.file.isReadable())
return updateFolderToIndex(folder_id, countForFolder);
const qint64 document_time = info.file.fileTime(QFile::FileModificationTime).toMSecsSinceEpoch();
const QString document_path = info.file.canonicalFilePath();
const bool currentlyProcessing = info.currentlyProcessing;
// Check and see if we already have this document
QSqlQuery q(m_db);
int existing_id = -1;
qint64 existing_time = -1;
if (!selectDocument(q, document_path, &existing_id, &existing_time)) {
handleDocumentError("ERROR: Cannot select document",
existing_id, document_path, q.lastError());
return updateFolderToIndex(folder_id, countForFolder);
}
// If we have the document, we need to compare the last modification time and if it is newer
// we must rescan the document, otherwise return
if (existing_id != -1 && !currentlyProcessing) {
Q_ASSERT(existing_time != -1);
if (document_time == existing_time) {
// No need to rescan, but we do have to schedule next
return updateFolderToIndex(folder_id, countForFolder);
}
if (!removeChunksByDocumentId(q, existing_id)) {
handleDocumentError("ERROR: Cannot remove chunks of document",
existing_id, document_path, q.lastError());
return updateFolderToIndex(folder_id, countForFolder);
}
updateCollectionStatistics();
}
// Update the document_time for an existing document, or add it for the first time now
int document_id = existing_id;
if (!currentlyProcessing) {
if (document_id != -1) {
if (!updateDocument(q, document_id, document_time)) {
handleDocumentError("ERROR: Could not update document_time",
document_id, document_path, q.lastError());
return updateFolderToIndex(folder_id, countForFolder);
}
} else {
if (!addDocument(q, folder_id, document_time, document_path, &document_id)) {
handleDocumentError("ERROR: Could not add document",
document_id, document_path, q.lastError());
return updateFolderToIndex(folder_id, countForFolder);
}
CollectionItem item = guiCollectionItem(folder_id);
item.totalDocs += 1;
updateGuiForCollectionItem(item);
}
}
// Get the embedding model for this folder
// FIXME(jared): there can be more than one since we allow one folder to be in multiple collections
QString embedding_model;
if (!sqlGetFolderEmbeddingModel(q, folder_id, embedding_model)) {
handleDocumentError("ERROR: Could not get embedding model",
document_id, document_path, q.lastError());
return updateFolderToIndex(folder_id, countForFolder);
}
Q_ASSERT(document_id != -1);
{
try {
m_chunkStreamer.setDocument(info, document_id, embedding_model);
} catch (const std::runtime_error &e) {
qWarning() << "LocalDocs ERROR:" << e.what();
goto dequeue;
}
}
switch (m_chunkStreamer.step()) {
case ChunkStreamer::Status::INTERRUPTED:
info.currentlyProcessing = true;
enqueueDocumentInternal(std::move(info), /*prepend*/ true);
return updateFolderToIndex(folder_id, countForFolder + 1);
case ChunkStreamer::Status::BINARY_SEEN:
/* When we see a binary file, we treat it like an empty file so we know not to
* scan it again. All existing chunks are removed, and in-progress embeddings
* are ignored when they complete. */
qInfo() << "LocalDocs: Ignoring file with binary data:" << document_path;
// this will also ensure in-flight embeddings are ignored
if (!removeChunksByDocumentId(q, existing_id))
handleDocumentError("ERROR: Cannot remove chunks of document", existing_id, document_path, q.lastError());
updateCollectionStatistics();
break;
case ChunkStreamer::Status::ERROR:
qWarning() << "error reading" << document_path;
break;
case ChunkStreamer::Status::DOC_COMPLETE:
;
}
dequeue:
auto item = guiCollectionItem(folder_id);
Q_ASSERT(item.currentBytesToIndex >= info.file.size());
if (item.currentBytesToIndex < info.file.size()) {
qWarning() << "Database ERROR: underflow in current bytes to index statistics";
item.currentBytesToIndex = 0;
} else {
item.currentBytesToIndex -= info.file.size();
}
updateGuiForCollectionItem(item);
return updateFolderToIndex(folder_id, countForFolder);
}
void Database::scanDocuments(int folder_id, const QString &folder_path)
{
#if defined(DEBUG)
qDebug() << "scanning folder for documents" << folder_path;
#endif
QDirIterator it(folder_path, QDir::Readable | QDir::Files | QDir::Dirs | QDir::NoDotAndDotDot,
QDirIterator::Subdirectories);
std::list<DocumentInfo> infos;
while (it.hasNext()) {
it.next();
QFileInfo fileInfo = it.fileInfo();
if (fileInfo.isDir()) {
addFolderToWatch(fileInfo.canonicalFilePath());
continue;
}
if (!m_scannedFileExtensions.contains(fileInfo.suffix(), Qt::CaseInsensitive))
continue;
infos.push_back({ folder_id, fileInfo });
}
if (!infos.empty()) {
CollectionItem item = guiCollectionItem(folder_id);
item.indexing = true;
updateGuiForCollectionItem(item);
enqueueDocuments(folder_id, std::move(infos));
} else {
updateFolderToIndex(folder_id, 0, false);
}
}
void Database::start()
{
connect(m_watcher, &QFileSystemWatcher::directoryChanged, this, &Database::directoryChanged);
connect(m_embLLM, &EmbeddingLLM::embeddingsGenerated, this, &Database::handleEmbeddingsGenerated);
connect(m_embLLM, &EmbeddingLLM::errorGenerated, this, &Database::handleErrorGenerated);
m_scanIntervalTimer->callOnTimeout(this, &Database::scanQueueBatch);
const QString modelPath = MySettings::globalInstance()->modelPath();
QList<CollectionItem> oldCollections;
if (!openLatestDb(modelPath, oldCollections)) {
m_databaseValid = false;
} else if (!initDb(modelPath, oldCollections)) {
m_databaseValid = false;
} else {
cleanDB();
ftsIntegrityCheck();
QSqlQuery q(m_db);
if (!refreshDocumentIdCache(q)) {
m_databaseValid = false;
} else {
addCurrentFolders();
}
}
if (!m_databaseValid)
emit databaseValidChanged();
}
void Database::addCurrentFolders()
{
#if defined(DEBUG)
qDebug() << "addCurrentFolders";
#endif
QSqlQuery q(m_db);
QList<CollectionItem> collections;
if (!selectAllFromCollections(q, &collections)) {
qWarning() << "ERROR: Cannot select collections" << q.lastError();
return;
}
guiCollectionListUpdated(collections);
scheduleUncompletedEmbeddings();
for (const auto &i : collections) {
if (!i.forceIndexing) {
addFolderToWatch(i.folder_path);
scanDocuments(i.folder_id, i.folder_path);
}
}
updateCollectionStatistics();
}
void Database::scheduleUncompletedEmbeddings()
{
QHash<IncompleteChunk, QStringList> chunkList;
QSqlQuery q(m_db);
if (!selectAllUncompletedChunks(q, chunkList)) {
qWarning() << "ERROR: Cannot select uncompleted chunks" << q.lastError();
return;
}
if (chunkList.isEmpty())
return;
// map of folder_id -> chunk count
QMap<int, int> folderNChunks;
for (auto it = chunkList.keyBegin(), end = chunkList.keyEnd(); it != end; ++it) {
int folder_id = it->folder_id;
if (folderNChunks.contains(folder_id)) continue;
int total = 0;
if (!selectCountChunks(q, folder_id, total)) {
qWarning() << "ERROR: Cannot count total chunks" << q.lastError();
return;
}
folderNChunks.insert(folder_id, total);
}
// map of (folder_id, collection) -> incomplete count
QMap<QPair<int, QString>, int> itemNIncomplete;
for (const auto &[chunk, collections]: std::as_const(chunkList).asKeyValueRange())
for (const auto &collection: std::as_const(collections))
itemNIncomplete[{ chunk.folder_id, collection }]++;
for (const auto &[key, nIncomplete]: std::as_const(itemNIncomplete).asKeyValueRange()) {
const auto &[folder_id, collection] = key;
/* FIXME(jared): this needs to be split by collection because different
* collections have different embedding models */
int total = folderNChunks.value(folder_id);
CollectionItem item = guiCollectionItem(folder_id);
item.totalEmbeddingsToIndex = total;
item.currentEmbeddingsToIndex = nIncomplete;
updateGuiForCollectionItem(item);
}
for (auto it = chunkList.keyBegin(), end = chunkList.keyEnd(); it != end;) {
QList<EmbeddingChunk> batch;
for (; it != end && batch.size() < s_batchSize; ++it)
batch.append({ /*model*/ it->embedding_model, /*folder_id*/ it->folder_id, /*chunk_id*/ it->chunk_id, /*chunk*/ it->text });
Q_ASSERT(!batch.isEmpty());
m_embLLM->generateDocEmbeddingsAsync(batch);
}
}
void Database::updateCollectionStatistics()
{
QSqlQuery q(m_db);
QList<CollectionItem> collections;
if (!selectAllFromCollections(q, &collections)) {
qWarning() << "ERROR: Cannot select collections" << q.lastError();
return;
}
for (const auto &i: std::as_const(collections)) {
int total_docs = 0;
int total_words = 0;
int total_tokens = 0;
if (!selectCountStatistics(q, i.folder_id, &total_docs, &total_words, &total_tokens)) {
qWarning() << "ERROR: could not count statistics for folder" << q.lastError();
} else {
CollectionItem item = guiCollectionItem(i.folder_id);
item.totalDocs = total_docs;
item.totalWords = total_words;
item.totalTokens = total_tokens;
updateGuiForCollectionItem(item);
}
}
}
int Database::checkAndAddFolderToDB(const QString &path)
{
QFileInfo info(path);
if (!info.exists() || !info.isReadable()) {
qWarning() << "ERROR: Cannot add folder that doesn't exist or not readable" << path;
return -1;
}
QSqlQuery q(m_db);
int folder_id = -1;
// See if the folder exists in the db
if (!selectFolder(q, path, &folder_id)) {
qWarning() << "ERROR: Cannot select folder from path" << path << q.lastError();
return -1;
}
// Add the folder
if (folder_id == -1 && !addFolderToDB(q, path, &folder_id)) {
qWarning() << "ERROR: Cannot add folder to db with path" << path << q.lastError();
return -1;
}
Q_ASSERT(folder_id != -1);
return folder_id;
}
void Database::forceIndexing(const QString &collection, const QString &embedding_model)
{
Q_ASSERT(!embedding_model.isNull());
QSqlQuery q(m_db);
QList<QPair<int, QString>> folders;
if (!selectFoldersFromCollection(q, collection, &folders)) {
qWarning() << "ERROR: Cannot select folders from collections" << collection << q.lastError();
return;
}
if (!setCollectionEmbeddingModel(q, collection, embedding_model)) {
qWarning().nospace() << "ERROR: Cannot set embedding model for collection " << collection << ": "
<< q.lastError();
return;
}
for (const auto &folder: std::as_const(folders)) {
CollectionItem item = guiCollectionItem(folder.first);
item.embeddingModel = embedding_model;
item.forceIndexing = false;
updateGuiForCollectionItem(item);
addFolderToWatch(folder.second);
scanDocuments(folder.first, folder.second);
}
}
void Database::forceRebuildFolder(const QString &path)
{
QSqlQuery q(m_db);
int folder_id;
if (!selectFolder(q, path, &folder_id)) {
qWarning().nospace() << "Database ERROR: Cannot select folder from path " << path << ": " << q.lastError();
return;
}
Q_ASSERT(!m_docsToScan.contains(folder_id));
transaction();
if (!sqlRemoveDocsByFolderPath(q, path)) {
qWarning().nospace() << "Database ERROR: Cannot remove chunks for folder " << path << ": " << q.lastError();
return rollback();
}
commit();
updateCollectionStatistics();
// We now have zero embeddings. Document progress will be updated by scanDocuments.
