K2 Horizon

pip install --upgrade pip
pip install -e "python"

from openai import OpenAI

client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY")
response = client.chat.completions.create(
    model="IFM/K2-Horizon-3.7B",
    messages=[{"role": "user", "content": "Hello"}],
)

message = response.choices[0].message
reasoning = getattr(message, "reasoning_content", None)
if reasoning is None:
    reasoning = (message.model_extra or {}).get("reasoning_content")
print("Reasoning:", reasoning)
print("Answer:", message.content)

Reasoning: We have a user who just said "Hello". We need to respond. As an AI, we should be friendly, perhaps ask how can we help. If says "Hello" alone, respond with greeting and offer assistance. Potentially ask what they need. There's no special instruction; just typical.

But maybe we must think about the context: The user gave just a greeting; we should respond with a greeting and ask how we can help. Could also ask if they want anything specific.

We should be concise but friendly: "Hello! How can I assist you today?" Provide some content.

But maybe the user expects something else. But in this scenario, we just reply with a greeting.

Thus answer: "Hello! What can I do for you?" Etc.

Check any style guidelines: We want to keep it short or perhaps ask follow-up. The user didn't ask any question, just saying hello. Could also respond: "Hello! ..."

Thus arguably the answer: "Hi there! How can I help you today?" Let's do that.

But also maybe we could ask if they need help with something specific. That's appropriate.

Thus the final answer: "Hello! How can I assist you today?" Possibly add friendly banter.

We'll respond.

Answer:
Hello! 👋 How can I help you today?

from openai import OpenAI

client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY")
tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Return weather for a city.",
            "parameters": {
                "type": "object",
                "properties": {"city": {"type": "string"}},
                "required": ["city"],
            },
        },
    }
]

response = client.chat.completions.create(
    model="IFM/K2-Horizon-32B",
    messages=[{"role": "user", "content": "Call get_weather for Paris."}],
    tools=tools,
    tool_choice={"type": "function", "function": {"name": "get_weather"}},
    max_tokens=1024,
)

message = response.choices[0].message
reasoning = getattr(message, "reasoning_content", None)
if reasoning is None:
    reasoning = (message.model_extra or {}).get("reasoning_content")
print("Reasoning:", reasoning)
print("Content:", message.content or "")
for tool_call in message.tool_calls or []:
    print(f"Tool call: {tool_call.function.name}({tool_call.function.arguments})")

Reasoning: Okay, the user wants me to call the get_weather function for Paris. Let me check the tools available. There's a function called get_weather that takes a city parameter. The required parameter is city, and it's a string. So I need to make sure to pass "Paris" as the city. I don't see any other parameters needed. The user didn't mention any other details, so I'll just use the city name. Let me structure the tool call correctly. The function name is get_weather, and the arguments should be a JSON object with "city": "Paris". I need to make sure the JSON is properly formatted. Alright, that should do it.

Content:
Tool call: get_weather({"city": "Paris"})