GitHub - NitheshK4/MIRA: An AI-powered platform that monitors competitor websites, detects meaningful changes using semantic analysis, scores business impact with a local LLM, and delivers actionable intelligence.

📑 Table of Contents

Click to expand

Overview
What's New — v2.0
Live Demo
System Architecture
ML Pipeline
Chrome Extension
Core Features
AI Strategy Co-Pilot (MIRA Oracle)
Competitive War Room Simulator
Auto-Generated Battlecards
Live Sales Kill Cards 🎯
Model Configurations
Quick Start
Environment Variables
API Reference
Integration Guides
Tech Stack
Project Structure
Deployment
Known Limitations
Contributing
License

🌟 Overview

MIRA (Market Intelligence & Research Automation) is an autonomous, self-healing competitor monitoring engine with an AI-powered strategy co-pilot and competitive war room simulator. It scrapes competitor websites on a configurable schedule, detects meaningful content changes using ONNX sentence embeddings, analyzes business impact with Gemini 3.6 Flash (with a local Qwen GGUF fallback), generates auto-battlecards, and pushes real-time alerts to Slack, Email, and Notion / Airtable CRM — all within a 512 MB RAM footprint.

🔬 Scrape	🧠 Detect	📊 Analyze & Score	⚔️ Simulate	🚨 Alert
Axios + Puppeteer	ONNX Embeddings	Gemini 3.6 Flash / Qwen	War Room + Oracle	Slack + Email + CRM
Static & JS pages	Cosine similarity	Impact 1–10 + Battlecards	Game-theory simulation	Real-time webhooks

✨ Why MIRA?

🧠 Semantic Understanding No brittle string diffs. Changes are compared by meaning, not characters. "Price is $100" ≈ "Current Price: $100" → no false alert.	🔄 Triple-Tier Fallback Cloud LLM → Local GGUF → Rule-based heuristics. MIRA always produces analysis, even fully offline.
🔮 AI Strategy Co-Pilot Ask MIRA Oracle anything — get instant competitive intelligence, pricing strategy advice, and sales objection scripts powered by your live data.	⚔️ War Room Simulator "What if we drop prices 30%?" — simulate competitive market reactions with risk scores, counter-offensives, and timeline predictions.
📢 Multi-Channel Alerts Slack (instant, high-impact ≥ 8), Email (periodic digest), Notion/Airtable (every change logged).	🩹 Self-Healing Failed CRM syncs auto-retry via a local SQLite queue. No data loss, ever.
🧩 Chrome Extension One-click competitor registration from any browser tab. Live badge shows unread intel count.	🃏 Auto-Battlecards LLM-generated competitive battlecards with strengths, weaknesses, killer questions, and sales landmines — ready for your next sales call.

🆕 What's New — v2.0

  ┌──────────────────────────────────────────────────────────────────┐
  │                    🚀  MIRA v2.0 HIGHLIGHTS                      │
  ├──────────────────────────────────────────────────────────────────┤
  │                                                                  │
  │  🔮  AI STRATEGY CO-PILOT (MIRA Oracle)                          │
  │      • Conversational AI with full workspace context             │
  │      • Instant pitch drafts, pricing analysis, SWOT audits       │
  │      • Quick-prompt chips for common strategic queries            │
  │                                                                  │
  │  ⚔️  COMPETITIVE WAR ROOM SIMULATOR                              │
  │      • "What-If" scenario engine with game-theory modeling       │
  │      • Dynamic risk scoring (1-10) with threat levels            │
  │      • Per-competitor predicted reactions & timelines             │
  │      • Step-by-step counter-offensive playbook generation        │
  │      • Pre-built scenario templates for rapid simulation         │
  │                                                                  │
  │  🃏  AUTO-GENERATED BATTLECARDS                                  │
  │      • LLM-powered competitive comparison cards                  │
  │      • Strengths / Weaknesses / Killer Questions / Landmines     │
  │      • One-click refresh from latest intelligence data           │
  │      • CRM sync status tracking (Notion / Airtable)              │
  │                                                                  │
  │  🎯  LIVE SALES OBJECTION "KILL CARDS"                            │
  │      • Real-time 15-second verbatim counter-scripts for pitch calls │
  │      • Live call HUD timer telemetry                             │
  │      • Dynamic displacement triggers & ROI defense per competitor │
  │      • 1-click script copy & AI co-pilot escalation              │
  │                                                                  │
  │  💻  DEVELOPER SDKs & INTEGRATIONS                               │
  │      • Multi-language SDKs for Node.js, Python, Go, and PHP      │
  │      • Scrape triggers, ONNX similarity, War Room & Webhook APIs │
  │      • 1-click Workspace API Key & ID credential copying         │
  │                                                                  │
  │  🔍  ENHANCED CONTENT CHANGE INSPECTOR & ANALYTICS               │
  │      • 3 Review Modes: Split Slider, Side-by-Side & Inline Diff  │
  │      • Color-coded text additions (+ green) & removals (- red)   │
  │      • Metadata bar: HTTP Status, Latency, Payload Delta         │
  │      • 7-Day Change Frequency Timeline & Severity Distribution  │
  │                                                                  │
  └──────────────────────────────────────────────────────────────────┘

