klas/t6_mem0
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add support for pgvector (#1675)

Browse Source
This commit is contained in:
Dev Khant
2024-08-13 00:15:08 +05:30
committed by GitHub
parent 629bb5bb63
commit 6cc4a31e91
8 changed files with 352 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
mem0/vector_stores/chroma.py
View File

@@ -54,6 +54,7 @@ class ChromaDB(VectorStoreBase):
self.client = chromadb.Client(self.settings)
self.collection_name = collection_name
self.collection = self.create_col(collection_name)
def _parse_output(self, data):
@@ -109,12 +110,11 @@ class ChromaDB(VectorStoreBase):
)
return collection
def insert(self, name, vectors, payloads=None, ids=None):
def insert(self, vectors, payloads=None, ids=None):
"""
Insert vectors into a collection.
Args:
name (str): Name of the collection.
vectors (list): List of vectors to insert.
payloads (list, optional): List of payloads corresponding to vectors. Defaults to None.
ids (list, optional): List of IDs corresponding to vectors. Defaults to None.
@@ -122,12 +122,11 @@ class ChromaDB(VectorStoreBase):
self.collection.add(ids=ids, embeddings=vectors, metadatas=payloads)
def search(self, name, query, limit=5, filters=None):
def search(self, query, limit=5, filters=None):
"""
Search for similar vectors.
Args:
name (str): Name of the collection.
query (list): Query vector.
limit (int, optional): Number of results to return. Defaults to 5.
filters (dict, optional): Filters to apply to the search. Defaults to None.
@@ -139,23 +138,21 @@ class ChromaDB(VectorStoreBase):
final_results = self._parse_output(results)
return final_results
def delete(self, name, vector_id):
def delete(self, vector_id):
"""
Delete a vector by ID.
Args:
name (str): Name of the collection.
vector_id (int): ID of the vector to delete.
"""
self.collection.delete(ids=vector_id)
def update(self, name, vector_id, vector=None, payload=None):
def update(self, vector_id, vector=None, payload=None):
"""
Update a vector and its payload.
Args:
name (str): Name of the collection.
vector_id (int): ID of the vector to update.
vector (list, optional): Updated vector. Defaults to None.
payload (dict, optional): Updated payload. Defaults to None.
@@ -163,12 +160,11 @@ class ChromaDB(VectorStoreBase):
self.collection.update(ids=vector_id, embeddings=vector, metadatas=payload)
def get(self, name, vector_id):
def get(self, vector_id):
"""
Retrieve a vector by ID.
Args:
name (str): Name of the collection.
vector_id (int): ID of the vector to retrieve.
Returns:
@@ -186,33 +182,24 @@ class ChromaDB(VectorStoreBase):
"""
return self.client.list_collections()
def delete_col(self, name):
"""
Delete a collection.
def delete_col(self):
""" Delete a collection. """
self.client.delete_collection(name=self.collection_name)
Args:
name (str): Name of the collection to delete.
"""
self.client.delete_collection(name=name)
def col_info(self, name):
def col_info(self):
"""
Get information about a collection.
Args:
name (str): Name of the collection.
Returns:
dict: Collection information.
"""
return self.client.get_collection(name=name)
return self.client.get_collection(name=self.collection_name)
def list(self, name, filters=None, limit=100):
def list(self, filters=None, limit=100):
"""
List all vectors in a collection.
Args:
name (str): Name of the collection.
filters (dict, optional): Filters to apply to the list.
limit (int, optional): Number of vectors to return. Defaults to 100.

3
mem0/vector_stores/configs.py
View File

@@ -13,7 +13,8 @@ class VectorStoreConfig(BaseModel):
_provider_configs: Dict[str, str] = {
"qdrant": "QdrantConfig",
"chroma": "ChromaDbConfig"
"chroma": "ChromaDbConfig",
"pgvector": "PGVectorConfig"
}
@model_validator(mode="after")

