<aside> 📜 TABLE OF CONTENTS

• Introduction
• Getting Started
• Architecture

Wrap-up

</aside>

Agent Architecture 📐

This section provides a high-level overview of the sequential, agentic architecture powering Ambitus Intelligence. Understanding how each agent fits into the pipeline and how data flows between them is essential for developers, researchers, and administrators looking to extend or maintain the system.

High-Level Overview

Ambitus Intelligence is built as a series of specialized AI agents, each responsible for a discrete stage of a market research workflow. The pipeline is part sequential (with one optional user choice) and part parallel (Agent 3 and 4 can be parallel), guaranteeing determinism and easy debugging:

1. Company Research Agent
2. Industry/Domain Analysis Agent
3. Market Data Agent
4. Competitive Landscape Agent
5. Gap Analysis Agent
6. Opportunity Agent
7. Report Synthesis Agent
8. Citation/Data Collector Agent (runs throughout as a tool)

Each agent receives structured inputs from its predecessor (or the user) and emits structured outputs for the next. Agents communicate via JSON or tabular data, and tooling is provided by Haystack pipelines, external APIs, and LLM prompts.

Components and Modules

Below are the core agent modules and their responsibilities:

1. Company Research Agent

• Purpose: Collect foundational data on the target company.

• Responsibilities:

  • Query web and API sources (Crunchbase [Pricey, Find Loophole] , Wikipedia)
  • Scrape/ingest documents and preprocess text [Not in Primary Stages, for later iterations]
  • Summarize into a structured company profile JSON

• Key Inputs: company_name (string)

• Key Outputs:

  
  {
    "name": "...",
    "industry": "...",
    "description": "...",
    "products": [...],
    "headquarters": "...",
    "sources": [...]
  }
  
  

2. Industry/Domain Analysis Agent

• Purpose: Identify and rank relevant domains where the company can expand.

• Responsibilities:

  • Analyze company profile with an LLM prompt chain
  • Generate a ranked list of candidate domains with rationales

• Key Inputs: Company profile JSON

• Key Outputs:

  
  [
    { "domain": "Retail Tech", "score": 0.92, "rationale": "...", "sources": [...] },
    …
  ]
  
  

3. Market Data Agent

• Purpose: Fetch quantitative market metrics for the chosen domain.

• Responsibilities:

  • Pause for optional domain selection by the user (default: top-most score)
  • Call trend and financial APIs
    • Google Trends [must be possible]
    • Statistica [see how much free tier offers]
    • Web Search
    • Pirated Knowledge Bases
  • Summarize into market statistics and narrative

• Key Inputs: chosen_domain (string)

• Key Outputs:

  
  {
    "market_size_usd": 5e9,
    "CAGR": 0.07,
    "key_drivers": [...],
    "sources": [...]
  }
  
  

4. Competitive Landscape Agent

• Purpose: Map out key competitors and their offerings.

• Responsibilities:

  • Query APIs (Crunchbase, CB Insights) if possible else web search
  • Compile a table of competitors, features, and market positions
  • Provide summary of competitor strengths/weaknesses

• Key Inputs: Company profile + chosen domain

• Key Outputs:

  [
    { "competitor": "A Corp", "product": "X", "market_share": 0.15, "note": "...", "source": "..." },
    …
  ]
  
  

5. Gap Analysis Agent

• Purpose: Detect unmet needs and strategic gaps in the market.

• Responsibilities:

  • Compare company capabilities vs. competitor features
  • Use an LLM to reason over data tables and narratives
  • Output a list of market gaps with evidence

• Key Inputs: Company profile, competitor list, and market stats

• Key Outputs:

  [
    { "gap": "Missing AI-driven personalization", "impact": "High", "evidence": "...", "source": "..." },
    …
  ]
  
  

6. Opportunity Agent

• Purpose: Generate prioritized growth opportunities and use-cases.

• Responsibilities:

  • Brainstorm and validate ideas via LLM prompts
  • Ground each opportunity in data from previous agents
  • Rank by feasibility and strategic impact

• Key Inputs: List of gaps + any additional context

• Key Outputs:

  [
    { "title": "Launch AI Chatbot", "priority": "High", "description": "...", "sources": [...] },
    …
  ]
  
  

7. Report Synthesis Agent

• Purpose: Compile a polished, citation-rich final report.
• Responsibilities:
• Key Inputs: Outputs from Agents 1–6
• Key Outputs: Full Market Research Report text, exportable as Markdown/HTML/PDF

8. Citation/Data Collector Agent [Served as a Tool]

• Purpose: Support all agents by retrieving and formatting citations.
• Responsibilities:
• Key Inputs: Citation requests from any agent
• Key Outputs: Snippet/text with { "url": "...", "excerpt": "...", "title": "..." }

Agentic Flow Diagram

Below is a placeholder for the agent pipeline diagram, illustrating the sequential handoff and optional user interaction for domain selection:

Understanding this agent architecture is foundational before you delve into implementation details, API integration, or production deployment. In the following sections, you’ll find per-agent specifications, sample workflows, and code templates.