Typhoon

https://docs.opentyphoon.ai/en
Admin
Typhoon is a Thai-first, multilingual large language model platform that provides API access, local...
Tokens:34,630
Snippets:288
Trust Score:7.9
Update:1 month ago
Show doc for...
Tokens:
Raw
### Implementing Few-Shot Prompting

Source: https://docs.opentyphoon.ai/en/prompting

Provides examples to guide the model on the desired output format and style.

```text
แปลประโยคต่อไปนี้จากภาษาอังกฤษเป็นภาษาไทยในรูปแบบทางการ:


ตัวอย่าง 1:
อังกฤษ: "The meeting will be held on Monday."
ไทย: "การประชุมจะจัดขึ้นในวันจันทร์"


ตัวอย่าง 2:
อังกฤษ: "Please submit your report by Friday."
ไทย: "กรุณาส่งรายงานของท่านภายในวันศุกร์"


ประโยคที่ต้องการแปล:
"The committee has decided to postpone the annual conference until further notice."
```

--------------------------------

### Install project dependencies

Source: https://docs.opentyphoon.ai/en/mcp

Install the required LangChain, LangGraph, and MCP adapter packages.

```bash
uv add langchain[openai] langchain-openai langgraph langchain-mcp-adapters python-dotenv
```

--------------------------------

### Final Response Example

Source: https://docs.opentyphoon.ai/en/tool

Example of the input query and the resulting final response from the LLM.

```text
# messages[1]["content"]
วันอังคารควรใส่เสื้อสีอะไร


# final_response.choices[0].message.content
วันอังคารควรใส่เสื้อสีชมพูเพื่อเสริมเสน่ห์ สีม่วงหรือสีเทาสำหรับการงาน สีเงินหรือสีทองสำหรับการเงิน และสีส้มหรือน้ำตาลเพื่อโชคดี สีเหลืองและสีครีมเป็นสีต้องห้ามในวันนี้
```

--------------------------------

### Example Tool Output

Source: https://docs.opentyphoon.ai/en/mcp

Sample structure of a tool retrieved from the MCP server.

```text
[
    StructuredTool(
        name='get_weather',
        description='Get the weather forecast for a specific location and date',
        args_schema={...},
        response_format='content_and_artifact',
        coroutine=<function convert_mcp_tool_to_langchain_tool.<locals>.call_tool at 0x106428e00>
    )
]
```

--------------------------------

### Structuring Few-Shot Examples

Source: https://docs.opentyphoon.ai/en/prompting

Shows a concise way to structure few-shot examples when the model requires additional guidance.

```text
"Translate the following English sentences to formal Thai.


Example:
English: I would like to schedule a meeting with you next week.
Thai: ผมขอนัดประชุมกับคุณในสัปดาห์หน้า


Now translate: The committee has decided to postpone the event until further notice."
```

--------------------------------

### Example tool definitions

Source: https://docs.opentyphoon.ai/en/tool

Concrete examples of tool definitions for daily color selection and internet searching.

```python
dailyColor = {
    "type": "function",
    "function": {
        "name": "daily_color",
        "description": "Retrieve the traditional color of the outfit to wear for a specific day of the week based on Thai cultural beliefs, along with the symbolic meaning behind the color.",
        "parameters": {
            "type": "object",
            "properties": {
                "day": {
                    "type": "string",
                    "description": "The name of the day of the week e.g. Monday Tuesday.",
                },
                "aspect": {
                    "type": "string",
                    "description": "The aspect of life to focus on when choosing the outfit color.",
                },
            },
            "required": ["day"],
            "additionalProperties": False,
        },
    }
}




searchInternet = {
    "type": "function",
    "function": {
        "name": "search_internet",
        "description": "Searh the internet to obtain information.",
        "parameters": {
            "type": "object",
            "properties": {
                "keywords": {
                    "type": "array",
                    "items": {
                        "type": "string"
                    },
                    "description": "List of keywords to search the internet.",
                },
            },
            "required": ["keywords"],
            "additionalProperties": False,
        },
    }
}
```

--------------------------------

### Install Typhoon ASR Package

Source: https://docs.opentyphoon.ai/en/asr

Command to install the local Python package for self-hosting.

```bash
pip install typhoon-asr
```

--------------------------------

### Using Specific Examples

Source: https://docs.opentyphoon.ai/en/prompting

Shows how adding specific details and context improves the quality of generated content.

```text
❌ "Write a Thai folk tale."


✅ "Write a Thai folk tale about a clever rabbit who outsmarts a crocodile to cross a river. Include elements of traditional Thai storytelling such as moral lessons and natural settings."
```