🚀 Live Demo

The app is deployed on Railway and publicly accessible:

👉 https://autonomous-competitor-intelligence-engine-production.up.railway.app/

🗺️ System Architecture

graph TD
    subgraph "📡 Input Sources"
        A["🧩 Chrome Extension"] -->|"One-Click API"| B("📋 Sequential Queue")
        C["⏰ Scheduled Runner"] -->|"Configurable Interval"| B
    end

    subgraph "⚙️ Core Pipeline"
        B --> D["🕸️ Double-Engine Scraper"]
        D -->|"Cleaned Text"| E["🧠 ONNX Sentence Embedder"]
        E -->|"Cosine Similarity"| F{"🔍 Semantic Change?"}
        F -->|"similarity >= 0.85"| G["✅ No Action"]
        F -->|"similarity < 0.85"| H["🧠 Gemini / Qwen LLM"]
    end

    subgraph "🃏 Battlecard Engine"
        H -->|"Change Analysis"| BC["🃏 Auto-Battlecard Generator"]
        BC -->|"Strengths / Weaknesses / Landmines"| BD["📋 Battlecard DB"]
    end

    subgraph "🔮 AI Strategy Layer"
        ORACLE["🔮 MIRA Oracle Co-Pilot"]
        WARROOM["⚔️ War Room Simulator"]
        BD --> ORACLE
        BD --> WARROOM
        ORACLE -->|"Conversational AI"| ORACLE_OUT["💬 Strategic Advice"]
        WARROOM -->|"Game Theory"| WAR_OUT["📊 Simulation Report"]
    end

    subgraph "📤 Integrations"
        H --> J["🔄 CRM Adapter"]
        J -->|"Idempotent Write"| K["📓 Notion"]
        J -->|"Idempotent Write"| L["📊 Airtable"]
        J -->|"On Failure"| M["💾 SQLite Retry Queue"]
        H -->|"Impact >= 8"| N["💬 Slack"]
        H -->|"Periodic Digest"| O["📧 Email"]
    end

    style K fill:#000000,stroke:#A78BFA,stroke-width:2px,color:#fff
    style N fill:#4A154B,stroke:#A78BFA,stroke-width:2px,color:#fff
    style O fill:#EA4335,stroke:#A78BFA,stroke-width:2px,color:#fff
    style B fill:#0d1117,stroke:#A78BFA,stroke-width:1px,color:#c9d1d9
    style D fill:#0d1117,stroke:#A78BFA,stroke-width:1px,color:#c9d1d9
    style E fill:#0d1117,stroke:#FFD21E,stroke-width:2px,color:#FFD21E
    style H fill:#0d1117,stroke:#A78BFA,stroke-width:1px,color:#c9d1d9
    style F fill:#161b22,stroke:#A78BFA,stroke-width:2px,color:#c9d1d9
    style ORACLE fill:#1a1040,stroke:#A78BFA,stroke-width:2px,color:#d8b4fe
    style WARROOM fill:#1a1040,stroke:#EC4899,stroke-width:2px,color:#f9a8d4
    style BC fill:#0d1117,stroke:#FFD21E,stroke-width:2px,color:#FFD21E

🧬 ML Pipeline

🧠 Stage 1 — Semantic Change Detection

Runtime: @huggingface/transformers (ONNX, JavaScript) Model: Xenova/all-MiniLM-L6-v2

const { pipeline } = require('@huggingface/transformers');

const embedder = await pipeline(
  'feature-extraction',
  'Xenova/all-MiniLM-L6-v2'
);

const oldEmbed = await embedder("Price is $100",
  { pooling: 'mean', normalize: true });
const newEmbed = await embedder("Current Price: $100",
  { pooling: 'mean', normalize: true });

// Cosine similarity → 0.93
// Same meaning — NO alert triggered ✅

❌ No string comparison — MIRA understands meaning, not characters.

