Skip to content
BranchesTags

Repository files navigation

Multi-Agent Competitive Intelligence System

An autonomous AI-powered competitive intelligence platform that continuously monitors competitor activity, performs strategic analysis, and generates professional weekly reports — all triggered from a clean dashboard.

Python FastAPI React Mistral AI Qdrant License

What It Does

This system deploys a crew of specialized AI agents that work together to produce a competitive intelligence report on demand. Click one button — the agents scrape the web, analyze the data, and deliver a structured PDF report with SWOT analysis, weak signal detection, and strategic recommendations.

Tracked competitors (configurable): OpenAI · Google DeepMind · Meta AI

Live Demo — Dashboard Pages

Page Description
Dashboard Live stats — articles collected, signals detected, vectors stored, run history
Live Pipeline Animated 6-stage pipeline visualizer with real-time log terminal
Agent Activity All 5 agents with roles, tools, status, and last activity timestamps
Intelligence Per-company SWOT analysis, sentiment momentum, signal velocity chart
Signal Graph D3 force-directed network graph of companies, topics, and signal relationships
Reports PDF download, inline report preview, SWOT summary, key developments
Data Store All collected articles, vector DB stats, search and filter

System Architecture

┌─────────────────────────────────────────────────────────────┐ │ React Dashboard (Port 8000) │ │ Dashboard · Pipeline · Intelligence · Reports │ └────────────────────────┬────────────────────────────────────┘ │ HTTP / REST API ┌────────────────────────▼────────────────────────────────────┐ │ FastAPI Server (api_server.py) │ │ /api/run · /api/status · /api/signals · /api/report │ └──────┬──────────────────┬──────────────────┬───────────────-┘ │ │ │ ┌──────▼──────┐ ┌────────▼────────┐ ┌──────▼──────────────┐ │ Manager │ │ Research Agent │ │ Analysis Agent │ │ Agent │ │ (per competitor)│ │ RAG + SWOT + │ │ Orchestrator│ │ Serper + Firecr │ │ Signal Scoring │ └─────────────┘ └────────┬────────┘ └──────┬──────────────┘ │ │ ┌────────▼────────┐ ┌──────▼──────────────┐ │ Ingestion │ │ Synthesizer Agent │ │ Pipeline │ │ Final Report │ │ Chunk+Embed │ │ Generation │ └────────┬────────┘ └──────┬──────────────┘ │ │ ┌────────▼────────┐ ┌──────▼──────────────┐ │ Qdrant Vector │ │ PDF Report │ │ Database │ │ (ReportLab) │ │ (Cloud) │ │ │ └─────────────────┘ └─────────────────────┘

Agent Roles

Agent Role Tools
Manager Agent Chief Intelligence Officer — orchestrates the crew CrewAI hierarchical process, task delegation
Research Agent Market Intelligence Collector — one per competitor Serper API, Firecrawl, rate-limited retries
Analysis Agent Strategy Analyst — SWOT + signal scoring RAG (Qdrant), Mistral AI, SWOT framework
Synthesizer Agent Executive Report Writer — unified report Multi-company synthesis, recommendation engine
Quality Guard Validation — error recovery + fallback Pydantic validation, retry logic (3 attempts)

Tech Stack

Backend

  • Python 3.10 — core language
  • CrewAI — multi-agent orchestration framework
  • LangChain + Mistral AI — LLM reasoning and RAG pipelines
  • Qdrant Cloud — vector database for semantic search
  • FastAPI + Uvicorn — REST API server
  • Sentence Transformers — local embeddings (all-MiniLM-L6-v2)
  • ReportLab — PDF report generation
  • Serper API — Google News search
  • Firecrawl — web scraping

Frontend

  • React 18 + Vite — fast modern frontend
  • TailwindCSS — utility-first styling
  • Chart.js + react-chartjs-2 — signal velocity charts
  • D3.js — force-directed signal graph
  • Lucide React — icons
  • date-fns — date formatting