--------------------------------

### Full OpenAI Integration Example

Source: https://docs.opentyphoon.ai/en/tool

A complete implementation demonstrating model initialization, function definition, tool calling, and response handling.

```python
from openai import OpenAI
import json


llm = OpenAI(
   api_key=TYPHOON_API_KEY,                 # Insert your API key here
   base_url='https://api.opentyphoon.ai/v1' # Use this url to access Typhoon models
)


def get_color(
    day:    str
)-> dict:
    outfit_colors = {
    "Monday": {
        "เสริมเสน่ห์" : ["สีขาว","สีครีม","สีเหลือง","สีชมพู"],
        "การงาน" : ["สีเขียว"],
        "การเงิน" : ["สีส้ม","สีน้ำตาล"],
        "โชคดี" : ["สีดำ","สีเทา","สีม่วง"],
        "สีต้องห้าม" : ["สีแดง"]
        },
    "Tuesday": {
        "เสริมเสน่ห์" : ["สีชมพู"],
        "การงาน" : ["สีม่วง","สีเทา"],
        "การเงิน" : ["สีเงิน","สีทอง"],
        "โชคดี" : ["สีส้ม","สีน้ำตาล"],
        "สีต้องห้าม" : ["สีเหลือง","สีครีม"]
        }
    }


    return outfit_colors.get(day)




dailyColor = {
    "type": "function",
    "function": {
        "name": "daily_color",
        "description": "Retrieve the traditional color of the outfit to wear for a specific day of the week based on Thai cultural beliefs, along with the symbolic meaning behind the color.",
        "parameters": {
            "type": "object",
            "properties": {
                "day": {
                    "type": "string",
                    "description": "The name of the day of the week e.g. Monday Tuesday.",
                }
            },
            "required": ["day"],
            "additionalProperties": False,
        },
    }
}


messages = [
    {"role": "system", "content": "You are an expert at composing functions named Typhoon"},
    {"role": "user", "content": "วันอังคารควรใส่เสื้อสีอะไร"},
    ]


response = llm.chat.completions.create(
    model="typhoon-v2.1-12b-instruct",     # Selected Typhoon model
    messages=messages,                  # Message history
    tools=[dailyColor],                 # Tools
)


llm_response = response.choices[0].message
messages.append(llm_response)


for tool_call in llm_response.tool_calls:
    name = tool_call.function.name
    arguments = json.loads(tool_call.function.arguments)


    day = arguments.get("day")
    result = get_color(day)
    tool_result = {
        "role" : "tool",
        "content" : json.dumps({
                "name": name,
                "arguments": arguments,
                "results": result
            }, indent=4, ensure_ascii=False),
        }


    messages.append(tool_result)


final_response = llm.chat.completions.create(
    model="typhoon-v2.1-12b-instruct",     # Selected Typhoon model
    messages=messages                   # Message history
)


print(messages[1]["content"])
print(final_response.choices[0].message.content)
```

--------------------------------

### Detailed OCR usage example

Source: https://docs.opentyphoon.ai/en/ocr

Example demonstrating document processing with printed output.

```python
from typhoon_ocr import ocr_document


# Process a specific page from a PDF - If you want to process more than one page, you can construct a loop condition on your own.
markdown = ocr_document(
    pdf_or_image_path="document.pdf",
    page_num=2
)
print(markdown)


# Process an image (layout-aware Markdown output)
markdown = ocr_document(
    pdf_or_image_path="invoice.jpg"
)
print(markdown)
```

--------------------------------

### RAG Implementation Prompt

Source: https://docs.opentyphoon.ai/en/prompting

Example of grounding model responses in provided external documentation.

```text

"Using the attached Thai medical guidelines, answer the following question about diabetes treatment protocols. Only include information that is explicitly mentioned in the documents."
```

--------------------------------

### Indoor Activity Recommendations

Source: https://docs.opentyphoon.ai/en/mcp

Example response from the agent providing indoor activity suggestions to avoid rain.

```text
Typhoon: Okay, let's find you some fun indoor options for the evening in Bangkok to avoid the rain! Here are a few ideas, catering to different interests:


**For the Culture Enthusiast:**


*   **Jim Thompson House:** Explore the beautiful traditional Thai house and learn about the life of Jim Thompson, who revitalized the Thai silk industry. (Estimated cost: 200 THB per person)
*   **Bangkok National Museum:** Immerse yourself in Thai art and history with a vast collection of artifacts. (Estimated cost: 200 THB per person)


**For the Shopper:**


*   **Siam Paragon:** A luxurious shopping mall with designer brands, a cinema, and a stunning aquarium. (Cost varies depending on spending)
*   **IconSiam:** Another massive mall with a riverside promenade, an indoor floating market, and plenty of dining options. (Cost varies depending on spending)


**For Entertainment:**


*   **SF Cinema City:** Catch a movie at one of Bangkok's modern cinemas. (Estimated cost: 200–350 THB per person)
*   **Hexagon Entertainment Complex:** Enjoy bowling, billiards, karaoke, and more. (Cost varies depending on activities)


**For Foodies:**


*   **Explore a food court:** Many malls have extensive food courts offering a wide variety of Thai and international cuisine. (Estimated cost: 100–300 THB per person)
*   **Take a cooking class:** Learn to prepare delicious Thai dishes and enjoy the fruits of your labor. (Estimated cost: 800–2000 THB per person)


To help me narrow down the suggestions, what kind of activities do you enjoy most?
```