📊 Stage 2 — LLM Analysis & Scoring

Cloud: Gemini 3.6 Flash API Local: Qwen2.5-0.5B GGUF via llama-cli (CPU)

Detected changes are analyzed in a single inference pass. The LLM returns structured output:

Output	Description
📂 Category	LLM-classified change type
📝 Summary	Plain-English explanation
❓ Why It Matters	Business impact analysis
📊 Impact Score	1–10 threat/opportunity rating
📋 Justification	Evidence-based reasoning
🎯 Recommendation	Action item with timeline
🃏 Battlecard	Auto-generated competitive card

🔄 Fallback Chain

┌─────────────────────┐     ┌──────────────────────────┐     ┌─────────────────────────┐
│  Gemini 3.6 Flash   │────▶│  Qwen2.5-0.5B GGUF       │────▶│  Rule-Based Heuristics  │
│  (Cloud, <1.5s)     │     │  (Local, CPU, 7-15s)      │     │  (Keyword match, <1ms)  │
└─────────────────────┘     └──────────────────────────┘     └─────────────────────────┘

When cloud inference is unavailable (rate limits, missing API key), MIRA falls back to local GGUF, then to keyword-based heuristics. Analysis always completes.

🧩 Chrome Extension

The MIRA Chrome Extension turns your browser into a competitor registration tool. Browse any competitor's website, click the extension icon, and it's instantly queued for monitoring.

Feature	Description
🖱️ One-Click Add	Auto-detects the active tab's URL and infers competitor name from the domain
🔴 Live Badge	Service worker polls every 60s — new intel lights up a badge counter on the icon
🔑 API Key Auth	All requests secured with `Bearer` token, configured via Options page
🎯 Scope Selector	Monitor full website, pricing page, or a specific section

📐 Architecture

graph LR
    subgraph "Chrome Extension — Manifest V3"
        POPUP["🖱️ popup.html + popup.js"]
        OPTIONS["⚙️ options.html + options.js"]
        BG["🔄 background.js"]
        STORAGE["💾 chrome.storage.sync"]
    end

    subgraph "MIRA Backend"
        API_ADD["/api/extension/add-competitor"]
        API_UNREAD["/api/extension/unread-count"]
    end

    POPUP -->|"POST + Bearer"| API_ADD
    BG -->|"GET every 60s"| API_UNREAD
    API_UNREAD -->|"{ unreadCount: N }"| BG
    BG -->|"setBadgeText(N)"| BADGE["🔴 Badge"]
    POPUP --> STORAGE
    OPTIONS --> STORAGE
    BG --> STORAGE

    style POPUP fill:#0d1117,stroke:#4285F4,stroke-width:2px,color:#c9d1d9
    style BG fill:#0d1117,stroke:#FF6D00,stroke-width:2px,color:#c9d1d9
    style OPTIONS fill:#0d1117,stroke:#4285F4,stroke-width:1px,color:#c9d1d9
    style BADGE fill:#38bdf8,stroke:#fff,stroke-width:1px,color:#000
    style API_ADD fill:#161b22,stroke:#A78BFA,stroke-width:1px,color:#c9d1d9
    style API_UNREAD fill:#161b22,stroke:#A78BFA,stroke-width:1px,color:#c9d1d9

🔧 Installation

Navigate to chrome://extensions/ and enable Developer mode
Click Load unpacked → select the extension/ directory
Click the MIRA extension icon → Configure Server Settings
Enter your server URL (e.g., http://localhost:3000) and Extension API Key (from dashboard Settings)
Click Verify and Save ✅

⚡ Core Features

🕸️ Intelligent Double-Engine Scraper

Engine	Library	Use Case
⚡ Fast Fetch	`axios` + `cheerio`	Static HTML — fast and lightweight
🌐 JS Render	`puppeteer` (headless Chromium)	SPAs & dynamic JS content

Smart Cleaning Pipeline:

🧹 Strips cookie banners, nav bars, footers, sidebars
🔄 Rotates User-Agent strings to avoid bot detection
🖼️ Blocks images/CSS in Puppeteer to minimize RAM
📸 Captures screenshots for audit trails

🔍 Tech Stack & DNS Enrichment

🌐 DNS Resolution — A-records & MX-records for server/email hosting detection
🔧 Header Inspection — Reads server, x-powered-by headers
📊 Dashboard Widget — Enriched tech profiles displayed in competitor sidebar

💼 Idempotent CRM Sync

🔒 Deduplication — Queries Notion/Airtable before writes
🔄 SQLite Retry Queue — Failed syncs queued locally, auto-retried
📊 Dynamic Schema Matching — Case-insensitive, whitespace-tolerant
✅ Status Tracking — synced / failed / pending per card

📢 Multi-Channel Alerts

Channel	Trigger	Content
💬 Slack	Impact ≥ 8	Immediate webhook with full intel card
📧 Email	Periodic digest	HTML summary of recent changes
📓 Notion	Every change	Structured database row
📊 Airtable	Every change	Structured record

🖥️ Rich Dashboard UI

Built with React 18 + Tailwind CSS + Vite, the dashboard features:

🔍 Command Palette — Cmd/Ctrl + K global search & quick navigation
📊 Intel Feed — Card-based timeline of all detected changes, filterable by impact
🗂️ Visual Diff Modal — Side-by-side content diff viewer for every scrape
🔔 Live Badge — Unread intel count synced with the Chrome Extension badge
⚙️ Settings Panel — Per-workspace CRM keys, SMTP, Slack webhook, and API key management
💀 Skeleton Loaders — Smooth loading states throughout the UI

🔮 AI Strategy Co-Pilot — MIRA Oracle

MIRA Oracle is an embedded AI strategy co-pilot that has real-time access to your entire competitive intelligence workspace — business profile, competitor radar, intel signals, and battlecards. Ask it anything and get actionable strategic advice instantly.

💡 What Can You Ask?

"How should I position against [Competitor X] in enterprise deals?"
"Draft a 60-second elevator pitch highlighting our advantages"
"What are the top 3 pricing risks if we go freemium?"
"Summarize all high-impact competitor moves this week"
"Write objection-handling scripts for [Competitor]'s latest feature"

⚡ Quick-Prompt Chips

Pre-built strategy prompts for instant access:

Chip	Action
🎯 Competitor SWOT	Full SWOT analysis of a tracked competitor
💰 Pricing Strategy	Price positioning based on market signals
📝 Pitch Draft	60-second elevator pitch generator
🛡️ Objection Handling	Counter-arguments for sales calls
📊 Weekly Briefing	Executive summary of all market movements

🏗️ How It Works

graph LR
    USER["👤 User Query"] --> CTX["📦 Context Assembler"]
    CTX --> |"Profile + Competitors + Intel + Battlecards"| LLM["🧠 Gemini 3.6 Flash"]
    LLM --> RESP["💬 Strategic Response"]
    RESP --> CONV["🔄 Conversation Memory"]
    CONV --> |"Multi-turn context"| CTX

    style USER fill:#0d1117,stroke:#A78BFA,stroke-width:2px,color:#c9d1d9
    style CTX fill:#0d1117,stroke:#FFD21E,stroke-width:2px,color:#FFD21E
    style LLM fill:#1a1040,stroke:#A78BFA,stroke-width:2px,color:#d8b4fe
    style RESP fill:#0d1117,stroke:#00C853,stroke-width:2px,color:#00C853
    style CONV fill:#161b22,stroke:#A78BFA,stroke-width:1px,color:#c9d1d9

Fallback: When Gemini is unavailable, Oracle generates intelligent contextual reports using your live workspace data — competitor counts, high-impact signals, and rule-based strategic recommendations.

⚔️ Competitive War Room Simulator

"What happens if we drop prices by 30%?" — The War Room runs a full market simulation powered by Gemini, modeling how each competitor is likely to react, with timelines, risk scores, and a step-by-step counter-offensive playbook.

