JSON output

Ask for JSON

import json
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://nokeep.ai/v1",
    api_key=os.environ["NOKEEP_API_KEY"],
)

reply = client.chat.completions.create(
    model="deepseek-v4.1-flash-uncensored",
    response_format={"type": "json_object"},
    messages=[
        {"role": "system", "content": 'Answer as JSON: {"title": string, "tags": string[]}'},
        {"role": "user", "content": "A short story about a lighthouse keeper's last night."},
    ],
)
data = json.loads(reply.choices[0].message.content)
print(data["title"], data["tags"])

response_format reaches DeepSeek as you send it, including json_schema. Describe the shape in the prompt as well: it is what the model reads.

With GLM

glm-5.3-flash-uncensored doesn’t take response_format; it is dropped and named in x-nokeep-ignored. Ask for JSON in the system message, and parse the reply the same way.

Check what comes back

• Parse with json.loads or JSON.parse inside a try, and retry once on a parse error.
• Check the fields you need are there and have the right types. A schema validator such as Pydantic or Zod does this in one line.
• Give the reply room: a max_tokens too small cuts the JSON off before it closes.