// FIXME(jared): this updates the folder, but these values should also depend on the collection
CollectionItem item = guiCollectionItem(folder_id);
item.currentEmbeddingsToIndex = item.totalEmbeddingsToIndex = 0;
updateGuiForCollectionItem(item);
scanDocuments(folder_id, path);
}
bool Database::addFolder(const QString &collection, const QString &path, const QString &embedding_model)
{
// add the folder, if needed
const int folder_id = checkAndAddFolderToDB(path);
if (folder_id == -1)
return false;
std::optional<CollectionItem> item;
QSqlQuery q(m_db);
if (!selectCollectionByName(q, collection, item)) {
qWarning().nospace() << "Database ERROR: Cannot select collection " << collection << ": " << q.lastError();
return false;
}
// add the collection, if needed
if (!item) {
item.emplace();
if (!addCollection(q, collection, QDateTime() /*start_update*/, QDateTime() /*last_update*/,
embedding_model /*embedding_model*/, *item)) {
qWarning().nospace() << "ERROR: Cannot add collection " << collection << ": " << q.lastError();
return false;
}
}
// link the folder and the collection, if needed
int res = addCollectionItem(q, item->collection_id, folder_id);
if (res < 0) { // error
qWarning().nospace() << "Database ERROR: Cannot add folder " << path << " to collection " << collection << ": "
<< q.lastError();
return false;
}
// add the new collection item to the UI
if (res == 1) { // new item added
item->folder_path = path;
item->folder_id = folder_id;
addGuiCollectionItem(item.value());
// note: this is the existing embedding model if the collection was found
if (!item->embeddingModel.isNull()) {
addFolderToWatch(path);
scanDocuments(folder_id, path);
}
}
return true;
}
void Database::removeFolder(const QString &collection, const QString &path)
{
#if defined(DEBUG)
qDebug() << "removeFolder" << path;
#endif
QSqlQuery q(m_db);
int folder_id = -1;
// See if the folder exists in the db
if (!selectFolder(q, path, &folder_id)) {
qWarning() << "ERROR: Cannot select folder from path" << path << q.lastError();
return;
}
// If we don't have a folder_id in the db, then something bad has happened
Q_ASSERT(folder_id != -1);
if (folder_id == -1) {
qWarning() << "ERROR: Collected folder does not exist in db" << path;
m_watcher->removePath(path);
return;
}
transaction();
if (removeFolderInternal(collection, folder_id, path)) {
commit();
} else {
rollback();
}
}
bool Database::removeFolderInternal(const QString &collection, int folder_id, const QString &path)
{
// Remove it from the collection
QSqlQuery q(m_db);
int nRemaining = removeCollectionFolder(q, collection, folder_id);
if (nRemaining == -1) {
qWarning().nospace() << "Database ERROR: Cannot remove collection " << collection << " from folder "
<< folder_id << ": " << q.lastError();
return false;
}
removeGuiFolderById(collection, folder_id);
if (!sqlPruneCollections(q)) {
qWarning() << "Database ERROR: Cannot prune collections:" << q.lastError();
return false;
}
// Keep folder if it is still referenced
if (nRemaining)
return true;
// Remove the last reference to a folder
// First remove all upcoming jobs associated with this folder
removeFolderFromDocumentQueue(folder_id);
// Get a list of all documents associated with folder
QList<int> documentIds;
if (!selectDocuments(q, folder_id, &documentIds)) {
qWarning() << "ERROR: Cannot select documents" << folder_id << q.lastError();
return false;
}
// Remove all chunks and documents associated with this folder
for (int document_id: std::as_const(documentIds)) {
if (!removeChunksByDocumentId(q, document_id)) {
qWarning() << "ERROR: Cannot remove chunks of document_id" << document_id << q.lastError();
return false;
}
if (!removeDocument(q, document_id)) {
qWarning() << "ERROR: Cannot remove document_id" << document_id << q.lastError();
return false;
}
}
if (!removeFolderFromDB(q, folder_id)) {
qWarning() << "ERROR: Cannot remove folder_id" << folder_id << q.lastError();
return false;
}
m_collectionMap.remove(folder_id);
removeFolderFromWatch(path);
return true;
}
void Database::addFolderToWatch(const QString &path)
{
#if defined(DEBUG)
qDebug() << "addFolderToWatch" << path;
#endif
// pre-check because addPath returns false for already watched paths