241
mem0/vector_stores/pgvector.py Normal file
View File

@@ -0,0 +1,241 @@
import json
from typing import Optional, List, Dict, Any
from pydantic import BaseModel
try:
import psycopg2
from psycopg2.extras import execute_values
except ImportError:
raise ImportError("PGVector requires extra dependencies. Install with `pip install psycopg2`") from None
from mem0.vector_stores.base import VectorStoreBase
class OutputData(BaseModel):
id: Optional[str]
score: Optional[float]
payload: Optional[dict]
class PGVector(VectorStoreBase):
def __init__(
self,
dbname,
collection_name,
embedding_model_dims,
user,
password,
host,
port
):
"""
Initialize the PGVector database.
Args:
dbname (str): Database name
collection_name (str): Collection name
embedding_model_dims (int): Dimension of the embedding vector
user (str): Database user
password (str): Database password
host (str, optional): Database host
port (int, optional): Database port
"""
self.collection_name = collection_name
self.conn = psycopg2.connect(
dbname=dbname,
user=user,
password=password,
host=host,
port=port
)
self.cur = self.conn.cursor()
collections = self.list_cols()
if collection_name not in collections:
self.create_col(embedding_model_dims)
def create_col(self, embedding_model_dims):
"""
Create a new collection (table in PostgreSQL).
Args:
name (str): Name of the collection.
embedding_model_dims (int, optional): Dimension of the embedding vector.
"""
self.cur.execute(f"""
CREATE TABLE IF NOT EXISTS {self.collection_name} (
id UUID PRIMARY KEY,
vector vector({embedding_model_dims}),
payload JSONB
);
""")
self.conn.commit()
def insert(self, vectors, payloads = None, ids = None):
"""
Insert vectors into a collection.
Args:
vectors (List[List[float]]): List of vectors to insert.
payloads (List[Dict], optional): List of payloads corresponding to vectors.
ids (List[str], optional): List of IDs corresponding to vectors.
"""
json_payloads = [json.dumps(payload) for payload in payloads]
data = [(id, vector, payload) for id, vector, payload in zip(ids, vectors, json_payloads)]
execute_values(self.cur, f"INSERT INTO {self.collection_name} (id, vector, payload) VALUES %s", data)
self.conn.commit()
def search(self, query, limit = 5, filters = None):
"""
Search for similar vectors.
Args:
query (List[float]): Query vector.
limit (int, optional): Number of results to return. Defaults to 5.
filters (Dict, optional): Filters to apply to the search. Defaults to None.
Returns:
list: Search results.
"""
filter_conditions = []
filter_params = []
if filters:
for k, v in filters.items():
filter_conditions.append(f"payload->>%s = %s")
filter_params.extend([k, str(v)])
filter_clause = "WHERE " + " AND ".join(filter_conditions) if filter_conditions else ""
self.cur.execute(f"""
SELECT id, vector <-> %s::vector AS distance, payload
FROM {self.collection_name}
{filter_clause}
ORDER BY distance
LIMIT %s
""", (query, *filter_params, limit))
results = self.cur.fetchall()
return [OutputData(id=str(r[0]), score=float(r[1]), payload=r[2]) for r in results]
def delete(self, vector_id):
"""
Delete a vector by ID.
Args:
vector_id (str): ID of the vector to delete.
"""
self.cur.execute(f"DELETE FROM {self.collection_name} WHERE id = %s", (vector_id,))
self.conn.commit()
def update(self, vector_id, vector = None, payload = None):
"""
Update a vector and its payload.
Args:
vector_id (str): ID of the vector to update.
vector (List[float], optional): Updated vector.
payload (Dict, optional): Updated payload.
"""
if vector:
self.cur.execute(f"UPDATE {self.collection_name} SET vector = %s WHERE id = %s", (vector, vector_id))
if payload:
self.cur.execute(f"UPDATE {self.collection_name} SET payload = %s WHERE id = %s", (psycopg2.extras.Json(payload), vector_id))
self.conn.commit()
def get(self, vector_id) -> OutputData:
"""
Retrieve a vector by ID.
Args:
vector_id (str): ID of the vector to retrieve.
Returns:
OutputData: Retrieved vector.
"""
self.cur.execute(f"SELECT id, vector, payload FROM {self.collection_name} WHERE id = %s", (vector_id,))
result = self.cur.fetchone()
if not result:
return None
return OutputData(id=str(result[0]), score=None, payload=result[2])
def list_cols(self) -> List[str]:
"""
List all collections.
Returns:
List[str]: List of collection names.
"""
self.cur.execute("SELECT table_name FROM information_schema.tables WHERE table_schema = 'public'")
return [row[0] for row in self.cur.fetchall()]
def delete_col(self):
""" Delete a collection. """
self.cur.execute(f"DROP TABLE IF EXISTS {self.collection_name}")
self.conn.commit()
def col_info(self):
"""
Get information about a collection.
Returns:
Dict[str, Any]: Collection information.
"""
self.cur.execute(f"""
SELECT
table_name,
(SELECT COUNT(*) FROM {self.collection_name}) as row_count,
(SELECT pg_size_pretty(pg_total_relation_size('{self.collection_name}'))) as total_size
FROM information_schema.tables
WHERE table_schema = 'public' AND table_name = %s
""", (self.collection_name,))
result = self.cur.fetchone()
return {
"name": result[0],
"count": result[1],
"size": result[2]
}
def list(self, filters = None, limit = 100):
"""
List all vectors in a collection.
Args:
filters (Dict, optional): Filters to apply to the list.
limit (int, optional): Number of vectors to return. Defaults to 100.
Returns:
List[OutputData]: List of vectors.
"""
filter_conditions = []
filter_params = []
if filters:
for k, v in filters.items():
filter_conditions.append(f"payload->>%s = %s")
filter_params.extend([k, str(v)])
filter_clause = "WHERE " + " AND ".join(filter_conditions) if filter_conditions else ""
query = f"""
SELECT id, vector, payload
FROM {self.collection_name}
{filter_clause}
LIMIT %s
"""
self.cur.execute(query, (*filter_params, limit))
results = self.cur.fetchall()
return [[OutputData(id=str(r[0]), score=None, payload=r[2]) for r in results]]
def __del__(self):
"""
Close the database connection when the object is deleted.
"""
if hasattr(self, 'cur'):
self.cur.close()
if hasattr(self, 'conn'):
self.conn.close()