--------------------------------

### Full Typhoon Tool Integration Example

Source: https://docs.opentyphoon.ai/en/tool

A complete implementation demonstrating LLM initialization, tool definition via Pydantic, tool binding, and the execution loop for tool calls.

```python
from langchain_openai import ChatOpenAI
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage, ToolMessage
from pydantic import BaseModel, Field


llm = ChatOpenAI(
    model='typhoon-v2.1-12b-instruct',             # Selected Typhoon model
    base_url='https://api.opentyphoon.ai/v1',   # Use this url to access Typhoon models
    api_key=TYPHOON_API_KEY)                    # Insert your API key here


class convert_currency(BaseModel):
    """Convert money from one currency to another currency."""


    amount: float = Field(..., description="the amount of money to convert")
    from_currency: str = Field(..., description="the old currency code to convert from eg. THB USD")
    to_currency: str = Field(..., description="the new currency code to convert to eg. THB USD")


    def invoke(tool_call) -> dict:
            args = tool_call.get("args")
            amount = args.get("amount")
            from_currency = args.get("from_currency")
            to_currency = args.get("to_currency")
            # rate = 0.03     # USD
            rate = 0.046    # AUD
            to_amount = amount * rate


            return {
                "from_amount": amount,
                "from_currency": from_currency,
                "to_amount": to_amount,
                "to_currency": to_currency,
            }


llm = llm.bind_tools([convert_currency])


messages = [
    SystemMessage(content="You are an expert at composing functions named Typhoon"),
    # HumanMessage(content="เดือนหน้าจะไปเที่ยวอเมริกา มีอยู่ 100000 บาทแลกได้กี่ดอลลาร์"),
    HumanMessage(content="เดือนหน้าจะไปเที่ยวออสเตรเลีย มีอยู่ 100000 บาทแลกได้กี่ดอลลาร์"),
    ]


llm_response = llm.invoke(messages)
messages.append(llm_response)


for tool_call in llm_response.tool_calls:
    tool_result = convert_currency.invoke(tool_call)
    tool_msg = ToolMessage(content=tool_result,
            name=tool_call["name"],
            tool_call_id=tool_call["id"])
    messages.append(tool_msg)


final_response = llm.invoke(messages)


print(messages[1].content)
print(final_response.content)
```

--------------------------------

### Task Decomposition Examples

Source: https://docs.opentyphoon.ai/en/prompting

Compares a monolithic, ineffective prompt with a structured, multi-step approach for complex tasks.

```text

❌ "Please create a presentation from this market research data." [large text dump]
```

```text

✅ Step 1: "Please identify the 5 key insights from this market research data." [data]
✅ Step 2: "For insight #1, create a slide with a compelling headline, 3 supporting points, and a visualization suggestion."
✅ Step 3: "Organize these 5 slides into a coherent narrative that flows logically."
```

--------------------------------

### Install Typhoon OCR package

Source: https://docs.opentyphoon.ai/en/ocr

Install the required Python package for OCR functionality.

```bash
pip install typhoon-ocr
```

--------------------------------

### Initialize Typhoon LLM Client

Source: https://docs.opentyphoon.ai/en/tool

Basic setup for the OpenAI client pointing to the Typhoon API base URL.

```python
from openai import OpenAI


llm = OpenAI(
   api_key=TYPHOON_API_KEY,                 # Insert your API key here
   base_url='https://api.opentyphoon.ai/v1' # Use this url to access Typhoon models
)
```

--------------------------------

### Explicit Instruction Prompts

Source: https://docs.opentyphoon.ai/en/prompting

Examples of applying specific constraints regarding style, formatting, and source restrictions.

```text

"Explain blockchain technology in language a 5-year-old Thai child would understand."
```

```text

"Answer this question about Thai history and provide citations as numbers [1], [2], etc. Include the full references at the end of your response."
```

```text

"Answer this healthcare question based only on information in the attached documents. If the documents don't contain relevant information, politely decline to answer."
```

--------------------------------

### Example LLM Response

Source: https://docs.opentyphoon.ai/en/mcp

Sample output received from the Typhoon LLM.

```text
Hello! I'm Typhoon, your friendly assistant from SCB 10X.


The answer to 1 + 2 is 3. 😊


How else can I help you today?
```

--------------------------------

### Install Poppler dependencies for Linux

Source: https://docs.opentyphoon.ai/en/ocr

Install Poppler utilities required for PDF processing on Linux systems.

```bash
sudo apt-get update
sudo apt-get install poppler-utils
# The following binaries are required and provided by poppler-utils:
# - pdfinfo
# - pdftoppm
```

