memorisdk 1.0.2__py3-none-any.whl → 2.0.0__py3-none-any.whl

memori/database/search_service.py

@@ -0,0 +1,548 @@
+"""
+SQLAlchemy-based search service for Memori v2.0
+Provides cross-database full-text search capabilities
+"""
+
+from datetime import datetime
+from typing import Any, Dict, List, Optional
+
+from loguru import logger
+from sqlalchemy import and_, desc, or_, text
+from sqlalchemy.orm import Session
+
+from .models import LongTermMemory, ShortTermMemory
+
+
+class SearchService:
+    """Cross-database search service using SQLAlchemy"""
+
+    def __init__(self, session: Session, database_type: str):
+        self.session = session
+        self.database_type = database_type
+
+    def search_memories(
+        self,
+        query: str,
+        namespace: str = "default",
+        category_filter: Optional[List[str]] = None,
+        limit: int = 10,
+        memory_types: Optional[List[str]] = None,
+    ) -> List[Dict[str, Any]]:
+        """
+        Search memories across different database backends
+
+        Args:
+            query: Search query string
+            namespace: Memory namespace
+            category_filter: List of categories to filter by
+            limit: Maximum number of results
+            memory_types: Types of memory to search ('short_term', 'long_term', or both)
+
+        Returns:
+            List of memory dictionaries with search metadata
+        """
+        if not query or not query.strip():
+            return self._get_recent_memories(
+                namespace, category_filter, limit, memory_types
+            )
+
+        results = []
+
+        # Determine which memory types to search
+        search_short_term = not memory_types or "short_term" in memory_types
+        search_long_term = not memory_types or "long_term" in memory_types
+
+        try:
+            # Try database-specific full-text search first
+            if self.database_type == "sqlite":
+                results = self._search_sqlite_fts(
+                    query,
+                    namespace,
+                    category_filter,
+                    limit,
+                    search_short_term,
+                    search_long_term,
+                )
+            elif self.database_type == "mysql":
+                results = self._search_mysql_fulltext(
+                    query,
+                    namespace,
+                    category_filter,
+                    limit,
+                    search_short_term,
+                    search_long_term,
+                )
+            elif self.database_type == "postgresql":
+                results = self._search_postgresql_fts(
+                    query,
+                    namespace,
+                    category_filter,
+                    limit,
+                    search_short_term,
+                    search_long_term,
+                )
+
+            # If no results or full-text search failed, fall back to LIKE search
+            if not results:
+                results = self._search_like_fallback(
+                    query,
+                    namespace,
+                    category_filter,
+                    limit,
+                    search_short_term,
+                    search_long_term,
+                )
+
+        except Exception as e:
+            logger.warning(f"Full-text search failed: {e}, falling back to LIKE search")
+            results = self._search_like_fallback(
+                query,
+                namespace,
+                category_filter,
+                limit,
+                search_short_term,
+                search_long_term,
+            )
+
+        return self._rank_and_limit_results(results, limit)
+
+    def _search_sqlite_fts(
+        self,
+        query: str,
+        namespace: str,
+        category_filter: Optional[List[str]],
+        limit: int,
+        search_short_term: bool,
+        search_long_term: bool,
+    ) -> List[Dict[str, Any]]:
+        """Search using SQLite FTS5"""
+        try:
+            # Build FTS query
+            fts_query = f'"{query.strip()}"'
+
+            # Build category filter
+            category_clause = ""
+            params = {"fts_query": fts_query, "namespace": namespace}
+
+            if category_filter:
+                category_placeholders = ",".join(
+                    [f":cat_{i}" for i in range(len(category_filter))]
+                )
+                category_clause = (
+                    f"AND fts.category_primary IN ({category_placeholders})"
+                )
+                for i, cat in enumerate(category_filter):
+                    params[f"cat_{i}"] = cat
+
+            # SQLite FTS5 search query
+            sql_query = f"""
+                SELECT
+                    fts.memory_id, fts.memory_type, fts.category_primary,
+                    CASE
+                        WHEN fts.memory_type = 'short_term' THEN st.processed_data
+                        WHEN fts.memory_type = 'long_term' THEN lt.processed_data
+                    END as processed_data,
+                    CASE
+                        WHEN fts.memory_type = 'short_term' THEN st.importance_score
+                        WHEN fts.memory_type = 'long_term' THEN lt.importance_score
+                        ELSE 0.5