📊 Simulation Output Structure

┌────────────────────────────────────────────────────────────────┐
│                    ⚔️  WAR ROOM SIMULATION                     │
├────────────────────────────────────────────────────────────────┤
│                                                                │
│  📋 SCENARIO:  "Drop pricing 30% across all tiers"             │
│                                                                │
│  🎯 RISK SCORE:     ████████░░  8/10 — HIGH RISK               │
│                                                                │
│  📈 MARKET IMPACT:                                             │
│     "Aggressive pricing will trigger a race-to-bottom in       │
│      the mid-market segment within 2-3 weeks..."               │
│                                                                │
│  👥 COMPETITOR RESPONSES:                                      │
│  ┌────────────┬──────────────────────┬───────┬─────────┐       │
│  │ Competitor  │ Predicted Action      │ Prob  │ When     │      │
│  ├────────────┼──────────────────────┼───────┼─────────┤       │
│  │ Rival A     │ Match pricing +       │ 85%   │ 1-2 wks  │      │
│  │             │  launch ad blitz      │       │          │      │
│  │ Rival B     │ Highlight premium     │ 60%   │ 2-4 wks  │      │
│  │             │  positioning          │       │          │      │
│  └────────────┴──────────────────────┴───────┴─────────┘       │
│                                                                │
│  🛡️ COUNTER-OFFENSIVE PLAYBOOK:                                │
│     Step 1 → Days 1-7:  Distribute battlecards to sales        │
│     Step 2 → Weeks 2-4: Publish customer success stories       │
│     Step 3 → Month 2:   Release differentiation features       │
│                                                                │
│  ⚖️ VERDICT: PROCEED WITH CAUTION                              │
│                                                                │
└────────────────────────────────────────────────────────────────┘

🎮 Pre-Built Scenario Templates

Template	Description
💰 Price Drop	"Drop prices by 30% across all tiers"
🚀 AI Feature Launch	"Launch an AI-powered feature competitor doesn't have"
🆓 Freemium Pivot	"Introduce a free tier to capture market share"
🏢 Enterprise Push	"Pivot to enterprise-only with premium support"

💡 Or type any custom hypothesis — e.g., "Acquire competitor X's top engineer and launch a PR campaign"

🃏 Auto-Generated Battlecards

Every tracked competitor gets an auto-generated battlecard powered by LLM analysis. Battlecards are living documents that evolve as new intelligence signals are captured.

📋 Card Contents

Each battlecard includes:

✅ Strengths — What they do well
❌ Weaknesses — Where they fall short
🎯 Killer Questions — Ask prospects to expose competitor gaps
💣 Sales Landmines — Trigger questions that shift the deal in your favor

🔄 Lifecycle

Stage	Mechanism
Generation	Auto-created on first competitor scrape
Refresh	One-click regenerate from latest intel data
Sync	Status tracked: `synced` / `pending` / `failed`
CRM Push	Auto-syncs to Notion / Airtable alongside intel cards

🎯 Live Sales Kill Cards

Never get caught off guard on a sales call again. The Live Sales Kill Cards Co-Pilot converts real-time competitor telemetry and AI battlecards into 15-second verbatim counter-scripts, customer migration triggers, pricing defenses, and landmine questions — strictly personalized per monitored competitor target.

⚡ Key Capabilities

⏱️ Live Call HUD Timer — Active call duration telemetry with start, pause, and reset controls
🎯 Target Competitor Filtering — Instant dropdown filtering across all active workspace competitors
⚡ 15-Second Verbatim Counter-Scripts — Numbered, out-loud responses for reps during live calls
💣 Landmine Questions — Trap questions reps ask prospects to expose competitor vulnerabilities
🛡️ Verified Evidence & Proof — Scraped domain evidence and price point comparison benchmarks
1️⃣ One-Click Script Copy & Oracle Escalation — Copy scripts instantly or push queries directly to MIRA Oracle AI

🔄 Dynamic Competitor Engine

Unlike legacy static cheat-sheets, MIRA Kill Cards are 100% dynamic:

Element	Source Data
🎯 Objection Cards	Synthesized from competitor `objection_handling` array
⚡ Displacement Cards	Generated from `switching_triggers` & `elevator_pitch`
💰 Pricing ROI Defense	Derived from `pricing_comparison` & `why_we_win`
🛡️ Evidence Snippets	Extracted from recent competitor web scrapes & intel signals
🔄 Auto-Generation	One-click AI synthesis for new competitor targets

🤖 Model Configurations

Component	Model	Runtime	Disk	RAM	Latency
🔍 Embeddings	`Xenova/all-MiniLM-L6-v2`	ONNX (JavaScript)	~90 MB	~80 MB	< 0.5s
🧠 LLM (Cloud)	`gemini-3.6-flash`	Google API	—	—	< 1.5s
🧠 LLM (Local)	`Qwen2.5-0.5B-Instruct`	llama-cli (GGUF Q4_K_M)	~382 MB	~350 MB	7–15s
🔧 Fallback	Rule-based heuristic	Node.js keyword matching	—	—	< 1ms
🔮 Oracle	`gemini-3.6-flash`	Google API (contextual)	—	—	< 2s
⚔️ War Room	`gemini-3.6-flash`	Google API (game-theory)	—	—	< 3s

💡 Configure your Gemini API key and other integration credentials via the dashboard Settings tab — no .env changes needed.

