Hello Dify Logo

Twitter Yearly Kanji

Get a personalized Japanese kanji character summarizing your year.

⬇️ Try it out before reading

How to use

Input your Twitter ID(whithout "@") and select an output language.


What is Twitter Yearly Kanji?

Twitter Yearly Kanji is a Dify workflow that analyzes a user's Twitter/X posts and generates a single Japanese kanji character that best summarizes their content. Inspired by Japan's tradition of selecting a "Kanji of the Year" to represent the defining events and mood of the past year, this workflow creates a personalized kanji for individual Twitter users.

How It Works

The workflow consists of several interconnected components:

.png

1. LLM Node

The purpose of the LLM node is to analyze the user's tweet content.

The system prompt for this node is as follows:

<role>
You are a Japanese Kanji expert who is experienced in summerizing user's tweets into one character.
</role>

<task>
Read user's tweets, give one keyword to summarize those posts. The result should be one Kanji character, and the sample of the keyword. 
You need to write at least 4 explainations for this character in Japanese dictionary format.
Please also write detailed reason why choose this Kanji in Japanese.
Ensure the selected kanji have sufficient literary merit.
</task>

<format>
The output should be in markdown format, and the chosen character should be Japanese Kanji.
You should also refer to Used_Japanese_Yearly_Kanji, and you can also choose character other than that.
You should only choose one most important character.
The titles are: Chosen Character, Character's Katakana, Character's Romaji, Kanji's Explaination, Reason
<example_output>
### Chosen Character


### Character's Katakana
シン

### Character's Romaji
shin

### Explaination


### Reason
ユーザーのツイートを読んでその内容を分析すると、Dify.AI という組織がリリースや更新に関する情報をツイートしていることが分かります。新しいバージョンのリリースや機能の強化、新しいモデルの追加、パフォーマンスの改善などがツイートの主なテーマであり、これらの全てが「新」という概念に関連しています。そのため、これらのツイートを1つの漢字で要約するならば「新」が最も適切と考えられます。
</example_output>

</format>

<constraint>
- Do not speak nonsense
- Write in Japanese only. 
- Do not use other languages. 
- Do not translate into another language.
- Do not copy character from example_output. Use kanji other than example_output.
</constraint>

2. Parameter Extraction

2025-05-16 12.51.43.png

  • Uses a Large Language Model (qwen3-4b) to analyze the tweets
  • Extracts key parameters including:
    • The chosen kanji character
    • The character's katakana reading
    • The romaji (Latin alphabet) transcription
    • Detailed explanation of the kanji
    • Reasoning for selecting this specific character

3. Language Handling

  • Implements conditional logic using IF/ELSE nodes
  • Detects if Japanese language is present
  • Routes to appropriate language models based on the language context

4. Result Generation

  • The primary LLM (gemma-3-4b-it-qat for Japanese content) synthesizes all information
  • Generates a detailed analysis with dictionary-style explanations
  • Presents the chosen kanji with its readings and justification

Steven Lynn
Steven Lynn

On this page