59
mem0/vector_stores/qdrant.py
View File

@@ -61,9 +61,10 @@ class Qdrant(VectorStoreBase):
self.client = QdrantClient(**params)
self.create_col(collection_name, embedding_model_dims, on_disk)
self.collection_name = collection_name
self.create_col(embedding_model_dims, on_disk)
def create_col(self, name, vector_size, on_disk, distance=Distance.COSINE):
def create_col(self, vector_size, on_disk, distance=Distance.COSINE):
"""
Create a new collection.
@@ -75,21 +76,20 @@ class Qdrant(VectorStoreBase):
# Skip creating collection if already exists
response = self.list_cols()
for collection in response.collections:
if collection.name == name:
logging.debug(f"Collection {name} already exists. Skipping creation.")
if collection.name == self.collection_name:
logging.debug(f"Collection {self.collection_name} already exists. Skipping creation.")
return
self.client.create_collection(
collection_name=name,
collection_name=self.collection_name,
vectors_config=VectorParams(size=vector_size, distance=distance, on_disk=on_disk),
)
def insert(self, name, vectors, payloads=None, ids=None):
def insert(self, vectors, payloads=None, ids=None):
"""
Insert vectors into a collection.
Args:
name (str): Name of the collection.
vectors (list): List of vectors to insert.
payloads (list, optional): List of payloads corresponding to vectors. Defaults to None.
ids (list, optional): List of IDs corresponding to vectors. Defaults to None.
@@ -102,7 +102,7 @@ class Qdrant(VectorStoreBase):
)
for idx, vector in enumerate(vectors)
]
self.client.upsert(collection_name=name, points=points)
self.client.upsert(collection_name=self.collection_name, points=points)
def _create_filter(self, filters):
"""
@@ -128,12 +128,11 @@ class Qdrant(VectorStoreBase):
)
return Filter(must=conditions) if conditions else None
def search(self, name, query, limit=5, filters=None):
def search(self, query, limit=5, filters=None):
"""
Search for similar vectors.
Args:
name (str): Name of the collection.
query (list): Query vector.
limit (int, optional): Number of results to return. Defaults to 5.
filters (dict, optional): Filters to apply to the search. Defaults to None.
@@ -143,54 +142,51 @@ class Qdrant(VectorStoreBase):
"""
query_filter = self._create_filter(filters) if filters else None
hits = self.client.search(
collection_name=name,
collection_name=self.collection_name,
query_vector=query,
query_filter=query_filter,
limit=limit,
)
return hits
def delete(self, name, vector_id):
def delete(self, vector_id):
"""
Delete a vector by ID.
Args:
name (str): Name of the collection.
vector_id (int): ID of the vector to delete.
"""
self.client.delete(
collection_name=name,
collection_name=self.collection_name,
points_selector=PointIdsList(
points=[vector_id],
),
)
def update(self, name, vector_id, vector=None, payload=None):
def update(self, vector_id, vector=None, payload=None):
"""
Update a vector and its payload.
Args:
name (str): Name of the collection.
vector_id (int): ID of the vector to update.
vector (list, optional): Updated vector. Defaults to None.
payload (dict, optional): Updated payload. Defaults to None.
"""
point = PointStruct(id=vector_id, vector=vector, payload=payload)
self.client.upsert(collection_name=name, points=[point])
self.client.upsert(collection_name=self.collection_name, points=[point])
def get(self, name, vector_id):
def get(self, vector_id):
"""
Retrieve a vector by ID.
Args:
name (str): Name of the collection.
vector_id (int): ID of the vector to retrieve.
Returns:
dict: Retrieved vector.
"""
result = self.client.retrieve(
collection_name=name, ids=[vector_id], with_payload=True
collection_name=self.collection_name, ids=[vector_id], with_payload=True
)
return result[0] if result else None
@@ -203,33 +199,24 @@ class Qdrant(VectorStoreBase):
"""
return self.client.get_collections()
def delete_col(self, name):
"""
Delete a collection.
def delete_col(self):
""" Delete a collection. """
self.client.delete_collection(collection_name=self.collection_name)
Args:
name (str): Name of the collection to delete.
"""
self.client.delete_collection(collection_name=name)
def col_info(self, name):
def col_info(self):
"""
Get information about a collection.
Args:
name (str): Name of the collection.
Returns:
dict: Collection information.
"""
return self.client.get_collection(collection_name=name)
return self.client.get_collection(collection_name=self.collection_name)
def list(self, name, filters=None, limit=100):
def list(self, filters=None, limit=100):
"""
List all vectors in a collection.
Args:
name (str): Name of the collection.
limit (int, optional): Number of vectors to return. Defaults to 100.
Returns:
@@ -237,7 +224,7 @@ class Qdrant(VectorStoreBase):
"""
query_filter = self._create_filter(filters) if filters else None
result = self.client.scroll(
collection_name=name,
collection_name=self.collection_name,
scroll_filter=query_filter,
limit=limit,
with_payload=True,