mem0-tcvectordb

mem0-tcvectordb adds Tencent Cloud’s TCVectorDB as a first-class vector store provider for mem0. Install it next to mem0ai to unlock Tencent Cloud hosting without patching the upstream library.

Highlights

• Registers a tcvectordb provider that works with Memory/AsyncMemory.
• Re-exports the configuration schema so you get validation and editor hints.
• Keeps close parity with the implementation that ships in the upstream mem0 repo.

Installation

pip install mem0-tcvectordb

Usage

import mem0_tcvectordb  # noqa: F401 - registers the provider with mem0
from mem0 import Memory

config = {
    "vector_store": {
        "provider": "tcvectordb",
        "config": {
            "collection_name": "memories",
            "database_name": "mem0",
            "embedding_model_dims": 1536,
            "url": "https://<instance-id>.ap-singapore.tcvdb.tencentcs.com",
            "username": "root",
            "api_key": "tcvdb_api_key",
        }
    }
}

memory = Memory.from_config(config)
memory.add("Alice loves sci-fi movies", user_id="alice")

Configuration Reference

mem0 ships embeddings, so TCVectorDB never runs an embedding model. Ensure embedding_model_dims matches the mem0 embedding size configured for your deployment.

Troubleshooting

• ValueError: TCVectorDB requires url, username, and api_key: supply credentials in the vector_store.config block or pass an authenticated VectorDBClient.
• VectorStoreError: ... failed: verify the Tencent Cloud endpoint is reachable from your runtime and that read_consistency, metric_type, and filters use values supported by TCVectorDB.