if (!m_watchedPaths.contains(path)) {
if (!m_watcher->addPath(path))
qWarning() << "Database::addFolderToWatch: failed to watch" << path;
// add unconditionally to suppress repeated warnings
m_watchedPaths << path;
}
}
void Database::removeFolderFromWatch(const QString &path)
{
#if defined(DEBUG)
qDebug() << "removeFolderFromWatch" << path;
#endif
QDirIterator it(path, QDir::Readable | QDir::Dirs | QDir::NoDotAndDotDot, QDirIterator::Subdirectories);
QStringList children { path };
while (it.hasNext())
children.append(it.next());
m_watcher->removePaths(children);
m_watchedPaths -= QSet(children.begin(), children.end());
}
QList<int> Database::searchEmbeddingsHelper(const std::vector<float> &query, QSqlQuery &q, int nNeighbors)
{
constexpr int BATCH_SIZE = 2048;
const int n_embd = query.size();
const us::metric_punned_t metric(n_embd, us::metric_kind_t::ip_k); // inner product
us::executor_default_t executor(std::thread::hardware_concurrency());
us::exact_search_t search;
QList<int> batchChunkIds;
QList<float> batchEmbeddings;
batchChunkIds.reserve(BATCH_SIZE);
batchEmbeddings.reserve(BATCH_SIZE * n_embd);
struct Result { int chunkId; us::distance_punned_t dist; };
QList<Result> results;
// The q parameter is expected to be the result of a QSqlQuery returning (chunk_id, embedding) pairs
while (q.at() != QSql::AfterLastRow) { // batches
batchChunkIds.clear();
batchEmbeddings.clear();
while (batchChunkIds.count() < BATCH_SIZE && q.next()) { // batch
batchChunkIds << q.value(0).toInt();
batchEmbeddings.resize(batchEmbeddings.size() + n_embd);
QVariant embdCol = q.value(1);
if (embdCol.userType() != QMetaType::QByteArray) {
qWarning() << "Database ERROR: Expected embedding to be blob, got" << embdCol.userType();
return {};
}
auto *embd = static_cast<const QByteArray *>(embdCol.constData());
const int embd_stride = n_embd * sizeof(float);
if (embd->size() != embd_stride) {
qWarning() << "Database ERROR: Expected embedding to be" << embd_stride << "bytes, got"
<< embd->size();
return {};
}
memcpy(&*(batchEmbeddings.end() - n_embd), embd->constData(), embd_stride);
}
int nBatch = batchChunkIds.count();
if (!nBatch)
break;
// get top-k nearest neighbors of this batch
int kBatch = qMin(nNeighbors, nBatch);
us::exact_search_results_t batchResults = search(
(us::byte_t const *)batchEmbeddings.data(), nBatch, n_embd * sizeof(float),
(us::byte_t const *)query.data(), 1, n_embd * sizeof(float),
kBatch, metric
);
for (int i = 0; i < kBatch; ++i) {
auto offset = batchResults.at(0)[i].offset;
us::distance_punned_t distance = batchResults.at(0)[i].distance;
results.append({batchChunkIds[offset], distance});
}
}
// get top-k nearest neighbors of combined results
nNeighbors = qMin(nNeighbors, results.size());
std::partial_sort(
results.begin(), results.begin() + nNeighbors, results.end(),
[](const Result &a, const Result &b) { return a.dist < b.dist; }
);
QList<int> chunkIds;
chunkIds.reserve(nNeighbors);
for (int i = 0; i < nNeighbors; i++)
chunkIds << results[i].chunkId;
return chunkIds;
}
QList<int> Database::searchEmbeddings(const std::vector<float> &query, const QList<QString> &collections,
int nNeighbors)
{
QSqlQuery q(m_db);
if (!q.exec(GET_COLLECTION_EMBEDDINGS_SQL.arg(collections.join("', '")))) {
qWarning() << "Database ERROR: Failed to exec embeddings query:" << q.lastError();
return {};
}
return searchEmbeddingsHelper(query, q, nNeighbors);
}
QList<int> Database::scoreChunks(const std::vector<float> &query, const QList<int> &chunks)
{
QList<QString> chunkStrings;
for (int id : chunks)
chunkStrings << QString::number(id);
QSqlQuery q(m_db);
if (!q.exec(GET_CHUNK_EMBEDDINGS_SQL.arg(chunkStrings.join(", ")))) {
qWarning() << "Database ERROR: Failed to exec embeddings query:" << q.lastError();
return {};
}
return searchEmbeddingsHelper(query, q, chunks.size());
}
QList<Database::BM25Query> Database::queriesForFTS5(const QString &input)
{
// Escape double quotes by adding a second double quote
QString escapedInput = input;
escapedInput.replace("\"", "\"\"");
static QRegularExpression spaces("\\s+");
QStringList oWords = escapedInput.split(spaces, Qt::SkipEmptyParts);
QList<BM25Query> queries;