+                    END as importance_score,
+                    CASE
+                        WHEN fts.memory_type = 'short_term' THEN st.created_at
+                        WHEN fts.memory_type = 'long_term' THEN lt.created_at
+                    END as created_at,
+                    fts.summary,
+                    rank as search_score,
+                    'sqlite_fts5' as search_strategy
+                FROM memory_search_fts fts
+                LEFT JOIN short_term_memory st ON fts.memory_id = st.memory_id AND fts.memory_type = 'short_term'
+                LEFT JOIN long_term_memory lt ON fts.memory_id = lt.memory_id AND fts.memory_type = 'long_term'
+                WHERE memory_search_fts MATCH :fts_query AND fts.namespace = :namespace
+                {category_clause}
+                ORDER BY rank, importance_score DESC
+                LIMIT {limit}
+            """
+
+            result = self.session.execute(text(sql_query), params)
+            return [dict(row) for row in result]
+
+        except Exception as e:
+            logger.debug(f"SQLite FTS5 search failed: {e}")
+            # Roll back the transaction to recover from error state
+            self.session.rollback()
+            return []
+
+    def _search_mysql_fulltext(
+        self,
+        query: str,
+        namespace: str,
+        category_filter: Optional[List[str]],
+        limit: int,
+        search_short_term: bool,
+        search_long_term: bool,
+    ) -> List[Dict[str, Any]]:
+        """Search using MySQL FULLTEXT"""
+        results = []
+
+        try:
+            # Search short-term memory if requested
+            if search_short_term:
+                short_query = self.session.query(ShortTermMemory).filter(
+                    ShortTermMemory.namespace == namespace
+                )
+
+                # Add FULLTEXT search
+                fulltext_condition = text(
+                    "MATCH(searchable_content, summary) AGAINST(:query IN NATURAL LANGUAGE MODE)"
+                ).params(query=query)
+                short_query = short_query.filter(fulltext_condition)
+
+                # Add category filter
+                if category_filter:
+                    short_query = short_query.filter(
+                        ShortTermMemory.category_primary.in_(category_filter)
+                    )
+
+                # Add relevance score
+                short_results = self.session.execute(
+                    short_query.statement.add_columns(
+                        text(
+                            "MATCH(searchable_content, summary) AGAINST(:query IN NATURAL LANGUAGE MODE) as search_score"
+                        ).params(query=query),
+                        text("'short_term' as memory_type"),
+                        text("'mysql_fulltext' as search_strategy"),
+                    )
+                ).fetchall()
+
+                results.extend([dict(row) for row in short_results])
+
+            # Search long-term memory if requested
+            if search_long_term:
+                long_query = self.session.query(LongTermMemory).filter(
+                    LongTermMemory.namespace == namespace
+                )
+
+                # Add FULLTEXT search
+                fulltext_condition = text(
+                    "MATCH(searchable_content, summary) AGAINST(:query IN NATURAL LANGUAGE MODE)"
+                ).params(query=query)
+                long_query = long_query.filter(fulltext_condition)
+
+                # Add category filter
+                if category_filter:
+                    long_query = long_query.filter(
+                        LongTermMemory.category_primary.in_(category_filter)
+                    )
+
+                # Add relevance score
+                long_results = self.session.execute(
+                    long_query.statement.add_columns(
+                        text(
+                            "MATCH(searchable_content, summary) AGAINST(:query IN NATURAL LANGUAGE MODE) as search_score"
+                        ).params(query=query),
+                        text("'long_term' as memory_type"),
+                        text("'mysql_fulltext' as search_strategy"),
+                    )
+                ).fetchall()
+
+                results.extend([dict(row) for row in long_results])
+
+            return results
+
+        except Exception as e:
+            logger.debug(f"MySQL FULLTEXT search failed: {e}")
+            # Roll back the transaction to recover from error state
+            self.session.rollback()
+            return []
+
+    def _search_postgresql_fts(
+        self,
+        query: str,
+        namespace: str,
+        category_filter: Optional[List[str]],
+        limit: int,
+        search_short_term: bool,
+        search_long_term: bool,
+    ) -> List[Dict[str, Any]]:
+        """Search using PostgreSQL tsvector"""
+        results = []
+
+        try:
+            # Prepare query for tsquery - handle spaces and special characters
+            # Convert simple query to tsquery format (join words with &)
+            tsquery_text = " & ".join(query.split())
+
+            # Search short-term memory if requested
+            if search_short_term:
+                short_query = self.session.query(ShortTermMemory).filter(