🚀 Quick Start

Prerequisites

✅ Node.js    v20+
✅ NPM        v10+
✅ OS         macOS / Linux / Windows (WSL)

1️⃣ Clone & Install

git clone https://github.com/NitheshK4/Autonomous-Competitor-Intelligence-Engine.git
cd Autonomous-Competitor-Intelligence-Engine

npm install
npm run install:all

💡 No Python required. All ML inference runs natively in Node.js.

2️⃣ Configure

cp .env.example .env
# Only PORT and NODE_ENV are needed — all API keys (Gemini, Slack,
# Notion, Airtable, SMTP) are configured via the dashboard Settings tab.

3️⃣ Launch

npm run dev

Service	URL
🖥️ Backend API	`http://localhost:3000`
🎨 Dashboard	`http://localhost:5173`

4️⃣ Test

npm test

Validates: Scraping → Semantic Detection → LLM Inference → Battlecards → CRM Sync

🔐 Environment Variables

Create a .env in the project root (copy from .env.example). Only PORT is strictly required — all integration credentials are managed via the dashboard Settings tab and persisted in SQLite.

Variable	Required	Default	Description
`PORT`	✅	`3000`	Express server port
`NODE_ENV`	—	`development`	`production` for optimized builds
`PUPPETEER_EXECUTABLE_PATH`	—	bundled	Override Chrome binary (e.g. `/usr/bin/google-chrome-stable`)

Note: GEMINI_API_KEY, SLACK_WEBHOOK_URL, Notion, Airtable, and SMTP credentials are all configured through the dashboard Settings tab — no .env entry needed.

📡 API Reference

Dashboard endpoints use X-Workspace-Id header (defaults to default). Extension endpoints require Authorization: Bearer <api_key>.

📋 Dashboard Endpoints

Method	Endpoint	Description
`GET`	`/health`	Health check
`GET`	`/api/profile`	Get business profile
`POST`	`/api/profile`	Save business profile
`GET`	`/api/competitors`	List competitors
`POST`	`/api/competitors`	Add competitor
`GET`	`/api/competitors/:id`	Competitor details
`PUT`	`/api/competitors/:id`	Update competitor
`DELETE`	`/api/competitors/:id`	Remove competitor
`POST`	`/api/competitors/:id/check`	Trigger immediate scrape
`GET`	`/api/intelligence`	Get intel cards
`PUT`	`/api/intelligence/:id`	Update card (read/unread)
`POST`	`/api/intelligence/read-all`	Mark all read
`POST`	`/api/intelligence/:id/retry`	Retry failed CRM sync
`GET`	`/api/settings`	Get settings
`POST`	`/api/settings`	Save settings
`POST`	`/api/settings/test-email`	Test SMTP
`GET`	`/api/debug-status`	Pipeline debug info

🃏 Battlecard Endpoints

Method	Endpoint	Description
`GET`	`/api/battlecards`	List all battlecards
`GET`	`/api/battlecards/:competitorId`	Get battlecard for competitor
`POST`	`/api/battlecards/:competitorId/generate`	Generate / refresh battlecard
`DELETE`	`/api/battlecards/:competitorId`	Delete battlecard

🔮 AI Strategy Co-Pilot & War Room Endpoints

Method	Endpoint	Description
`POST`	`/api/oracle/chat`	Send message to MIRA Oracle (AI Co-Pilot)
`POST`	`/api/warroom/simulate`	Run War Room "What-If" simulation

Oracle Request Body:

{
  "message": "How should we counter Competitor X's new AI feature?",
  "history": [
    { "role": "user", "text": "Previous question..." },
    { "role": "assistant", "text": "Previous answer..." }
  ]
}

War Room Request Body:

{
  "move": "Drop pricing by 30% across all tiers"
}

War Room Response Structure:

{
  "scenario": "Drop pricing by 30% across all tiers",
  "risk_score": 8,
  "risk_level": "HIGH",
  "market_impact_summary": "Aggressive pricing will trigger...",
  "competitor_responses": [
    {
      "competitor_name": "Rival A",
      "predicted_action": "Match pricing + launch ad blitz",
      "likelihood_pct": 85,
      "timeframe": "1-2 Weeks",
      "threat_severity": "High"
    }
  ],
  "counter_offensive_playbook": [
    {
      "step": 1,
      "phase": "Immediate (Days 1-7)",
      "action": "Distribute battlecards to SDRs",
      "details": "Equip sales team before competitors adjust messaging"
    }
  ],
  "strategic_verdict": "PROCEED WITH CAUTION"
}

🧩 Extension Endpoints

Method	Endpoint	Description
`GET`	`/api/extension/status`	Connection check
`GET`	`/api/extension/unread-count`	Unread card count (badge)
`POST`	`/api/extension/add-competitor`	Register competitor from browser

🔌 Integration Guides

📓 Notion CRM

Create an integration at Notion My Integrations
Create a database with these properties:

Property	Type
Title	`Title`
Competitor Name	`Select`
URL	`URL`
Category	`Select`
Impact Score	`Number`
Recommended Action	`Text`
Summary	`Text`
Justification	`Text`
Screenshot URL	`URL`

Connect integration to database → ... → Connect to
Copy Database ID from page URL
Enter credentials in dashboard Settings

📊 Airtable CRM

Generate a PAT with data.records:write at Airtable Tokens
Create a Base → Table named Competitor Intel
Enter Base ID, Table name, and Token in dashboard Settings

💬 Slack

Create an Incoming Webhook in your workspace
Paste the webhook URL in dashboard Settings
Impact Score ≥ 8 → instant Slack alert ⚡

📧 Email (SMTP)

For Gmail: generate an App Password at security.google.com
Configure in Settings: smtp.gmail.com / Port 587
Click Test SMTP Connection to verify
Periodic digests arrive automatically 📬

🛠️ Tech Stack

Layer	Technologies
🖥️ Backend	Node.js 20, Express 4, SQLite (`sqlite3`), UUID
🎨 Frontend	React 18, Vite 5, Tailwind CSS 3, Lucide React, PHP (SSR Dashboard)
🐘 PHP SDK & SSR	`MiraClient.php`, Server-Side Rendered Dashboard & CSV Exporter
🕸️ Scraping	Axios, Cheerio, Puppeteer (headless Chromium)
🧠 AI / ML	HuggingFace Transformers (ONNX), Gemini 3.6 Flash, Qwen GGUF
🔮 Strategy AI	MIRA Oracle (Co-Pilot), War Room Simulator (Game Theory)
🃏 Battlecards	LLM-powered auto-generation with CRM sync
🔌 Integrations	Notion SDK, Airtable REST, Slack Webhooks, Nodemailer
🧩 Extension	Chrome Manifest V3, Service Workers
🔧 Tooling	Concurrently, Nodemon, Docker