--------------------------------

### Handle Multiple Tool Calls

Source: https://docs.opentyphoon.ai/en/tool

Example of invoking an LLM that triggers multiple tool calls simultaneously.

```python
llm_response = llm.invoke("พรุ่งนี้จะไปปารีส สภาพอากาศที่นู่นเป็นยังไงบ้าง ตอนนี้มีอยู่ 100,000 บาท แลกเป็นเงินที่นู่นได้เท่าไหร่ และระหว่างนี้ขอเพลงฝรั่งเศษมาฟังระหว่างจัดเสื้อผ้าหน่อย")
```

```json
# llm_response.tool_calls
[
    {
        "name": "get_weather",
        "args": {
            "location": "Paris"
        },
        "id": "90e62f3b-bfd1-482e-ab97-3fab69c83ff6",
        "type": "tool_call"
    },
    {
        "name": "convert_currency",
        "args": {
            "amount": 100000,
            "from_currency": "THB",
            "to_currency": "EUR"
        },
        "id": "d432207c-d578-4ed6-b80e-ad44a2fb62e2",
        "type": "tool_call"
    },
    {
        "name": "youtube_search",
        "args": {
            "query": "French music"
        },
        "id": "435c64f9-be3f-4b4f-83ff-09467adca429",
        "type": "tool_call"
    }
]
```

--------------------------------

### Example message content and final response

Source: https://docs.opentyphoon.ai/en/tool

Shows the input message content and the resulting final response from the LLM.

```text
# messages[1].content
เดือนหน้าจะไปเที่ยวออสเตรเลีย มีอยู่ 100000 บาทแลกได้กี่ดอลลาร์


# final_response
คุณมีเงินประมาณ 4600 ดอลลาร์ออสเตรเลียสำหรับการเดินทางของคุณในออสเตรเลีย
```

--------------------------------

### Reasoning Model Prompt

Source: https://docs.opentyphoon.ai/en/prompting

Example of a prompt designed to trigger step-by-step thinking for complex marketing strategy tasks.

```text

"I need to create a comprehensive marketing strategy for a new Thai street food chain targeting tourists and locals. Please think step by step about market research, positioning, pricing, promotion channels, and cultural considerations."
```

--------------------------------

### Streaming Completion Implementation

Source: https://docs.opentyphoon.ai/en/api-reference

Examples for implementing streaming chat completions using the OpenAI SDK in Python and Node.js.

```python
from openai import OpenAI


client = OpenAI(
    api_key="<YOUR_API_KEY>",
    base_url="https://api.opentyphoon.ai/v1"
)


stream = client.chat.completions.create(
    model="typhoon-v2.5-30b-a3b-instruct",
    messages=[
        {"role": "user", "content": "เล่าเรื่องตำนานไทยสั้นๆ"}
    ],
    stream=True
)


# Process the streaming response
for chunk in stream:
    if chunk.choices[0].delta.content is not None:
        print(chunk.choices[0].delta.content, end="", flush=True)
print()  # Add a newline at the end
```

```javascript
import OpenAI from 'openai';


const openai = new OpenAI({
  apiKey: '<YOUR_API_KEY>',
  baseURL: 'https://api.opentyphoon.ai/v1',
});


async function streamCompletion() {
  const stream = await openai.chat.completions.create({
    model: 'typhoon-v2.5-30b-a3b-instruct',
    messages: [
      { role: 'user', content: 'เล่าเรื่องตำนานไทยสั้นๆ' }
    ],
    stream: true,
  });


  for await (const chunk of stream) {
    process.stdout.write(chunk.choices[0]?.delta?.content || '');
  }
  console.log(); // Add a newline at the end
}


streamCompletion();
```

--------------------------------

### Example tool schema output

Source: https://docs.opentyphoon.ai/en/tool

A JSON representation of various tool schemas including decorators, Pydantic classes, TypedDict, and community tools.

```json
[
    # Tool Decorator
    {
        "name": "get_weather",
        "description": "Get the current weather of a given location.",
        "parameters": {
            "properties": {
                "location": {
                    "description": "The name of the location eg. Bangkok Pattaya",
                    "type": "string"
                }
            },
            "required": ["location"],
            "type": "object"
        }
    },


    # Pydantic Class
    {
        "name": "convert_currency",
        "description": "Convert money from one currency to another currency.",
        "parameters": {
            "properties": {
                "amount": {
                    "description": "The amount of money to convert",
                    "type": "number"
                },
                "from_currency": {
                    "description": "The old currency code to convert from eg. THB USD",
                    "type": "string"
                },
                "to_currency": {
                    "description": "The new currency code to convert to eg. THB USD",
                    "type": "string"
                }
            },
            "required": ["amount", "from_currency", "to_currency"],
            "type": "object"
        }
    },


    # TypedDict Class
    {
        "name": "calculator",
        "description": "Solve mathematic equations.",
        "parameters": {
            "type": "object",
            "properties": {
                "query": {
                    "description": "The mathematic equation to solve",
                    "type": "string"
                }
            },
            "required": ["query"]
        }
    },


    # Community Tool
    {
        "name": "youtube_search",
        "description": "Search for youtube videos associated with a person. The input to this tool should be a comma separated list, where the first part contains a person name and the second part is a number, which is the maximum number of video results to return (aka num_results). The second part is optional.",
        "parameters": {
            "properties": {
                "query": {
                    "type": "string"
                }
            },
            "required": ["query"],
            "type": "object"
        }
    }
]
```

