Palette K-LLM · API reference

Overview

Seven routes. One key. Streams when you ask.

7

5

1M

Base URL

Authorization

Content types

Routes

Method	Path	Auth	Purpose

Authentication

Send the key on every call except health.

Header

Key format

Revocation

Quotas

curl

# every route except GET /v1/health
export PALETTE_API_KEY=pk_...

curl https://llm.pltt.dev/v1/models \
  -H "Authorization: Bearer $PALETTE_API_KEY"

401 · missing header

{"detail": "Authorization: Bearer <palette key> 헤더가 필요합니다."}

401 · unknown or revoked key

{"detail": "유효하지 않거나 폐기된 키입니다."}

Endpoints

Seven routes, documented from the code that serves them.

»openapi.json

Schemas

The objects, field by field.

Models

Five seats behind one model field.

default

palette-llm

Palette LLM · Palette managed endpoint · 1M · text only

KMMLU 0.723 · n=300

0.850CLIcK · n=300

vision

palette-llm-max

Palette LLM Max · Palette managed endpoint · 1M

KMMLU 0.736 · n=300

0.802CLIcK · n=300

vision

palette-vlm

Palette Vision · Palette managed endpoint · 1M

alias: palette-llm-fast

4.0 sKorean invoice, total verified

own gpu

palette-llm-ko

Palette LLM Korean · Palette inference server, Palette GPU · 256K

KMMLU 0.640 · n=300 · vision 6.4 s

0.869CLIcK · n=300

Same 300 items, frontier for context

Model	CLIcK n=300	KMMLU n=300	Vision	Context	Where	Status
palette-llm-ko	0.869	0.640	Yes	256K	Palette GPU, Palette inference server	served
palette-llm	0.850	0.723	No	1M	Palette managed endpoint	served
palette-llm-max	0.802	0.736	Yes	1M	Palette managed endpoint	served
palette-vlm	not run	not run	Yes	1M	Palette managed endpoint	served
palette-island	tool calls	FunctionChat 82.2 n=500	No	4K	Palette CPU, warm	served
GPT-6 Astra	0.965	0.853	-	-	not served here	context
Gemini 3.1 Pro	0.957	0.819	-	-	not served here	context
Claude Fable 5.1	0.930	0.803	-	-	not served here	context

Reasoning effort

One switch, four levels, two implementations.

Effort	Shared seats (native thinking)	palette-llm-ko (Korean protocol)	Default max_tokens
low			1024
medium (default)			2048
high			8192
xhigh			16384 (draft) + 4096 (review)

Rules the gateway applies

01
02
03
04
05
06

Streaming

Server-sent events, OpenAI chunk shape.

01
02
03
04

python · consume a stream

stream = client.chat.completions.create(
    model="palette-llm",
    messages=[{"role": "user", "content": "부산에서 서울까지 KTX로 얼마나 걸리나요? 한 문장으로."}],
    stream=True,
    extra_body={"reasoning_effort": "low"},
)
for chunk in stream:
    if not chunk.choices:          # final usage chunk may have empty choices
        continue
    d = chunk.choices[0].delta
    if getattr(d, "reasoning_content", None):
        print(d.reasoning_content, end="", flush=True)
    if d.content:
        print(d.content, end="", flush=True)

Images

Send an image as a content part.

Shape

Per request

Measured

python · palette-vlm

import base64
import os
from openai import OpenAI

client = OpenAI(base_url="https://llm.pltt.dev/v1", api_key=os.environ["PALETTE_API_KEY"])

with open("invoice.png", "rb") as f:
    b64 = base64.b64encode(f.read()).decode()

r = client.chat.completions.create(
    model="palette-vlm",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "이 세금계산서의 합계 금액을 읽고, 품목 금액의 합과 일치하는지 확인해 주세요."},
            {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{b64}"}},
        ],
    }],
    extra_body={"reasoning_effort": "low"},
)
print(r.choices[0].message.content)

Errors

FastAPI-shaped bodies. Korean detail for auth.

Status	When	detail

Limits

What the gateway enforces or inherits.