// Start by trying to match the entire input
BM25Query e;
e.isExact = true;
e.input = oWords.join(" ");
e.query = "\"" + oWords.join(" ") + "\"";
e.qlength = oWords.size();
e.ilength = oWords.size();
queries << e;
// https://github.com/igorbrigadir/stopwords?tab=readme-ov-file
// Lucene, Solr, Elastisearch
static const QSet<QString> stopWords = {
"a", "an", "and", "are", "as", "at", "be", "but", "by",
"for", "if", "in", "into", "is", "it", "no", "not", "of",
"on", "or", "such", "that", "the", "their", "then", "there",
"these", "they", "this", "to", "was", "will", "with"
};
QStringList quotedWords;
for (const QString &w : oWords)
if (!stopWords.contains(w.toLower()))
quotedWords << "\"" + w + "\"";
BM25Query b;
b.input = oWords.join(" ");
b.query = "(" + quotedWords.join(" OR ") + ")";
b.qlength = 1; // length of phrase
b.ilength = oWords.size();
b.rlength = oWords.size() - quotedWords.size();
queries << b;
return queries;
}
QList<int> Database::searchBM25(const QString &query, const QList<QString> &collections, BM25Query &bm25q, int k)
{
struct SearchResult { int chunkId; float score; };
QList<BM25Query> bm25Queries = queriesForFTS5(query);
QSqlQuery sqlQuery(m_db);
sqlQuery.prepare(SELECT_CHUNKS_FTS_SQL.arg(collections.join("', '"), QString::number(k)));
QList<SearchResult> results;
for (auto &bm25Query : std::as_const(bm25Queries)) {
sqlQuery.addBindValue(bm25Query.query);
if (!sqlQuery.exec()) {
qWarning() << "Database ERROR: Failed to execute BM25 query:" << sqlQuery.lastError();
return {};
}
if (sqlQuery.next()) {
// Save the query that was used to produce results
bm25q = bm25Query;
break;
}
}
if (sqlQuery.at() != QSql::AfterLastRow) {
do {
const int chunkId = sqlQuery.value(0).toInt();
const float score = sqlQuery.value(1).toFloat();
results.append({chunkId, score});
} while (sqlQuery.next());
}
k = qMin(k, results.size());
std::partial_sort(
results.begin(), results.begin() + k, results.end(),
[](const SearchResult &a, const SearchResult &b) { return a.score < b.score; }
);
QList<int> chunkIds;
chunkIds.reserve(k);
for (int i = 0; i < k; i++)
chunkIds << results[i].chunkId;
return chunkIds;
}
float Database::computeBM25Weight(const Database::BM25Query &bm25q)
{
float bmWeight = 0.0f;
if (bm25q.isExact) {
bmWeight = 0.9f; // the highest we give
} else {
// qlength is the length of the phrases in the query by number of distinct words
// ilength is the length of the natural language query by number of distinct words
// rlength is the number of stop words removed from the natural language query to form the query
// calculate the query length weight based on the ratio of query terms to meaningful terms.
// this formula adjusts the weight with the empirically determined insight that BM25's
// effectiveness decreases as query length increases.
float queryLengthWeight = 1 / powf(float(bm25q.ilength - bm25q.rlength), 2);
queryLengthWeight = qBound(0.0f, queryLengthWeight, 1.0f);
// the weighting is bound between 1/4 and 3/4 which was determined empirically to work well
// with the beir nfcorpus, scifact, fiqa and trec-covid datasets along with our embedding
// model
bmWeight = 0.25f + queryLengthWeight * 0.50f;
}
#if 0
qDebug()
<< "bm25q.type" << bm25q.type
<< "bm25q.qlength" << bm25q.qlength
<< "bm25q.ilength" << bm25q.ilength
<< "bm25q.rlength" << bm25q.rlength
<< "bmWeight" << bmWeight;
#endif
return bmWeight;
}
QList<int> Database::reciprocalRankFusion(const std::vector<float> &query, const QList<int> &embeddingResults,
const QList<int> &bm25Results, const BM25Query &bm25q, int k)
{
// We default to the embedding results and augment with bm25 if any
QList<int> results = embeddingResults;
QList<int> missingScores;
QHash<int, int> bm25Ranks;
for (int i = 0; i < bm25Results.size(); ++i) {
if (!results.contains(bm25Results[i]))
missingScores.append(bm25Results[i]);
bm25Ranks[bm25Results[i]] = i + 1;
}
if (!missingScores.isEmpty()) {
QList<int> scored = scoreChunks(query, missingScores);
results << scored;
}
QHash<int, int> embeddingRanks;
for (int i = 0; i < results.size(); ++i)
embeddingRanks[results[i]] = i + 1;
const float bmWeight = bm25Results.isEmpty() ? 0 : computeBM25Weight(bm25q);