📁 Project Structure

📦 MIRA — Autonomous Competitor Intelligence Engine
│
├── 📂 php/                        # Complete PHP Subsystem & Intelligence Portal
│   ├── 📂 src/
│   │   ├── 📂 Console/            # PHP CLI Command Runners
│   │   │   └── MiraCliCommand.php
│   │   ├── 📂 Core/               # Configuration & Environment Manager
│   │   │   └── Config.php
│   │   └── 📂 Services/           # War Room Simulator & Battlecard Services
│   │       ├── BattlecardService.php
│   │       └── WarRoomSimulator.php
│   ├── 📂 tests/                  # PHP Subsystem Unit Test Suite
│   │   └── MiraClientTest.php
│   ├── MiraClient.php             # Object-Oriented PHP SDK API Client
│   ├── mira.php                   # Executable PHP CLI tool
│   ├── index.php                  # Server-Side Rendered PHP Dashboard
│   ├── export.php                 # PHP CSV/JSON Intelligence Exporter
│   ├── composer.json              # PSR-4 Package Manifest
│   └── README.md                  # PHP Subsystem Documentation
│
├── 📂 client/                     # React + Vite + Tailwind CSS dashboard
│   ├── src/
│   │   ├── App.jsx                # Root dashboard application (all views)
│   │   ├── index.css              # Global + Tailwind styles
│   │   ├── main.jsx               # React entry point
│   │   └── components/
│   │       ├── BattlecardsView.jsx   # 🃏 Auto-generated battlecards UI
│   │       ├── CommandPalette.jsx    # 🔍 Cmd+K global search & quick actions
│   │       ├── Sidebar.jsx           # 📱 Navigation sidebar
│   │       ├── SkeletonLoader.jsx    # 💀 Loading state placeholders
│   │       ├── StrategyCopilotModal.jsx # 🔮 MIRA Oracle AI chat modal
│   │       ├── TopBar.jsx            # 🔝 Top navigation bar
│   │       ├── VisualDiffModal.jsx   # 🗂️ Side-by-side content diff viewer
│   │       └── WarRoomView.jsx       # ⚔️ War Room scenario simulator
│   ├── index.html
│   ├── tailwind.config.js         # Tailwind theme configuration
│   └── vite.config.js             # Vite config with API proxy to :3000
│
├── 📂 server/                     # Node.js + Express backend
│   └── src/
│       ├── index.js               # Express server, routes, scheduler
│       ├── scraper.js             # Double-engine scraper (Axios + Puppeteer)
│       ├── detector.js            # Semantic change detection (ONNX)
│       ├── llm.js                 # LLM inference + Oracle + War Room + Battlecards
│       ├── crm.js                 # Notion & Airtable CRM adapter
│       ├── queue.js               # Sequential processing queue
│       ├── slack.js               # Slack webhook alerts
│       ├── mailer.js              # SMTP email digest
│       ├── enrichment.js          # DNS & header enrichment
│       ├── db.js                  # SQLite database layer
│       └── verify-test.js         # Integration tests
│
├── 📂 extension/                  # Chrome Extension (MV3)
│   ├── manifest.json
│   ├── popup.html / popup.js      # One-click competitor registration
│   ├── options.html / options.js  # Server URL & API key config
│   ├── background.js              # Badge polling service worker
│   └── icon*.png                  # Icons (16, 48, 128)
│
├── 📂 docs/                       # Documentation & screenshots
│   ├── images/
│   ├── railway_deployment_guide.md
│   └── walkthrough.md
├── data_exporter.py               # CSV/Markdown export utility
├── Dockerfile                     # Production container (Node 20 + Chrome)
├── package.json                   # Workspace orchestrator
└── .env.example                   # Environment template

🐳 Deployment

Railway (Recommended)

The Dockerfile installs Google Chrome Stable, downloads ONNX models, builds the Vite client, and starts Express — all automatically.

Create a project on Railway
Link your GitHub repo
Set env vars: PORT=3000, NODE_ENV=production
Deploy 🚀

All integration credentials (Gemini, Slack, Notion, etc.) are configured via the dashboard Settings after deploy — no extra env vars needed.

Docker (Manual)

docker build -t mira .
docker run -p 3000:3000 --env-file .env mira

⚠️ Known Limitations

Issue	Details
⏳ Cold Start	First run downloads models (~90 MB ONNX + ~382 MB GGUF). Cached after.
🤖 Anti-Bot	Some sites block headless scrapers → graceful Axios fallback.
📋 Sequential Queue	One-at-a-time processing to stay under 512 MB RAM.
🔄 Rate Limits	Gemini may throttle → auto-fallback to GGUF → heuristics.
🔮 Oracle Fallback	Without Gemini key, Oracle provides data-driven reports (not conversational AI).
⚔️ War Room Fallback	Without Gemini key, simulations use deterministic rule-based modeling.

🤝 Contributing

Fork the repository
Branch → git checkout -b feature/your-feature
Commit → git commit -m 'Add your feature'
Push → git push origin feature/your-feature
PR → Open a Pull Request

Please ensure npm test passes before submitting.

📜 License

Licensed under the MIT License — see LICENSE for details.

_{🔮 MIRA Oracle • ⚔️ War Room Simulator • 🃏 Auto-Battlecards • 🧠 Semantic Intelligence • 🕸️ Smart Scraping}

Name		Name	Last commit message	Last commit date
Latest commit History 141 Commits
.agents		.agents
client		client
docs		docs
extension		extension
go		go
php		php
python		python
server		server
.env.example		.env.example
.gitignore		.gitignore
Dockerfile		Dockerfile
Dockerfile.php		Dockerfile.php
README.md		README.md
data_exporter.py		data_exporter.py
docker-compose.yml		docker-compose.yml
package-lock.json		package-lock.json
package.json		package.json

Folders and files

Latest commit

History

Repository files navigation