Skip to content
TheChatArchiveThe Chat Archive
Qwenqwen3-235b-a22b-no-thinking· Alibaba (Qwen)Hugging Face LMArena Human Preference 140K

Untitled conversation

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.5-flash’s answer.
  1. User
    flowchart TD
        subgraph "Internal Workflow for 'ask_llm_for_analysis' Tool"
            A(Start<br/>Input: query, context) --> B{Generate<br/>Structured Prompt};
            B -- "Prompt includes required<br/>JSON output schema" --> C[Execute LLM Call<br/>(e.g., Gemini API)];
            C --> D{Parse & Validate<br/>LLM Response};
            D -- "Is output valid JSON<br/>& matches schema?" --> |✅ Yes| G[Return Structured Data<br/>(e.g., a validated JSON object)];
            D -- "❌ No" --> E{Retry/Self-Correct};
            E -- "Re-prompt LLM with error<br/>and request corrected format" --> C;
            E -- "Max retries reached" --> F[Return Error];
            G --> H(End);
            F --> H;
        end
    
    Error: Error: Parse error on line 4:
    ...xecute LLM Call<br/>(e.g., Gemini API)];
    -----------------------^
    Expecting 'SQE', 'DOUBLECIRCLEEND', 'PE', '-)', 'STADIUMEND', 'SUBROUTINEEND', 'PIPE', 'CYLINDEREND', 'DIAMOND_STOP', 'TAGEND', 'TRAPEND', 'INVTRAPEND', 'UNICODE_TEXT', 'TEXT', 'TAGSTART', got 'PS'
    
  2. Qwen Assistant

    The error you're encountering comes from Mermaid's parser, which doesn't recognize certain HTML-like syntax in node labels when it's expecting specific Mermaid syntax tokens. In your case, the issue is with the <br/> tag inside the node label:

    C[Execute LLM Call<br/>(e.g., Gemini API)];
    

    Problem:

    Mermaid does not support HTML tags like <br/> by default in all contexts or environments — especially if you're using an older version or a strict parser.


    ✅ Solution 1: Use \n for Line Breaks (Recommended)

    Replace <br/> with \n to insert a newline character in the label:

    flowchart TD
        subgraph "Internal Workflow for 'ask_llm_for_analysis' Tool"
            A(Start\nInput: query, context) --> B{Generate\nStructured Prompt};
            B -- "Prompt includes required\nJSON output schema" --> C[Execute LLM Call\n(e.g., Gemini API)];
            C --> D{Parse & Validate\nLLM Response};
            D -- "Is output valid JSON\n& matches schema?" --> |✅ Yes| G[Return Structured Data\n(e.g., a validated JSON object)];
            D -- "❌ No" --> E{Retry/Self-Correct};
            E -- "Re-prompt LLM with error\nand request corrected format" --> C;
            E -- "Max retries reached" --> F[Return Error];
            G --> H(End);
            F --> H;
        end
    

    This should work across most Mermaid versions and renderers.


    ✅ Solution 2: Use HTML in Labels (Advanced, if supported)

    If your environment supports HTML in Mermaid, you can use ::<br />:: syntax or wrap the text in HTML explicitly:

    C[<div>Execute LLM Call<br />(e.g., Gemini API)</div>]
    

    But again, this depends on the renderer supporting HTML in Mermaid diagrams.


    ✅ Bonus Tip: Clean Up Arrows

    Ensure that arrow labels are also clean and quoted properly:

    D -- "❌ No" --> E
    E -- "Re-prompt..." --> C
    

    Avoid extra formatting unless necessary.


    Let me know if you'd like help rendering this as a working diagram or converting it into another format like PlantUML!

  3. User
    flowchart TD
        subgraph "Internal Workflow for 'ask_llm_for_analysis' Tool"
            A(Start<br/>Input: query, context) --> B{Generate<br/>Structured Prompt}
            B -- "Prompt includes required<br/>JSON output schema" --> C["Execute LLM Call<br/>(e.g., Gemini API)"]
            C --> D{Parse & Validate<br/>LLM Response}
            D -- "Is output valid JSON<br/>& matches schema?" --> |✅ Yes| G["Return Structured Data<br/>(e.g., a validated JSON object)"]
            D -- "❌ No" --> E{Retry/Self-Correct}
            E -- "Re-prompt LLM with error<br/>and request corrected format" --> C
            E -- "Max retries reached" --> F[Return Error]
            G --> H(End)
            F --> H
        end
    
    Error: Error: Parse error on line 6:
    ...atches schema?" --> |✅ Yes| G["Return St
    -----------------------^
    Expecting 'AMP', 'COLON', 'DOWN', 'DEFAULT', 'NUM', 'COMMA', 'NODE_STRING', 'BRKT', 'MINUS', 'MULT', 'UNICODE_TEXT', got 'PIPE'
    Sample Diagrams
    
  4. Qwen Assistant

    The error you're seeing:

    Expecting 'AMP', 'COLON', 'DOWN', 'DEFAULT', 'NUM', 'COMMA', 'NODE_STRING', 'BRKT', 'MINUS', 'MULT', 'UNICODE_TEXT', got 'PIPE'
    

    ...is caused by the use of the |✅ Yes| label in your Mermaid flowchart.

    ❌ Problem:

    Mermaid does not support the |...| syntax for arrow labels in the way you're using it. That syntax is not valid Mermaid for labeling edges — even though you might have seen it used in some examples or tools (like Mermaid Live Editor with experimental features enabled).

    The pipe characters (|) are confusing the parser.


    ✅ Solution: Use Quoted Labels Instead

    Replace:

    D -- "Is output valid JSON<br/>& matches schema?" --> |✅ Yes| G["Return Structured Data<br/>(e.g., a validated JSON object)"]
    

    With:

    D -- "✅ Yes: Output valid JSON & matches schema" --> G
    

    And optionally, if you want to keep the "no" branch readable:

    D -- "❌ No: Invalid or mismatched schema" --> E
    

    ✅ Updated Working Version

    Here's your corrected diagram:

    flowchart TD
        subgraph "Internal Workflow for 'ask_llm_for_analysis' Tool"
            A(Start<br/>Input: query, context) --> B{Generate<br/>Structured Prompt}
            B -- "Prompt includes required<br/>JSON output schema" --> C["Execute LLM Call<br/>(e.g., Gemini API)"]
            C --> D{Parse & Validate<br/>LLM Response}
            D -- "✅ Yes: Valid JSON & matches schema" --> G["Return Structured Data<br/>(e.g., a validated JSON object)"]
            D -- "❌ No: Invalid or mismatched schema" --> E{Retry/Self-Correct}
            E -- "Re-prompt LLM with error<br/>and request corrected format" --> C
            E -- "Max retries reached" --> F[Return Error]
            G --> H(End)
            F --> H
        end
    

    🔍 Additional Notes:

    • Use <br/> only if your environment supports HTML-like syntax in Mermaid (many do, including Mermaid Live Editor and VSCode extensions).
    • Avoid special syntax like |label| unless you're using a version or fork of Mermaid that explicitly supports it.
    • Always wrap labels containing special characters (like parentheses or quotes) in double quotes.

    Let me know if you'd like to generate a PNG or SVG version of this diagram, or convert it to another format like PlantUML or Graphviz!

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.