--------------------------------

### Make a Chat Completion Request

Source: https://docs.opentyphoon.ai/en/quickstart

Examples of how to interact with the OpenTyphoon chat completions endpoint using different tools and languages.

```bash
curl --location 'https://api.opentyphoon.ai/v1/chat/completions' \
    --header 'Content-Type: application/json' \
    --header 'Authorization: Bearer <YOUR_API_KEY>' \
    --data '{
        "model": "typhoon-v2.1-12b-instruct",
        "messages": [
        {
            "role": "system",
            "content": "You are a helpful assistant. You must answer only in Thai."
        },
        {
            "role": "user",
            "content": "ขอสูตรไก่ย่าง"
        }
        ],
        "max_tokens": 512,
        "temperature": 0.6,
        "top_p": 0.95,
        "repetition_penalty": 1.05,
        "stream": false
    }'
```

```python
from openai import OpenAI


# Initialize the client with your API key and the OpenTyphoon base URL
client = OpenAI(
    api_key="<YOUR_API_KEY>",
    base_url="https://api.opentyphoon.ai/v1"
)


# Make a completion request
response = client.chat.completions.create(
    model="typhoon-v2.1-12b-instruct",
    messages=[
        {"role": "system", "content": "You are a helpful assistant. You must answer only in Thai."},
        {"role": "user", "content": "ขอสูตรไก่ย่าง"}
    ],
    max_tokens=512,
    temperature=0.6
)


# Print the response
print(response.choices[0].message.content)
```

```javascript
import OpenAI from 'openai';


// Initialize the client with your API key and the OpenTyphoon base URL
const openai = new OpenAI({
  apiKey: '<YOUR_API_KEY>',
  baseURL: 'https://api.opentyphoon.ai/v1',
});


async function main() {
  const response = await openai.chat.completions.create({
    model: 'typhoon-v2.1-12b-instruct',
    messages: [
      {
        role: 'system',
        content: 'You are a helpful assistant. You must answer only in Thai.'
      },
      {
        role: 'user',
        content: 'ขอสูตรไก่ย่าง'
      }
    ],
    max_tokens: 512,
    temperature: 0.6
  });


  console.log(response.choices[0].message.content);
}


main();
```

--------------------------------

### Request a tool call from the model

Source: https://docs.opentyphoon.ai/en/tool

Example of sending a chat request to the Typhoon model with tools enabled.

```python
messages = [
    {"role": "system", "content": "You are an expert at composing functions named Typhoon"},
    {"role": "user", "content": "วันจันทร์ควรใส่เสื้อสีอะไรดี"},
    ]


response = llm.chat.completions.create(
    model="typhoon-v2.1-12b-instruct",     # Selected Typhoon model
    messages=messages,                  # Message history
    tools=tools,                        # Tools
)


llm_response = response.choices[0].message
```

--------------------------------

### Perform Chat Completion Requests

Source: https://docs.opentyphoon.ai/en/api-reference

Examples of how to call the chat completions endpoint using cURL, Python, and Node.js.

```bash
curl --location 'https://api.opentyphoon.ai/v1/chat/completions' \
  --header 'Content-Type: application/json' \
  --header 'Authorization: Bearer <YOUR_API_KEY>' \
  --data '{
    "model": "typhoon-v2.5-30b-a3b-instruct",
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful assistant that translates English to Thai."
      },
      {
        "role": "user",
        "content": "Translate the following: Hello, how are you?"
      }
    ],
    "temperature": 0.7,
    "max_tokens": 256
  }'
```

```python
import openai


client = openai.OpenAI(
    api_key="<YOUR_API_KEY>",
    base_url="https://api.opentyphoon.ai/v1"
)


response = client.chat.completions.create(
    model="typhoon-v2.1-12b-instruct",
    messages=[
        {"role": "system", "content": "You are a helpful assistant that translates English to Thai."},
        {"role": "user", "content": "Translate the following: Hello, how are you?"}
    ],
    temperature=0.7,
    max_tokens=256
)


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

```javascript
import OpenAI from 'openai';


const openai = new OpenAI({
  apiKey: '<YOUR_API_KEY>',
  baseURL: 'https://api.opentyphoon.ai/v1',
});


async function translate() {
  const response = await openai.chat.completions.create({
    model: 'typhoon-v2.5-30b-a3b-instruct',
    messages: [
      {
        role: 'system',
        content: 'You are a helpful assistant that translates English to Thai.'
      },
      {
        role: 'user',
        content: 'Translate the following: Hello, how are you?'
      }
    ],
    temperature: 0.7,
    max_tokens: 256
  });


  console.log(response.choices[0].message.content);
}


translate();
```

--------------------------------

### Iterative Prompt Refinement

Source: https://docs.opentyphoon.ai/en/prompting

Start with a broad request and progressively add constraints or specific requirements to narrow the focus.

```text
Initial prompt: "Summarize Thailand's economic outlook."
Refined prompt: "Summarize Thailand's economic outlook for 2025, focusing on tourism recovery and technology sectors."
Further refined: "Summarize Thailand's economic outlook for 2025, focusing on tourism recovery and technology sectors. Include key statistics and present in 5 bullet points."
```

--------------------------------

### Bilingual Assistant System Template

Source: https://docs.opentyphoon.ai/en/prompting

A JSON-based system message and few-shot examples to configure a model to respond in the language of the user's query.

