StreamLake

StreamLake provides an OpenAI-compatible chat completions API. Use the streamlake/ prefix with a model name or an endpoint ID from your StreamLake account.

Since v1.109.0-dev.2

Provider support was added in PR #41006. Use this development version or a later build that includes the provider.

Overview

Property	Details
Provider prefix	streamlake/
API key environment variable	STREAMLAKE_API_KEY
Default API base	https://vanchin.streamlake.ai/api/gateway/v1/endpoints
Optional API base override	STREAMLAKE_API_BASE or the api_base parameter
Upstream endpoint	/chat/completions
Provider documentation	StreamLake API documentation

Setup

Create an API key in your StreamLake account and export it before you start the SDK or proxy:

export STREAMLAKE_API_KEY="your-streamlake-api-key"

Use streamlake/<model-name> for a model available to your account, or streamlake/<endpoint-id> for a deployed endpoint. Replace the example ep-your-endpoint-id below with your actual endpoint ID. LiteLLM removes the streamlake/ prefix before it sends the model value to StreamLake.

LiteLLM Python SDK

from litellm import completion

response = completion(
    model="streamlake/ep-your-endpoint-id",
    messages=[{"role": "user", "content": "Hello from LiteLLM"}],
    max_tokens=128,
)

print(response.choices[0].message.content)

For streaming, pass stream=True and iterate over the response chunks:

from litellm import completion

response = completion(
    model="streamlake/ep-your-endpoint-id",
    messages=[{"role": "user", "content": "Hello from LiteLLM"}],
    max_tokens=128,
    stream=True,
)

for chunk in response:
    print(chunk.choices[0].delta.content or "", end="")

LiteLLM maps max_tokens to StreamLake's max_completion_tokens. Tool calling and reasoning support depend on the selected model.

LiteLLM Gateway

Save this configuration as config.yaml:

model_list:
  - model_name: streamlake-model
    litellm_params:
      model: streamlake/ep-your-endpoint-id
      api_key: os.environ/STREAMLAKE_API_KEY

Start the proxy with the API key exported in its environment:

litellm --config config.yaml

Send requests using the configured model alias:

curl http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "streamlake-model",
    "messages": [{"role": "user", "content": "Hello from LiteLLM"}],
    "max_tokens": 128
  }'

If your proxy requires authentication, also send Authorization: Bearer <your-litellm-key>. This is your LiteLLM key, not your StreamLake API key.

LiteLLM also accepts Anthropic Messages and Responses API requests for chat models through its translation layer. These requests use StreamLake's chat completions endpoint, not native StreamLake Messages or Responses endpoints.