// From the paper: "Reciprocal Rank Fusion outperforms Condorcet and individual Rank Learning Methods"
// doi: 10.1145/1571941.157211
const int fusion_k = 60;
std::stable_sort(
results.begin(), results.end(),
[&](const int &a, const int &b) {
// Reciprocal Rank Fusion (RRF)
const int aBm25Rank = bm25Ranks.value(a, bm25Results.size() + 1);
const int aEmbeddingRank = embeddingRanks.value(a, embeddingResults.size() + 1);
Q_ASSERT(embeddingRanks.contains(a));
const int bBm25Rank = bm25Ranks.value(b, bm25Results.size() + 1);
const int bEmbeddingRank = embeddingRanks.value(b, embeddingResults.size() + 1);
Q_ASSERT(embeddingRanks.contains(b));
const float aBm25Score = 1.0f / (fusion_k + aBm25Rank);
const float bBm25Score = 1.0f / (fusion_k + bBm25Rank);
const float aEmbeddingScore = 1.0f / (fusion_k + aEmbeddingRank);
const float bEmbeddingScore = 1.0f / (fusion_k + bEmbeddingRank);
const float aWeightedScore = bmWeight * aBm25Score + (1.f - bmWeight) * aEmbeddingScore;
const float bWeightedScore = bmWeight * bBm25Score + (1.f - bmWeight) * bEmbeddingScore;
// Higher RRF score means better ranking, so we use greater than for sorting
return aWeightedScore > bWeightedScore;
}
);
k = qMin(k, results.size());
results.resize(k);
return results;
}
QList<int> Database::searchDatabase(const QString &query, const QList<QString> &collections, int k)
{
std::vector<float> queryEmbd = m_embLLM->generateQueryEmbedding(query);
if (queryEmbd.empty()) {
qDebug() << "ERROR: generating embeddings returned a null result";
return { };
}
const QList<int> embeddingResults = searchEmbeddings(queryEmbd, collections, k);
BM25Query bm25q;
const QList<int> bm25Results = searchBM25(query, collections, bm25q, k);
return reciprocalRankFusion(queryEmbd, embeddingResults, bm25Results, bm25q, k);
}
void Database::retrieveFromDB(const QList<QString> &collections, const QString &text, int retrievalSize,
QList<ResultInfo> *results)
{
#if defined(DEBUG)
qDebug() << "retrieveFromDB" << collections << text << retrievalSize;
#endif
QList<int> searchResults = searchDatabase(text, collections, retrievalSize);
if (searchResults.isEmpty())
return;
QSqlQuery q(m_db);
if (!selectChunk(q, searchResults)) {
qDebug() << "ERROR: selecting chunks:" << q.lastError();
return;
}
QHash<int, ResultInfo> tempResults;
while (q.next()) {
const int rowid = q.value(0).toInt();
const QString document_path = q.value(2).toString();
const QString chunk_text = q.value(3).toString();
const QString date = QDateTime::fromMSecsSinceEpoch(q.value(1).toLongLong()).toString("yyyy, MMMM dd");
const QString file = q.value(4).toString();
const QString title = q.value(5).toString();
const QString author = q.value(6).toString();
const int page = q.value(7).toInt();
const int from = q.value(8).toInt();
const int to = q.value(9).toInt();
const QString collectionName = q.value(10).toString();
ResultInfo info;
info.collection = collectionName;
info.path = document_path;
info.file = file;
info.title = title;
info.author = author;
info.date = date;
info.text = chunk_text;
info.page = page;
info.from = from;
info.to = to;
tempResults.insert(rowid, info);
#if defined(DEBUG)
qDebug() << "retrieve rowid:" << rowid
<< "chunk_text:" << chunk_text;
#endif
}
for (int id : searchResults)
if (tempResults.contains(id))
results->append(tempResults.value(id));
}
bool Database::ftsIntegrityCheck()
{
QSqlQuery q(m_db);
// Returns an error executing sql if it the integrity check fails
// See: https://www.sqlite.org/fts5.html#the_integrity_check_command
const bool success = q.exec(FTS_INTEGRITY_SQL);
if (!success && q.lastError().nativeErrorCode() != "267" /*SQLITE_CORRUPT_VTAB from sqlite header*/) {
qWarning() << "ERROR: Cannot prepare sql for fts integrity check" << q.lastError();
return false;
}
if (!success && !q.exec(FTS_REBUILD_SQL)) {
qWarning() << "ERROR: Cannot exec sql for fts rebuild" << q.lastError();
return false;
}
return true;
}
// FIXME This is very slow and non-interruptible and when we close the application and we're
// cleaning a large table this can cause the app to take forever to shut down. This would ideally be
// interruptible and we'd continue 'cleaning' when we restart
bool Database::cleanDB()
{
#if defined(DEBUG)