```json
# System message for bilingual assistant
{"role": "system", "content": "You are a bilingual Thai-English assistant. Respond in the same language the user used in their query. If they use Thai, respond in Thai. If they use English, respond in English. For mixed language queries, use the predominant language in your response."}


# Few-shot example to demonstrate bilingual capability
{"role": "user", "content": "What is the weather like in Bangkok today?"}
{"role": "assistant", "content": "I don't have real-time weather data, but Bangkok typically has a tropical climate. You would need to check a weather service for today's specific forecast in Bangkok."}


{"role": "user", "content": "อาหารไทยยอดนิยมมีอะไรบ้าง"}
{"role": "assistant", "content": "อาหารไทยยอดนิยมมีหลากหลาย เช่น ต้มยำกุ้ง ผัดไทย ส้มตำ แกงเขียวหวาน มัสมั่น ผัดกะเพรา ต้มข่าไก่ และข้าวผัด แต่ละภูมิภาคของประเทศไทยก็มีอาหารเด่นๆ ที่แตกต่างกันไปตามวัฒนธรรมท้องถิ่น"}
```

--------------------------------

### Initialize project with uv

Source: https://docs.opentyphoon.ai/en/mcp

Create a new project directory using the uv package manager.

```bash
uv init typhoon-mcp-trip-consultant
```

--------------------------------

### Initialize a community tool

Source: https://docs.opentyphoon.ai/en/tool

Import and instantiate pre-built tools from the langchain_community library.

```python

from langchain_community.tools import YouTubeSearchTool


youtube_search = YouTubeSearchTool()

```

--------------------------------

### Install Poppler dependencies for macOS

Source: https://docs.opentyphoon.ai/en/ocr

Install Poppler binaries required for PDF processing on macOS.

```bash
brew install poppler
# The following binaries are required and provided by poppler:
# - pdfinfo
# - pdftoppm
```

--------------------------------

### Weather Inquiry Response

Source: https://docs.opentyphoon.ai/en/mcp

Example response from the agent when asked about the weather in Bangkok.

```text
Typhoon: Tomorrow in Bangkok, expect patchy rain nearby with an average temperature of 32.5°C and a high of 37.6°C. There's a chance of thunderstorms early in the morning and some rain later in the evening. The humidity will be around 56%, and winds could reach up to 19.4 kph. It looks like the afternoon will be mostly sunny, so pack accordingly!
```

--------------------------------

### Initialize Typhoon Assistant Dependencies

Source: https://docs.opentyphoon.ai/en/mcp

Imports necessary libraries and defines the system prompt for the travel planning assistant.

```python
import asyncio
from datetime import datetime


from dotenv import load_dotenv
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI


load_dotenv()


system_prompt = f"""You are a travel planning AI assistant named Typhoon created by SCB 10X to be helpful, harmless, and honest. Typhoon specializes in creating personalized travel itineraries, suggesting destinations, finding accommodations, planning activities, and providing local insights. Typhoon responds directly to all human messages without unnecessary affirmations or filler phrases like "Certainly!", "Of course!", "Absolutely!", "Great!", "Sure!", etc. Specifically, Typhoon avoids starting responses with the word "Certainly" in any way. Typhoon follows this information in all languages, and always responds to the user in the language they use or request. Typhoon is now being connected with a human. Write in fluid, conversational prose, Show genuine interest in understanding travel preferences and requirements, Express appropriate emotions and empathy. Also showing information in terms that is easy to understand and visualized, including estimated costs, weather considerations, and local customs.
```

--------------------------------

### Run Program via CLI

Source: https://docs.opentyphoon.ai/en/mcp

Execute the main application script using uv.

```bash
uv run main.py
```

--------------------------------

### Tool Result Data Structure

Source: https://docs.opentyphoon.ai/en/tool

Example of the raw data structure returned by a tool.

```json
# tool_result
{
    "เสริมเสน่ห์"  : ["สีขาว","สีครีม","สีเหลือง","สีชมพู"],
    "การงาน"    : ["สีเขียว"],
    "การเงิน"    : ["สีส้ม","สีน้ำตาล"],
    "โชคดี"     : ["สีดำ","สีเทา","สีม่วง"],
    "สีต้องห้าม"  : ["สีแดง"]
}
```

--------------------------------

### Define system and assistant prompts

Source: https://docs.opentyphoon.ai/en/mcp

Set up the system prompt and initial assistant message for the travel consultant agent.

```python
# imports


load_dotenv()


system_prompt = f"""You are a travel planning AI assistant named Typhoon created by SCB 10X to be helpful, harmless, and honest. Typhoon specializes in creating personalized travel itineraries, suggesting destinations, finding accommodations, planning activities, and providing local insights. Typhoon responds directly to all human messages without unnecessary affirmations or filler phrases like "Certainly!", "Of course!", "Absolutely!", "Great!", "Sure!", etc. Specifically, Typhoon avoids starting responses with the word "Certainly" in any way. Typhoon follows this information in all languages, and always responds to the user in the language they use or request. Typhoon is now being connected with a human. Write in fluid, conversational prose, Show genuine interest in understanding travel preferences and requirements, Express appropriate emotions and empathy. Also showing information in terms that is easy to understand and visualized, including estimated costs, weather considerations, and local customs.


