Generative AI

Weaviate's integration with DeepSeek's API allows you to access their generative models' capabilities directly from Weaviate.

Configure a Weaviate collection to use a generative AI model with DeepSeek. Weaviate will perform retrieval augmented generation (RAG) using the specified model and your DeepSeek API key.

More specifically, Weaviate will perform a search, retrieve the most relevant objects, and then pass them to the DeepSeek generative model to generate outputs.

Your Weaviate instance must be configured with the DeepSeek generative AI integration (generative-deepseek) module.

You must provide a valid DeepSeek API key to Weaviate for this integration. Go to DeepSeek to sign up and obtain an API key.

Provide the API key to Weaviate using one of the following methods:

• Set the DEEPSEEK_APIKEY environment variable that is available to Weaviate.
• Provide the API key at runtime, as shown in the examples below.

# Recommended: save sensitive data as environment variables
deepseek_key = os.getenv("DEEPSEEK_APIKEY")

Always set model explicitly. The module's built-in default is a retired model alias, so a collection configured without a model points at a model that DeepSeek no longer serves. See Available models for the current model names.

Configure a Weaviate index as follows to use a DeepSeek generative model:

from weaviate.classes.config import Configureclient.collections.create(    "DemoCollection",    generative_config=Configure.Generative.deepseek(        model="deepseek-v4-flash"    )    # Additional parameters not shown)

Specify any current DeepSeek model name. See Available models for the current names, and Generative parameters for the other settings you can configure alongside it.

You can also override the model at query time.

Configure the following generative parameters to customize the model behavior.

from weaviate.classes.config import Configureclient.collections.create(    "DemoCollection",    generative_config=Configure.Generative.deepseek(        model="deepseek-v4-flash",        # # These parameters are optional        # temperature=0.7,        # max_tokens=500,        # frequency_penalty=0.0,        # presence_penalty=0.0,        # top_p=1.0,        # base_url="https://api.deepseek.com",        # stop=["\n\n"],    ))

Weaviate checks maxTokens against a built-in ceiling only for the retired deepseek-chat and deepseek-reasoner aliases. For any current model, a maxTokens above the model's limit is accepted when you create the collection and fails later, as an error from DeepSeek at query time.

For further details on model parameters, see the DeepSeek API documentation.

Aside from setting the default model provider when creating the collection, you can also override it at query time.

from weaviate.classes.config import Configurefrom weaviate.classes.generate import GenerativeConfigcollection = client.collections.use("DemoCollection")response = collection.generate.near_text(    query="A holiday film",    limit=2,    grouped_task="Write a tweet promoting these two movies",    generative_provider=GenerativeConfig.deepseek(        # # These parameters are optional        model="deepseek-v4-pro",        # temperature=0.7,        # max_tokens=500,        # frequency_penalty=0.0,        # presence_penalty=0.0,        # top_p=1.0,        # base_url="https://api.deepseek.com",        # stop=["\n\n"],    ),    # Additional parameters not shown)

You can provide the API key as well as some optional parameters at runtime through additional headers in the request. The following headers are available:

• X-Deepseek-Api-Key: The DeepSeek API key.
• X-Deepseek-Baseurl: The base URL to use (e.g. a proxy) instead of the default DeepSeek URL.

X-Deepseek-Api-Key takes precedence over the DEEPSEEK_APIKEY environment variable. The API key is never part of the collection configuration, so if neither the header nor the environment variable is set, the request fails with api key: no api key found.

X-Deepseek-Baseurl takes precedence over a baseURL set at query time, which in turn takes precedence over the baseURL in the collection configuration. If none of them are set, Weaviate uses https://api.deepseek.com. Provide an API root rather than a full endpoint path, because Weaviate appends /chat/completions to it.

Provide the headers as shown in the API credentials examples above.

After configuring the generative AI integration, perform RAG operations, either with the single prompt or grouped task method.

To generate text for each object in the search results, use the single prompt method.

The example below generates outputs for each of the n search results, where n is specified by the limit parameter.

When creating a single prompt query, use braces {} to interpolate the object properties you want Weaviate to pass on to the language model. For example, to pass on the object's title property, include {title} in the query.

collection = client.collections.use("DemoCollection")response = collection.generate.near_text(    query="A holiday film",  # The model provider integration will automatically vectorize the query    single_prompt="Translate this into French: {title}",    limit=2)for obj in response.objects:    print(obj.properties["title"])    print(f"Generated output: {obj.generated}")  # Note that the generated output is per object

To generate one text for the entire set of search results, use the grouped task method.

In other words, when you have n search results, the generative model generates one output for the entire group.

collection = client.collections.use("DemoCollection")response = collection.generate.near_text(    query="A holiday film",  # The model provider integration will automatically vectorize the query    grouped_task="Write a fun tweet to promote readers to check out these films.",    limit=2)print(f"Generated output: {response.generative.text}")  # Note that the generated output is per queryfor obj in response.objects:    print(obj.properties["title"])

Weaviate forwards the configured model name to DeepSeek as-is. There is no allowlist on the Weaviate side, so any current DeepSeek model name is accepted.

The current model names are deepseek-v4-flash and deepseek-v4-pro. For the full list of models and pricing, see the DeepSeek pricing page and the DeepSeek API documentation.

The generative-deepseek module returns only the model's message content. If you use a reasoning model, its separate reasoning (chain-of-thought) output is not surfaced through the integration.

Reasoning models can take much longer to respond than non-reasoning models. Weaviate applies the MODULES_CLIENT_TIMEOUT environment variable to the whole request, including reading the response, and it defaults to 50 seconds. If reasoning queries time out, raise this value on your Weaviate instance.

Once the integration is configured at the collection, the data management and search operations in Weaviate work identically to any other collection. See the following model-agnostic examples:

• The How-to: Manage collections and How-to: Manage objects guides show how to perform data operations (i.e. create, read, update, delete collections and objects within them).
• The How-to: Query & Search guides show how to perform search operations (i.e. vector, keyword, hybrid) as well as retrieval augmented generation.

• DeepSeek API documentation

Have a question or feedback? Here's how to reach us.