Project Structure

Multi-Agent Competitive Intelligence System/ ├── agents/ │ ├── analysis_agent.py # RAG + SWOT + signal analysis │ ├── manager_agent.py # CrewAI orchestrator │ ├── research_agent.py # News scraping per competitor │ └── synthesizer_agent.py # Final report generation ├── config/ │ ├── competitors.yaml # Define which companies to track │ └── settings.py # Central config loader ├── crew/ │ └── intelligence_crew.py # CrewAI crew assembly ├── frontend/ │ ├── src/ │ │ ├── pages/ # Dashboard, Pipeline, Intelligence... │ │ ├── components/ # Sidebar, TopBar, ConnectionStatus │ │ └── services/api.js # All API calls │ └── dist/ # Built frontend (served by FastAPI) ├── monitoring/ │ ├── logger.py # Rich console + JSON file logging │ └── run_tracker.py # Per-run stats and history ├── pipelines/ │ ├── chunker.py # Semantic text chunking │ ├── ingestion_pipeline.py # Full scrape→embed→upsert pipeline │ └── signal_detector.py # Weak signal detection + scoring ├── reports/ │ └── pdf_renderer.py # ReportLab PDF generation ├── storage/ │ ├── database.py # Local JSON article store │ └── vector_store.py # Qdrant client wrapper ├── tasks/ │ ├── analysis_task.py # CrewAI task definitions │ ├── research_task.py │ └── synthesis_task.py ├── tools/ │ ├── rag_tool.py # RAG retrieval from Qdrant │ ├── scraper_tool.py # Firecrawl scraping │ ├── search_tool.py # Serper news search │ └── signal_scorer.py # Velocity + sentiment scoring ├── api_server.py # FastAPI server + pipeline trigger ├── main.py # Entry point ├── start.bat # One-click startup script ├── requirements.txt # Python dependencies ├── .env.example # Environment variable template └── README.md # This file

Quick Start

Prerequisites

  • Python 3.10+
  • Node.js 18+
  • Git

1. Clone the repository

git clone https://github.com/Subrahmanyeswar/multi-agent-competitive-intelligence.git
cd multi-agent-competitive-intelligence

2. Set up Python environment

python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # Mac/Linux
pip install -r requirements.txt

3. Configure environment variables

cp .env.example .env

Open .env and fill in your API keys:

Key Where to get it Free tier
MISTRAL_API_KEY console.mistral.ai Yes
SERPER_API_KEY serper.dev 2500 searches/month
FIRECRAWL_API_KEY firecrawl.dev 500 pages/month
QDRANT_URL + QDRANT_API_KEY cloud.qdrant.io 1GB free

4. Configure competitors

Edit config/competitors.yaml to track the companies you want:

competitors:
  - name: "OpenAI"
    domain: "openai.com"
    keywords: ["OpenAI", "ChatGPT", "GPT-4o"]
    categories: ["product", "partnership", "funding"]

5. Build and run

# Windows — double-click or run:
.\start.bat

# Manual start:
cd frontend && npm install && npm run build && cd ..
python main.py

Open http://localhost:8000 in your browser.

6. Generate your first report

Click Run Pipeline in the dashboard. The system will:

  1. Scrape latest news for all competitors
  2. Chunk and embed articles into Qdrant
  3. Run SWOT + signal analysis via Mistral AI
  4. Generate a PDF competitive intelligence report

Full run takes approximately 5–10 minutes on free API tiers.

Report Output

Each pipeline run produces:

  • PDF report with executive summary, SWOT analysis, key developments, weak signals, strategic recommendations, and 30-day outlook
  • JSON analyses stored in storage/analyses.json
  • Run history logged in storage/run_history.json

Configuration

Adding new competitors

Edit config/competitors.yaml:

competitors:
  - name: "Anthropic"
    domain: "anthropic.com"
    keywords: ["Anthropic", "Claude", "Constitutional AI"]
    categories: ["product", "research", "funding"]

Changing the LLM model

In .env: MISTRAL_MODEL=mistral-large-latest # More accurate, slower MISTRAL_MODEL=mistral-small-latest # Faster, good for free tier

Scheduling weekly runs

python main.py --schedule

Runs every Monday at 08:00 automatically.

API Endpoints

Method Endpoint Description
POST /api/run Trigger pipeline run
GET /api/status Current pipeline status
GET /api/competitors All tracked competitors + stats
GET /api/articles Collected articles (filter by company)
GET /api/signals Weak signal detection results
GET /api/analyses Latest SWOT analyses
GET /api/report/latest Latest report metadata
GET /api/report/download Download latest PDF report
GET /api/vector-stats Qdrant vector DB statistics
GET /api/runs Full run history

Full API docs available at http://localhost:8000/docs

Important Notes

  • Free tier safe — designed to work within free API limits
  • Rate limiting handled — automatic retry with exponential backoff
  • No data committed — .env and storage/ are gitignored
  • Lazy loading — heavy AI libraries load only when pipeline runs

License

MIT License — see LICENSE for details.

👨‍💻 Author

Bhogeswarareddy Katakam

Built with CrewAI, LangChain, Mistral AI, Qdrant, React, and D3.js.

About

No description, website, or topics provided.

Resources

Readme
MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages