my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

#1	[Supabase](https://supabase.com/) is an open-source Firebase alternative that provides a PostgreSQL database with pgvector extension for vector similarity search. It offers a powerful and scalable solution for storing and querying vector embeddings.
#2
#3	Create a [Supabase](https://supabase.com/dashboard/projects) account and project, then get your connection string from Project Settings > Database. See the [docs](https://supabase.github.io/vecs/hosting/) for details.
#4
#5	### Usage
#6
#7	<CodeGroup>
#8	```python Python
#9	import os
#10	from mem0 import Memory
#11
#12	os.environ["OPENAI_API_KEY"] = "sk-xx"
#13
#14	config = {
#15	"vector_store": {
#16	"provider": "supabase",
#17	"config": {
#18	"connection_string": "postgresql://user:password@host:port/database",
#19	"collection_name": "memories",
#20	"index_method": "hnsw", # Optional: defaults to "auto"
#21	"index_measure": "cosine_distance" # Optional: defaults to "cosine_distance"
#22	}
#23	}
#24	}
#25
#26	m = Memory.from_config(config)
#27	messages = [
#28	{"role": "user", "content": "I'm planning to watch a movie tonight. Any recommendations?"},
#29	{"role": "assistant", "content": "How about thriller movies? They can be quite engaging."},
#30	{"role": "user", "content": "I'm not a big fan of thriller movies but I love sci-fi movies."},
#31	{"role": "assistant", "content": "Got it! I'll avoid thriller recommendations and suggest sci-fi movies in the future."}
#32	]
#33	m.add(messages, user_id="alice", metadata={"category": "movies"})
#34	```
#35
#36	```typescript Typescript
#37	import { Memory } from "mem0ai/oss";
#38
#39	const config = {
#40	vectorStore: {
#41	provider: "supabase",
#42	config: {
#43	collectionName: "memories",
#44	embeddingModelDims: 1536,
#45	supabaseUrl: process.env.SUPABASE_URL \|\| "",
#46	supabaseKey: process.env.SUPABASE_KEY \|\| "",
#47	tableName: "memories",
#48	},
#49	},
#50	}
#51
#52	const memory = new Memory(config);
#53
#54	const messages = [
#55	{"role": "user", "content": "I'm planning to watch a movie tonight. Any recommendations?"},
#56	{"role": "assistant", "content": "How about thriller movies? They can be quite engaging."},
#57	{"role": "user", "content": "I'm not a big fan of thriller movies but I love sci-fi movies."},
#58	{"role": "assistant", "content": "Got it! I'll avoid thriller recommendations and suggest sci-fi movies in the future."}
#59	]
#60
#61	await memory.add(messages, { userId: "alice", metadata: { category: "movies" } });
#62	```
#63	</CodeGroup>
#64
#65	### SQL Migrations for TypeScript Implementation
#66
#67	The following SQL migrations are required to enable the vector extension and create the memories table:
#68
#69	```sql
#70	-- Enable the vector extension
#71	create extension if not exists vector;
#72
#73	-- Create the memories table
#74	create table if not exists memories (
#75	id text primary key,
#76	embedding vector(1536),
#77	metadata jsonb,
#78	created_at timestamp with time zone default timezone('utc', now()),
#79	updated_at timestamp with time zone default timezone('utc', now())
#80	);
#81
#82	-- Create the vector similarity search function
#83	create or replace function match_vectors(
#84	query_embedding vector(1536),
#85	match_count int,
#86	filter jsonb default '{}'::jsonb
#87	)
#88	returns table (
#89	id text,
#90	similarity float,
#91	metadata jsonb
#92	)
#93	language plpgsql
#94	as $$
#95	begin
#96	return query
#97	select
#98	t.id::text,
#99	1 - (t.embedding <=> query_embedding) as similarity,
#100	t.metadata
#101	from memories t
#102	where case
#103	when filter::text = '{}'::text then true
#104	else t.metadata @> filter
#105	end
#106	order by t.embedding <=> query_embedding
#107	limit match_count;
#108	end;
#109	$$;
#110	```
#111
#112	Go to [Supabase](https://supabase.com/dashboard/projects) and run the above SQL migrations in the SQL Editor.
#113
#114	### Config
#115
#116	Here are the parameters available for configuring Supabase:
#117
#118	<Tabs>
#119	<Tab title="Python">
#120	\| Parameter \| Description \| Default Value \|
#121	\| --- \| --- \| --- \|
#122	\| `connection_string` \| PostgreSQL connection string (required) \| None \|
#123	\| `collection_name` \| Name for the vector collection \| `mem0` \|
#124	\| `embedding_model_dims` \| Dimensions of the embedding model \| `1536` \|
#125	\| `index_method` \| Vector index method to use \| `auto` \|
#126	\| `index_measure` \| Distance measure for similarity search \| `cosine_distance` \|
#127	</Tab>
#128	<Tab title="TypeScript">
#129	\| Parameter \| Description \| Default Value \|
#130	\| --- \| --- \| --- \|
#131	\| `collectionName` \| Name for the vector collection \| `mem0` \|
#132	\| `embeddingModelDims` \| Dimensions of the embedding model \| `1536` \|
#133	\| `supabaseUrl` \| Supabase URL \| None \|
#134	\| `supabaseKey` \| Supabase key \| None \|
#135	\| `tableName` \| Name for the vector table \| `memories` \|
#136	</Tab>
#137	</Tabs>
#138
#139	### Index Methods
#140
#141	The following index methods are supported:
#142
#143	- `auto`: Automatically selects the best available index method
#144	- `hnsw`: Hierarchical Navigable Small World graph index (faster search, more memory usage)
#145	- `ivfflat`: Inverted File Flat index (good balance of speed and memory)
#146
#147	### Distance Measures
#148
#149	Available distance measures for similarity search:
#150
#151	- `cosine_distance`: Cosine similarity (recommended for most embedding models)
#152	- `l2_distance`: Euclidean distance
#153	- `l1_distance`: Manhattan distance
#154	- `max_inner_product`: Maximum inner product similarity
#155
#156	### Best Practices
#157
#158	1. Index Method Selection:
#159	- Use `hnsw` for fastest search performance when memory is not a constraint
#160	- Use `ivfflat` for a good balance of search speed and memory usage
#161	- Use `auto` if unsure, it will select the best method based on your data
#162
#163	2. Distance Measure Selection:
#164	- Use `cosine_distance` for most embedding models (OpenAI, Hugging Face, etc.)
#165	- Use `max_inner_product` if your vectors are normalized
#166	- Use `l2_distance` or `l1_distance` if working with raw feature vectors
#167
#168	3. Connection String:
#169	- Always use environment variables for sensitive information in the connection string
#170	- Format: `postgresql://user:password@host:port/database`
#171

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public