Store knowledge in Milvus

The milvus backend embeds knowledge chunks and stores them in a Milvus collection. Use it when you already operate Milvus or need local persistence through Milvus Lite. The VeADK index maps to the Milvus collection name.

Prerequisites

Install the VeADK extensions and configure an embedding model:

pip install "veadk-python[extensions]"

A remote deployment also requires a Milvus URI and credentials. For local validation, point DATABASE_MILVUS_URI to a Milvus Lite database file.

Use Milvus Lite

export DATABASE_MILVUS_URI="./data/company_faq.db"
export MODEL_EMBEDDING_API_KEY="your-ark-api-key"

from veadk import Agent
from veadk.knowledgebase import KnowledgeBase

kb = KnowledgeBase(backend="milvus", index="company_faq")
kb.add_from_text("The standard annual leave allowance is 15 days per year.")

agent = Agent(
    name="company_assistant",
    instruction="Use the knowledge base for company policy questions.",
    knowledgebase=kb,
)

The first run creates data in ./data/company_faq.db. Later processes using the same file and collection name can retrieve the stored content.

Connect to remote Milvus

export DATABASE_MILVUS_URI="https://milvus.example.com:19530"
export DATABASE_MILVUS_TOKEN="your-milvus-token"
export DATABASE_MILVUS_DB_NAME="default"
export MODEL_EMBEDDING_API_KEY="your-ark-api-key"

You can also pass the public configuration object:

from veadk.configs.database_configs import MilvusConfig
from veadk.knowledgebase import KnowledgeBase

kb = KnowledgeBase(
    backend="milvus",
    backend_config={
        "index": "company_faq",
        "milvus_config": MilvusConfig(
            uri="https://milvus.example.com:19530",
            token="your-milvus-token",
            db_name="default",
        ),
    },
)

Inject credentials through environment variables or a secret manager. Do not commit real tokens to source control.

Parameters

Constructor parameters

Parameter	Type	Default	Description
`index`	`str`	None	Milvus collection name. It must start with a letter or underscore, contain only letters, digits, and underscores, and be no longer than 255 characters.
`milvus_config`	`MilvusConfig`	Read from `DATABASE_MILVUS_*`	Milvus connection and collection configuration.
`embedding_config`	`EmbeddingModelConfig`	Read from `MODEL_EMBEDDING_*`	Embedding model configuration.

Milvus configuration

Setting	Environment variable	Type	Default	Description
`uri`	`DATABASE_MILVUS_URI`	`str`	`""`	A Milvus Lite file path or remote Milvus URI. Required.
`token`	`DATABASE_MILVUS_TOKEN`	`str`	`""`	Authentication token. Takes precedence over user and password.
`user`	`DATABASE_MILVUS_USER`	`str`	`""`	Username; used only with `password` when `token` is unset.
`password`	`DATABASE_MILVUS_PASSWORD`	`str`	`""`	Password; used only with `user` when `token` is unset.
`db_name`	`DATABASE_MILVUS_DB_NAME`	`str`	`default`	Database name.
`overwrite`	`DATABASE_MILVUS_OVERWRITE`	`bool`	`false`	Whether initialization replaces an existing collection with the same name.
`timeout`	`DATABASE_MILVUS_TIMEOUT`	`float \| None`	`None`	Connection and request timeout, in the unit expected by the Milvus client.
`output_fields`	`DATABASE_MILVUS_OUTPUT_FIELDS`	`list[str] \| str`	`[]`	Additional fields returned during retrieval. The environment variable accepts a comma-separated value or JSON string array.

Setting DATABASE_MILVUS_OVERWRITE=true replaces the collection and its data. Use it only when the data can be rebuilt.

Embedding configuration

Setting	Environment variable	Type	Default	Description
`name`	`MODEL_EMBEDDING_NAME`	`str`	`doubao-embedding-vision-250615`	Embedding model name.
`dim`	`MODEL_EMBEDDING_DIM`	`int`	`2048`	Vector dimension; it must match the model output.
`api_base`	`MODEL_EMBEDDING_API_BASE`	`str`	`https://ark.cn-beijing.volces.com/api/v3/`	Embedding service endpoint.
`api_key`	`MODEL_EMBEDDING_API_KEY`	`str`	Falls back to `MODEL_AGENT_API_KEY` or an automatically acquired Ark token	Embedding service credential.

Overview

Get started

Core components

Observability

Extensions

Frontend

Security

Prerequisites

Use Milvus Lite

Connect to remote Milvus

Parameters

Constructor parameters

Milvus configuration

Embedding configuration