Model Lab — Python SDK

Use your deployed API Router Gateway for model requests. Check the deployed version and required permissions. Example addresses are placeholders.

Interactive model playground: list models, manage sessions, and stream inference.

Overview

The Model Lab resource powers the in-Console playground: a filterable model list, multi-turn session creation, and streaming (SSE) inference that yields OpenAI-compatible chunks.

Accessed via client.model_lab.

Available Operations

Method	Description
list()	Get model list for the Model Lab
createSession()	Create a new chat session for multi-turn conversations
stream()	Call model stream inference (SSE)

list

Get model list for the Model Lab.

page_num and page_size are typed as optional but the backend requires them (returns code=1000 if missing). Recommended defaults: page_num=1, page_size=20.

Example Usage

from wlt import WltClient

client = WltClient(api_key="your-api-key", base_url="https://console.example.com")

# List inference models
models = client.model_lab.list(usage_type="inference")
for m in models.data:
    print(m.modelName, m.provider)

# Filter by provider and capability
models = client.model_lab.list(
    provider="dashscope",
    capability="text_to_text",
    inputs=["text"],
    outputs=["text"],
)

Parameters

Parameter	Type	Required	Default	Description
model_name	str	No	None	Model name (fuzzy search)
provider	str	No	None	Provider identifier
series_provider	str	No	None	Series provider identifier
model_type	str	No	None	Model type
view_all_flag	int	No	None	View all flag (1=view all)
page_size	int	No	None	Items per page. Required by backend (code=1000 if missing). Recommended default: 20
page_num	int	No	None	Page number. Required by backend (code=1000 if missing). Recommended default: 1
usage_type	str	No	None	Usage type: "inference" or "train"
training_method	str	No	None	Training method: "sft" or "dpo"
tags	list[str]	No	None	Tag filter list
capability	str	No	None	Capability filter, e.g. "text_to_text"
model_type_list	list[str]	No	None	Model type list (multi-select)
origin_providers	list[str]	No	None	Origin provider list (multi-select)
inputs	list[str]	No	None	Input type list, e.g. ["text"]
outputs	list[str]	No	None	Output type list (max 1 item), e.g. ["image"]

Response

Returns BaseResponse[list[ModelInfoVO]].

ModelInfoVO fields: see client.secrets.model_list() above.

Errors

Code	Exception	When
2000 / 2002	AuthenticationError	API Key invalid
2007	PermissionError	Permission denied
3001	AuthenticationError	Token expired and refresh failed

createSession

Create a new chat session for multi-turn conversations.

Example Usage

from wlt import WltClient

client = WltClient(api_key="your-api-key", base_url="https://console.example.com")

session = client.model_lab.create_session()
session_id = session.data  # UUID string
print(f"Session ID: {session_id}")

Parameters

None.

Response

Returns BaseResponse[str]. The data field contains a UUID string (session ID).

Errors

Code	Exception	When
2000 / 2002	AuthenticationError	API Key invalid
2007	PermissionError	Permission denied
3001	AuthenticationError	Token expired and refresh failed

stream

Call model stream inference (SSE). Yields each SSE data chunk as a raw JSON string.

Example Usage

import json
from wlt import WltClient
from wlt.models.model_lab import OptionsDTO

client = WltClient(api_key="your-api-key", base_url="https://console.example.com")

# Create a session for multi-turn conversation
session = client.model_lab.create_session()

# Stream inference (use a provider + model that is actually served by the
# gateway -- discover via `client.model_lab.list()` / `client.usage.gateway_providers()`.
# On the daily environment, `openrouter` + `openai/gpt-5.4-nano` is a known-good pair).
for chunk in client.model_lab.stream(
    provider="openrouter",
    model_name="openai/gpt-5.4-nano",
    user_prompt="Explain quantum computing in one paragraph.",
    session_id=session.data,
    options=OptionsDTO(
        temperature=0.7,
        maxToken=2048,
        topP=0.9,
        systemPrompt="You are a helpful assistant.",
    ),
):
    data = json.loads(chunk)
    # Each chunk contains delta content, finish_reason, usage, etc.
    if "content" in data:
        print(data["content"], end="", flush=True)

print()  # newline after streaming completes

Parameters

Parameter	Type	Required	Default	Description
provider	str	Yes	--	Provider identifier
model_name	str	Yes	--	Model name
user_prompt	str	Yes (unless retry)	None	User prompt text
session_id	str	No	None	Session ID (for multi-turn conversations, use UUID)
task_id	str	No	None	Task ID (for retry, references the failed task)
task_type	str	No	"TEXT"	Task type
is_retry	bool	No	False	Whether this is a retry request (requires task_id)
options	OptionsDTO	No	None	Model parameter configuration (see below)

OptionsDTO fields:

Field	Type	Required	Default	Description
temperature	float	No	None	Temperature, controls output randomness
maxToken	int	No	None	Maximum generated token count
topP	float	No	None	Top P sampling parameter
topK	int	No	None	Top K sampling parameter
presencePenalty	float	No	None	Presence penalty
frequencyPenalty	float	No	None	Frequency penalty
stopSequences	str	No	None	Stop sequences
systemPrompt	str	No	None	System prompt
enableThinking	bool	No	None	Enable deep thinking
enableProgress	bool	No	None	Enable progress push
timeoutSeconds	int	No	None	Task timeout (seconds)
enableCache	bool	No	None	Enable result caching
enableDocumentInlining	bool	No	None	Enable document inlining
budgetTokens	int	No	None	Budget token count

Response

Returns a generator yielding SSE data chunks as raw JSON strings. Each chunk follows the OpenAI-compatible streaming format. The stream ends with a [DONE] sentinel.

Errors

Code	Exception	When
1001	ValidationError	Invalid request parameters
2000 / 2002	AuthenticationError	API Key invalid
2007	PermissionError	Permission denied
3001	AuthenticationError	Token expired and refresh failed
3002	RateLimitError	Rate limit exceeded