Limit	Value
Concurrency, shared seats
Concurrency, gateway
Concurrency, palette-llm-ko
Context
Images per request
Request timeout
Default max_tokens
Key cache

Examples

Ready to run. Set PALETTE_API_KEY first.

curl · non-stream · low

curl https://llm.pltt.dev/v1/chat/completions \
  -H "Authorization: Bearer $PALETTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "palette-llm",
    "reasoning_effort": "low",
    "messages": [
      {"role": "user", "content": "공급사 A(단가 낮음, 납기 2주 지연)와 B(단가 8% 높음, 납기 준수) 중 하나를 고르고 이유를 두 문장으로 답하세요."}
    ]
  }'

curl · stream · high

curl -N https://llm.pltt.dev/v1/chat/completions \
  -H "Authorization: Bearer $PALETTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "palette-llm",
    "reasoning_effort": "high",
    "max_tokens": 4000,
    "stream": true,
    "messages": [
      {"role": "user", "content": "공급사 A(단가 낮음, 납기 2주 지연)와 B(단가 8% 높음, 납기 준수) 중 하나를 고르고 이유를 두 문장으로 답하세요."}
    ]
  }'
# chunks arrive as "data: {...}"; reasoning in delta.reasoning_content, answer in delta.content, then "data: [DONE]"

python · openai SDK · non-stream

import os
from openai import OpenAI

client = OpenAI(base_url="https://llm.pltt.dev/v1", api_key=os.environ["PALETTE_API_KEY"])

r = client.chat.completions.create(
    model="palette-llm",
    messages=[{"role": "user", "content": "공급사 A(단가 낮음, 납기 2주 지연)와 B(단가 8% 높음, 납기 준수) 중 하나를 고르고 이유를 두 문장으로 답하세요."}],
    max_tokens=4000,
    extra_body={"reasoning_effort": "high"},
)
msg = r.choices[0].message
print("reasoning:", getattr(msg, "reasoning_content", None))
print("answer:", msg.content)

python · openai SDK · stream

stream = client.chat.completions.create(
    model="palette-llm",
    messages=[{"role": "user", "content": "부산에서 서울까지 KTX로 얼마나 걸리나요? 한 문장으로."}],
    stream=True,
    extra_body={"reasoning_effort": "low"},
)
for chunk in stream:
    if not chunk.choices:
        continue
    d = chunk.choices[0].delta
    if getattr(d, "reasoning_content", None):
        print(d.reasoning_content, end="", flush=True)
    if d.content:
        print(d.content, end="", flush=True)

javascript · openai SDK

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://llm.pltt.dev/v1",
  apiKey: process.env.PALETTE_API_KEY,
});

const r = await client.chat.completions.create({
  model: "palette-llm",
  messages: [{ role: "user", content: "공급사 A(단가 낮음, 납기 2주 지연)와 B(단가 8% 높음, 납기 준수) 중 하나를 고르고 이유를 두 문장으로 답하세요." }],
  max_tokens: 4000,
  reasoning_effort: "high",
});
console.log("reasoning:", r.choices[0].message.reasoning_content);
console.log("answer:", r.choices[0].message.content);

// Streaming
const stream = await client.chat.completions.create({
  model: "palette-llm",
  messages: [{ role: "user", content: "부산에서 서울까지 KTX로 얼마나 걸리나요? 한 문장으로." }],
  stream: true,
  reasoning_effort: "low",
});
for await (const chunk of stream) {
  const d = chunk.choices[0]?.delta ?? {};
  if (d.reasoning_content) process.stdout.write(d.reasoning_content);
  if (d.content) process.stdout.write(d.content);
}

python · langchain · init_chat_model

import os
from langchain.chat_models import init_chat_model

llm = init_chat_model(
    "palette-llm",
    model_provider="openai",
    base_url="https://llm.pltt.dev/v1",
    api_key=os.environ["PALETTE_API_KEY"],
    max_tokens=4000,
    extra_body={"reasoning_effort": "high"},
)

msg = llm.invoke("공급사 A(단가 낮음, 납기 2주 지연)와 B(단가 8% 높음, 납기 준수) 중 하나를 고르고 이유를 두 문장으로 답하세요.")
print("answer:", msg.content)
print("reasoning:", msg.additional_kwargs.get("reasoning_content"))

»Try it in the Playground »Download openapi.json

One endpoint. OpenAI shape. Korean by default.