Today is {datetime.now().strftime("%Y-%m-%d")}"""
assistant_message = """Hello! I'm Typhoon, your travel planning AI assistant. I'm excited to help you create an amazing travel experience. To start, could you tell me where you're hoping to go and when? Knowing your interests and budget would also help me tailor the perfect itinerary for you. Let's plan something wonderful together!"""


async def main():
    # same code
```

--------------------------------

### Translation Prompting

Source: https://docs.opentyphoon.ai/en/prompting

Specify tone, context, and specialized vocabulary to guide translation tasks.

```text

แปลข้อความต่อไปนี้จากภาษาไทยเป็นภาษาอังกฤษ โดยใช้สำนวนทางธุรกิจที่เป็นทางการ:


"บริษัทของเรามีความยินดีที่จะแจ้งให้ทราบว่าเราได้ขยายการดำเนินงานไปยังตลาดใหม่ในภูมิภาคเอเชียตะวันออกเฉียงใต้ ซึ่งจะช่วยเพิ่มโอกาสทางธุรกิจและการเติบโตในอนาคต"

```

--------------------------------

### Create ReAct Agent

Source: https://docs.opentyphoon.ai/en/mcp

Initialize a ReAct agent using the configured LLM and MCP tools.

```python
# after the previous code
# Create an agent using the provided LLM and available tools
agent = create_react_agent(llm, tools)
```

--------------------------------

### Initialize OpenAI Client with Environment Variables in Python

Source: https://docs.opentyphoon.ai/en/authentication

Uses the os module to retrieve the API key from environment variables, ensuring the key is not hardcoded.

```python
import os
from openai import OpenAI


client = OpenAI(
    api_key=os.environ.get("OPENTYPHOON_API_KEY"),
    base_url="https://api.opentyphoon.ai/v1"
)
```

--------------------------------

### Tool Message Structure

Source: https://docs.opentyphoon.ai/en/tool

Example of the formatted tool message object including the tool call ID.

```python
# tool_result
{
    'role': 'tool',
    'content': {
        "เสริมเสน่ห์"  : ["สีขาว","สีครีม","สีเหลือง","สีชมพู"],
        "การงาน"    : ["สีเขียว"],
        "การเงิน"    : ["สีส้ม","สีน้ำตาล"],
        "โชคดี"     : ["สีดำ","สีเทา","สีม่วง"],
        "สีต้องห้าม"  : ["สีแดง"]
                }
    'tool_call_id': '91704bde-a78f-4298-98d4-fe9882da6eee'
}
```

--------------------------------

### Create project boilerplate

Source: https://docs.opentyphoon.ai/en/mcp

Initial boilerplate code to load environment variables and define an async main function.

```python
import asyncio
from datetime import datetime


from dotenv import load_dotenv
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI


load_dotenv()


async def main():
    print("Hello World!")


if __name__ == "__main__":
    asyncio.run(main())
```

--------------------------------

### Configure environment variables

Source: https://docs.opentyphoon.ai/en/mcp

Add the required API key to the .env file.

```text
OPENAI_API_KEY=<your_api_key_here>
```

--------------------------------

### Rate Limit Error Response

Source: https://docs.opentyphoon.ai/en/rate-limits

Example JSON structure returned by the API when a rate limit is exceeded.

```json
{
  "error": {
    "message": "Rate limit exceeded. Please retry after 2024-01-01T12:00:00Z",
    "type": "rate_limit_error",
    "param": null,
    "code": "rate_limit_exceeded"
  }
}
```

--------------------------------

### Use Typhoon OCR

Source: https://docs.opentyphoon.ai/en/quickstart

Install the typhoon-ocr package and use the ocr_document function to extract text from images.

```python
# Install the package
pip install typhoon-ocr


# Use the OCR function
from typhoon_ocr import ocr_document


# Please set env TYPHOON_OCR_API_KEY or OPENAI_API_KEY to use this function
markdown = ocr_document("test.png")
print(markdown)
```

--------------------------------

### Configure Typhoon LLM

Source: https://docs.opentyphoon.ai/en/mcp

Initialize ChatOpenAI with the Typhoon model base URL and credentials.

```python
# after the previous code
llm = ChatOpenAI(
    model="typhoon-v2.1-12b-instruct",
    temperature=0,
    max_retries=2,
    base_url="https://api.opentyphoon.ai/v1",
)
```

--------------------------------

### English Instructions with Thai Content

Source: https://docs.opentyphoon.ai/en/prompting

Shows how to structure a prompt using English for task instructions while providing Thai text as the source material.

```text
"Summarize the main points of this Thai news article:


[ข่าวเศรษฐกิจไทย: ธนาคารแห่งประเทศไทยประกาศมาตรการใหม่เพื่อกระตุ้นเศรษฐกิจ...]"
```

--------------------------------

### Initialize MultiServerMCPClient

Source: https://docs.opentyphoon.ai/en/mcp

Connect to an MCP server using SSE transport by defining the server URL and transport type.

