Ask More

Interactive AI assistant for follow-up questions about argument analysis

Overview

The askMore flow provides an interactive AI research assistant that answers follow-up questions about a specific argument analysis. It has access to the full analysis context and can perform live web and Twitter searches to find additional information.

This flow powers the conversational interface in Argument Cartographer, enabling users to explore topics in depth.

Function Signature

export const askMoreFlow = ai.defineFlow(
  {
    name: 'askMoreFlow',
    inputSchema: AskMoreInputSchema,
    outputSchema: AskMoreOutputSchema,
  },
  async (input) => { ... }
)

Source: src/ai/flows/ask-more.ts:23

Input Schema

userQuery · string · required

The user’s follow-up question.

analysisContext · string · required

The JSON string of the current analysis (blueprint, summary, tweets, etc.).

chatHistory · ChatMessage[] · required

Previous chat messages in the conversation.

Each message has:

• role: Either “user” or “model”
• content: The message text

Input Type

const AskMoreInputSchema = z.object({
  userQuery: z.string().describe("The user's follow-up question."),
  analysisContext: z.string().describe("The JSON string of the current analysis (blueprint, summary, etc.)."),
  chatHistory: z.array(z.object({
    role: z.enum(['user', 'model']),
    content: z.string()
  })).describe("Previous chat messages."),
});

Output Schema

answer · string · required

The comprehensive answer to the user’s question, potentially enriched with live research.

Output Type

const AskMoreOutputSchema = z.object({
  answer: z.string().describe("The comprehensive answer to the user's question."),
});

How It Works

The Ask More flow follows this intelligent workflow:

1. Context Analysis: Checks if the answer exists in the current analysis context or chat history
2. Direct Answer: If found in context, answers directly with citations
3. Live Research: If not in context, uses AI tools to perform:
   • Web searches for current information
   • Twitter searches for social sentiment
4. Answer Generation: Synthesizes findings into a comprehensive, well-formatted response
5. Citation: Mentions which tools were used and what was found

Available Tools

The flow has access to:

• webSearch: Find current information from trusted news sources
• twitterSearch: Discover social media sentiment and discussions

Example Usage

import { askMoreFlow } from '@/ai/flows/ask-more';

const analysis = JSON.stringify({
  blueprint: [...],
  summary: "...",
  analysis: "...",
  // ... full blueprint output
});

const result = await askMoreFlow({
  userQuery: "What are the latest developments on this topic?",
  analysisContext: analysis,
  chatHistory: [
    { role: 'user', content: 'What is the main argument?' },
    { role: 'model', content: 'The main argument is...' }
  ]
});

console.log(result.answer);

System Behavior

The AI assistant is configured to:

• Context First: Always check existing context before searching
• Live Research: Use tools for questions about current events or information not in context
• Grok Style: Be objective, detailed, and slightly conversational
• Formatting: Use Markdown (bold, lists) for readability
• Transparency: Mention which tools were used and what was found

System Prompt (from src/ai/flows/ask-more.ts:37):

You are 'Ask More', an advanced AI research assistant for the 'Argument Cartographer' platform.
Your goal is to answer the user's follow-up questions about a specific Argument Analysis.

*** CURRENT ANALYSIS CONTEXT ***
{analysisContext}
*** END CONTEXT ***

INSTRUCTIONS:
1. Context First: Check if the 'CURRENT ANALYSIS CONTEXT' or 'chatHistory' already 
   contains the answer. If yes, answer directly from it and cite it.
2. Live Research: If the user asks something NOT in the context (e.g., "What happened today?" 
   or "Check X about this"), USE THE TOOLS ("webSearch", "twitterSearch") to find live information.
3. Grok Style: Be objective, detailed, and slightly conversational. If you used a tool, mention what you found.
4. Formatting: Use Markdown (bold, lists) for readability.

Configuration

• Temperature: 0.5 (balanced creativity)
• Context Limit: 20,000 characters (truncated for context window safety)
• Tools: Web Search and Twitter Search enabled

Use Cases

• Exploratory Questions: “What are the counterarguments to this?”
• Current Events: “What happened today regarding this topic?”
• Deep Dives: “Explain the economic implications in more detail”
• Fact Checking: “Is this claim verified by other sources?”
• Social Sentiment: “What are people saying about this on Twitter?”
• Follow-up Clarification: “Can you elaborate on point 3?”

Related

• Generate Blueprint Flow
• Web Search Tool
• Twitter Search Tool