+                    ShortTermMemory.namespace == namespace
+                )
+
+                # Add tsvector search
+                ts_query = text(
+                    "search_vector @@ to_tsquery('english', :query)"
+                ).params(query=tsquery_text)
+                short_query = short_query.filter(ts_query)
+
+                # Add category filter
+                if category_filter:
+                    short_query = short_query.filter(
+                        ShortTermMemory.category_primary.in_(category_filter)
+                    )
+
+                # Add relevance score
+                short_results = self.session.execute(
+                    short_query.statement.add_columns(
+                        text(
+                            "ts_rank(search_vector, to_tsquery('english', :query)) as search_score"
+                        ).params(query=tsquery_text),
+                        text("'short_term' as memory_type"),
+                        text("'postgresql_fts' as search_strategy"),
+                    ).order_by(text("search_score DESC"))
+                ).fetchall()
+
+                results.extend([dict(row) for row in short_results])
+
+            # Search long-term memory if requested
+            if search_long_term:
+                long_query = self.session.query(LongTermMemory).filter(
+                    LongTermMemory.namespace == namespace
+                )
+
+                # Add tsvector search
+                ts_query = text(
+                    "search_vector @@ to_tsquery('english', :query)"
+                ).params(query=tsquery_text)
+                long_query = long_query.filter(ts_query)
+
+                # Add category filter
+                if category_filter:
+                    long_query = long_query.filter(
+                        LongTermMemory.category_primary.in_(category_filter)
+                    )
+
+                # Add relevance score
+                long_results = self.session.execute(
+                    long_query.statement.add_columns(
+                        text(
+                            "ts_rank(search_vector, to_tsquery('english', :query)) as search_score"
+                        ).params(query=tsquery_text),
+                        text("'long_term' as memory_type"),
+                        text("'postgresql_fts' as search_strategy"),
+                    ).order_by(text("search_score DESC"))
+                ).fetchall()
+
+                results.extend([dict(row) for row in long_results])
+
+            return results
+
+        except Exception as e:
+            logger.debug(f"PostgreSQL FTS search failed: {e}")
+            # Roll back the transaction to recover from error state
+            self.session.rollback()
+            return []
+
+    def _search_like_fallback(
+        self,
+        query: str,
+        namespace: str,
+        category_filter: Optional[List[str]],
+        limit: int,
+        search_short_term: bool,
+        search_long_term: bool,
+    ) -> List[Dict[str, Any]]:
+        """Fallback LIKE-based search"""
+        results = []
+        search_pattern = f"%{query}%"
+
+        # Search short-term memory
+        if search_short_term:
+            short_query = self.session.query(ShortTermMemory).filter(
+                and_(
+                    ShortTermMemory.namespace == namespace,
+                    or_(
+                        ShortTermMemory.searchable_content.like(search_pattern),
+                        ShortTermMemory.summary.like(search_pattern),
+                    ),
+                )
+            )
+
+            if category_filter:
+                short_query = short_query.filter(
+                    ShortTermMemory.category_primary.in_(category_filter)
+                )
+
+            short_results = (
+                short_query.order_by(
+                    desc(ShortTermMemory.importance_score),
+                    desc(ShortTermMemory.created_at),
+                )
+                .limit(limit)
+                .all()
+            )
+
+            for result in short_results:
+                memory_dict = {
+                    "memory_id": result.memory_id,
+                    "memory_type": "short_term",
+                    "processed_data": result.processed_data,
+                    "importance_score": result.importance_score,
+                    "created_at": result.created_at,
+                    "summary": result.summary,
+                    "category_primary": result.category_primary,
+                    "search_score": 0.4,  # Fixed score for LIKE search
+                    "search_strategy": f"{self.database_type}_like_fallback",
+                }
+                results.append(memory_dict)
+
+        # Search long-term memory
+        if search_long_term:
+            long_query = self.session.query(LongTermMemory).filter(
+                and_(
+                    LongTermMemory.namespace == namespace,
+                    or_(
+                        LongTermMemory.searchable_content.like(search_pattern),
+                        LongTermMemory.summary.like(search_pattern),
+                    ),
+                )
+            )
+
+            if category_filter:
+                long_query = long_query.filter(