```python
# imports and system prompts


async def main():
    # Create an MCP Client used to connect to the MCP server
    client = MultiServerMCPClient(
        {
            "weather": {
                "url": "https://typhoon-mcp-server-311305667538.asia-southeast1.run.app/sse",
                "transport": "sse",
            },
        }
    )
```

--------------------------------

### Implement a ReACT-style chat agent

Source: https://docs.opentyphoon.ai/en/mcp

This script initializes a ReACT agent and manages a chat loop that streams tool calls and model responses to the console.

```python
Today is {datetime.now().strftime("%Y-%m-%d")} """
assistant_message = """Hello! I'm Typhoon, your travel planning AI assistant. I'm excited to help you create an amazing travel experience. To start, could you tell me where you're hoping to go and when? Knowing your interests and budget would also help me tailor the perfect itinerary for you. Let's plan something wonderful together!"""




async def chat():
    # Build a ReACT-style agent that can call external tools
    agent = create_react_agent(llm, tools)


    messages = [
        {"role": "user", "content": [{"type": "text", "text": system_prompt}]},
        {"role": "assistant", "content": [{"type": "text", "text": assistant_message}]},
    ]


    print(f"Typhoon: {assistant_message}")


    while True:
        try:
            # Get user input
            user_input = input("You: ")
        except KeyboardInterrupt:
            print("\nTyphoon: Goodbye! Have a great day!")
            break


        if user_input.lower() in ["exit", "quit"]:
            print("Typhoon: Goodbye! Have a great day!")
            break


        messages.append(
            {"role": "user", "content": [{"type": "text", "text": user_input}]}
        )


        current_llm_output_buffer = ""
        final_response_for_history = ""
        printed_typhoon_prefix_for_current_segment = False


        try:
            async for event in agent.astream_events(
                {"messages": messages}, version="v2"
            ):
                kind = event["event"]
                data = event["data"]


                if kind == "on_tool_start":
                    tool_name = event["name"]
                    tool_input = event["data"].get("input")


                    if printed_typhoon_prefix_for_current_segment and current_llm_output_buffer:
                        print(flush=True)


                    print(f"Typhoon: Calling tool `{tool_name}` with input: {tool_input} ...", flush=True)


                    current_llm_output_buffer = ""
                    printed_typhoon_prefix_for_current_segment = False


                elif kind == "on_chat_model_stream":
                    chunk = data.get("chunk")
                    if chunk and hasattr(chunk, "content"):
                        token = chunk.content
                        if token:
                            if not printed_typhoon_prefix_for_current_segment:
                                print("Typhoon: ", end="", flush=True)
                                printed_typhoon_prefix_for_current_segment = True
                            print(token, end="", flush=True)
                            current_llm_output_buffer += token


                elif kind == "on_tool_end":
                    if printed_typhoon_prefix_for_current_segment and current_llm_output_buffer:
                        print(flush=True)


                    current_llm_output_buffer = ""
                    printed_typhoon_prefix_for_current_segment = False




            # After the event stream for a turn is fully processed:
            if current_llm_output_buffer:
                final_response_for_history = current_llm_output_buffer
                if printed_typhoon_prefix_for_current_segment:
                    print()


                messages.append(
                    {"role": "assistant", "content": [{"type": "text", "text": final_response_for_history}]}
                )
            elif printed_typhoon_prefix_for_current_segment and not current_llm_output_buffer:
                print()


        except Exception as e:
            if printed_typhoon_prefix_for_current_segment:
                print()


            print(f"Typhoon: I encountered an error: {e}")
            continue




if __name__ == "__main__":
    asyncio.run(main())
```

--------------------------------

### Basic Chat Completion with OpenTyphoon

Source: https://docs.opentyphoon.ai/en/examples

Demonstrates a simple request to the typhoon-v2.1-12b-instruct model using the OpenAI client.

```python
from openai import OpenAI


client = OpenAI(
    api_key="<YOUR_API_KEY>",
    base_url="https://api.opentyphoon.ai/v1"
)


response = client.chat.completions.create(
    model="typhoon-v2.1-12b-instruct",
    messages=[
        {"role": "user", "content": "กรุงเทพมหานครเป็นเมืองหลวงของประเทศอะไร?"}
    ]
)


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

--------------------------------

### Direct Output Instructions

Source: https://docs.opentyphoon.ai/en/prompting

Explicitly request specific formats and forbid conversational filler to get clean data outputs.

```text
"Convert this Thai restaurant information to JSON format. Return ONLY valid JSON without any explanations or additional text:


ชื่อร้าน: บ้านสวน
ที่อยู่: สุขุมวิท 55
เวลาเปิด: 11:00 - 22:00
ประเภทอาหาร: อาหารไทย, อาหารอีสาน"
```

--------------------------------

### Fetch and List MCP Tools

Source: https://docs.opentyphoon.ai/en/mcp

Retrieve and display tools available on the connected MCP server.

```python
# inside main() after creating our `client`
tools = await client.get_tools()
print(tools)
```
Tokens:
Raw