qDebug() << "cleanDB";
#endif
// Scan all folders in db to make sure they still exist
QSqlQuery q(m_db);
QList<CollectionItem> collections;
if (!selectAllFromCollections(q, &collections)) {
qWarning() << "ERROR: Cannot select collections" << q.lastError();
return false;
}
transaction();
for (const auto &i: std::as_const(collections)) {
// Find the path for the folder
QFileInfo info(i.folder_path);
if (!info.exists() || !info.isReadable()) {
#if defined(DEBUG)
qDebug() << "clean db removing folder" << i.folder_id << i.folder_path;
#endif
if (!removeFolderInternal(i.collection, i.folder_id, i.folder_path)) {
rollback();
return false;
}
}
}
// Scan all documents in db to make sure they still exist
if (!q.prepare(SELECT_ALL_DOCUMENTS_SQL)) {
qWarning() << "ERROR: Cannot prepare sql for select all documents" << q.lastError();
rollback();
return false;
}
if (!q.exec()) {
qWarning() << "ERROR: Cannot exec sql for select all documents" << q.lastError();
rollback();
return false;
}
while (q.next()) {
int document_id = q.value(0).toInt();
QString document_path = q.value(1).toString();
QFileInfo info(document_path);
if (info.exists() && info.isReadable() && m_scannedFileExtensions.contains(info.suffix(), Qt::CaseInsensitive))
continue;
#if defined(DEBUG)
qDebug() << "clean db removing document" << document_id << document_path;
#endif
// Remove all chunks and documents that either don't exist or have become unreadable
QSqlQuery query(m_db);
if (!removeChunksByDocumentId(query, document_id)) {
qWarning() << "ERROR: Cannot remove chunks of document_id" << document_id << query.lastError();
rollback();
return false;
}
if (!removeDocument(query, document_id)) {
qWarning() << "ERROR: Cannot remove document_id" << document_id << query.lastError();
rollback();
return false;
}
}
commit();
return true;
}
void Database::changeChunkSize(int chunkSize)
{
if (chunkSize == m_chunkSize)
return;
#if defined(DEBUG)
qDebug() << "changeChunkSize" << chunkSize;
#endif
QSqlQuery q(m_db);
// Scan all documents in db to make sure they still exist
if (!q.prepare(SELECT_ALL_DOCUMENTS_SQL)) {
qWarning() << "ERROR: Cannot prepare sql for select all documents" << q.lastError();
return;
}
if (!q.exec()) {
qWarning() << "ERROR: Cannot exec sql for select all documents" << q.lastError();
return;
}
transaction();
while (q.next()) {
int document_id = q.value(0).toInt();
// Remove all chunks and documents to change the chunk size
QSqlQuery query(m_db);
if (!removeChunksByDocumentId(query, document_id)) {
qWarning() << "ERROR: Cannot remove chunks of document_id" << document_id << query.lastError();
return rollback();
}
if (!removeDocument(query, document_id)) {
qWarning() << "ERROR: Cannot remove document_id" << document_id << query.lastError();
return rollback();
}
}
commit();
m_chunkSize = chunkSize;
addCurrentFolders();
updateCollectionStatistics();
}
void Database::changeFileExtensions(const QStringList &extensions)
{
#if defined(DEBUG)
qDebug() << "changeFileExtensions";
#endif
m_scannedFileExtensions = extensions;
if (cleanDB())
updateCollectionStatistics();
QSqlQuery q(m_db);
QList<CollectionItem> collections;
if (!selectAllFromCollections(q, &collections)) {
qWarning() << "ERROR: Cannot select collections" << q.lastError();
return;
}
for (const auto &i: std::as_const(collections)) {
if (!i.forceIndexing)
scanDocuments(i.folder_id, i.folder_path);
}
}
void Database::directoryChanged(const QString &path)
{
#if defined(DEBUG)
qDebug() << "directoryChanged" << path;
#endif
// search for a collection that contains this folder (we watch subdirectories)
int folder_id = -1;
QDir dir(path);
for (;;) {
QSqlQuery q(m_db);
if (!selectFolder(q, dir.path(), &folder_id)) {
qWarning() << "ERROR: Cannot select folder from path" << dir.path() << q.lastError();
return;
}
if (folder_id != -1)
break;
// check next parent
if (!dir.cdUp()) {
if (!dir.isRoot()) break; // folder removed
Q_ASSERT(false);
qWarning() << "ERROR: Watched folder does not exist in db" << path;
m_watcher->removePath(path);
return;
}
}
// Clean the database
if (cleanDB())
updateCollectionStatistics();
// Rescan the documents associated with the folder
if (folder_id != -1)
scanDocuments(folder_id, path);
}