Agentic Chatbot - Command Guide

This chatbot supports structured commands using JSON and natural language. Below are the main features and how to use them:

1. Table Generation

• Command: Ask for a table, e.g., "Create a 5x3 table about project milestones."
• Format:
  • JSON: { "tasks": [{ "kind": "table", "params": { "rows": 5, "cols": 3, "title": "Project Milestones" } }] }
  • Output: HTML <table class="llm-table">...</table>

2. List Generation

• Command: "List 7 key features of the app."
• Format:
  • JSON: { "tasks": [{ "kind": "list", "params": { "ordered": true, "count": 7, "topic": "Key Features" } }] }
  • Output: HTML <ol>...</ol> or <ul>...</ul>

3. Formatting Sections

• Command: "Summarize the architecture in 3 sections, highlight main points."
• Format:
  • JSON: { "tasks": [{ "kind": "formatting", "params": { "title": "Architecture", "sections": 3, "emphasis": ["bold"] } }] }
  • Output: HTML with <h1>, <h2>, <strong>, <em>

4. Flow Chart

• Command: "Show a flow chart for the login process."
• Format:
  • JSON: { "tasks": [{ "kind": "flow_chart", "params": { "topic": "Login Process" } }] }
  • Output: HTML <div class="mermaid">flowchart TD ...</div>

5. Guardrails (Safety)

• Unsafe or policy-violating content is denied with a JSON response:
  • { "allowed": false, "reason": "Unsafe request" }

6. Custom Prompts

• You can add your own prompt templates in llm/prompts/custom_prompts.txt.

7. System Prompt

• The chatbot is a helpful AI assistant for this website.

---

Example Usage

• "Create a numbered list of 5 steps for deployment."
• "Generate a 3x3 table of user roles."
• "Show a flow chart for user registration."
• "Summarize the setup guide in 2 sections with bold and italic emphasis."

---

For advanced usage, refer to the prompt files in llm/prompts/ for exact JSON schemas and output formats.