+                    LongTermMemory.category_primary.in_(category_filter)
+                )
+
+            long_results = (
+                long_query.order_by(
+                    desc(LongTermMemory.importance_score),
+                    desc(LongTermMemory.created_at),
+                )
+                .limit(limit)
+                .all()
+            )
+
+            for result in long_results:
+                memory_dict = {
+                    "memory_id": result.memory_id,
+                    "memory_type": "long_term",
+                    "processed_data": result.processed_data,
+                    "importance_score": result.importance_score,
+                    "created_at": result.created_at,
+                    "summary": result.summary,
+                    "category_primary": result.category_primary,
+                    "search_score": 0.4,  # Fixed score for LIKE search
+                    "search_strategy": f"{self.database_type}_like_fallback",
+                }
+                results.append(memory_dict)
+
+        return results
+
+    def _get_recent_memories(
+        self,
+        namespace: str,
+        category_filter: Optional[List[str]],
+        limit: int,
+        memory_types: Optional[List[str]],
+    ) -> List[Dict[str, Any]]:
+        """Get recent memories when no search query is provided"""
+        results = []
+
+        search_short_term = not memory_types or "short_term" in memory_types
+        search_long_term = not memory_types or "long_term" in memory_types
+
+        # Get recent short-term memories
+        if search_short_term:
+            short_query = self.session.query(ShortTermMemory).filter(
+                ShortTermMemory.namespace == namespace
+            )
+
+            if category_filter:
+                short_query = short_query.filter(
+                    ShortTermMemory.category_primary.in_(category_filter)
+                )
+
+            short_results = (
+                short_query.order_by(desc(ShortTermMemory.created_at))
+                .limit(limit // 2)
+                .all()
+            )
+
+            for result in short_results:
+                memory_dict = {
+                    "memory_id": result.memory_id,
+                    "memory_type": "short_term",
+                    "processed_data": result.processed_data,
+                    "importance_score": result.importance_score,
+                    "created_at": result.created_at,
+                    "summary": result.summary,
+                    "category_primary": result.category_primary,
+                    "search_score": 1.0,
+                    "search_strategy": "recent_memories",
+                }
+                results.append(memory_dict)
+
+        # Get recent long-term memories
+        if search_long_term:
+            long_query = self.session.query(LongTermMemory).filter(
+                LongTermMemory.namespace == namespace
+            )
+
+            if category_filter:
+                long_query = long_query.filter(
+                    LongTermMemory.category_primary.in_(category_filter)
+                )
+
+            long_results = (
+                long_query.order_by(desc(LongTermMemory.created_at))
+                .limit(limit // 2)
+                .all()
+            )
+
+            for result in long_results:
+                memory_dict = {
+                    "memory_id": result.memory_id,
+                    "memory_type": "long_term",
+                    "processed_data": result.processed_data,
+                    "importance_score": result.importance_score,
+                    "created_at": result.created_at,
+                    "summary": result.summary,
+                    "category_primary": result.category_primary,
+                    "search_score": 1.0,
+                    "search_strategy": "recent_memories",
+                }
+                results.append(memory_dict)
+
+        return results
+
+    def _rank_and_limit_results(
+        self, results: List[Dict[str, Any]], limit: int
+    ) -> List[Dict[str, Any]]:
+        """Rank and limit search results"""
+        # Calculate composite score
+        for result in results:
+            search_score = result.get("search_score", 0.4)
+            importance_score = result.get("importance_score", 0.5)
+            recency_score = self._calculate_recency_score(result.get("created_at"))
+
+            # Weighted composite score
+            result["composite_score"] = (
+                search_score * 0.5 + importance_score * 0.3 + recency_score * 0.2
+            )
+
+        # Sort by composite score and limit
+        results.sort(key=lambda x: x.get("composite_score", 0), reverse=True)
+        return results[:limit]
+
+    def _calculate_recency_score(self, created_at) -> float:
+        """Calculate recency score (0-1, newer = higher)"""
+        try:
+            if not created_at:
+                return 0.0
+
+            if isinstance(created_at, str):
+                created_at = datetime.fromisoformat(created_at.replace("Z", "+00:00"))
+
+            days_old = (datetime.now() - created_at).days
+            return max(0, 1 - (days_old / 30))  # Full score for recent, 0 after 30 days
+        except:
+            return 0.0