AIStat is a lightweight, local-first Chromium extension engineered to measure, analyze, and aggregate user interactions across major Artificial Intelligence conversation platforms. Built on Chrome Extensions Manifest V3, AIStat provides telemetry on message volume, platform distribution, hourly activity density, and historical usage trends with a strict zero-cloud privacy guarantee.
AIStat operates using a multi-tiered execution model that strictly separates content inspection, event arbitration, persistent storage, and analytical presentation.
flowchart TD
subgraph BrowserContext["Target Web Pages (ChatGPT, Claude, Gemini, DeepSeek, Perplexity)"]
A["Main World Execution Context"] -->|"window.fetch / XHR Interception"| B["Network Interceptor\n(network-interceptor.js)"]
B -->|"window.postMessage (AISTAT_NET_INTERCEPT)"| C["Isolated Extension Context\n(tracker.js)"]
D["DOM Event Listeners (Enter Key / Submit Buttons)"] -->|"Direct Event Fallback"| C
end
subgraph ExtensionCore["Extension Service Worker Layer"]
C -->|"chrome.runtime.sendMessage (RECORD_PROMPT)"| E["Background Service Worker\n(service-worker.js)"]
E -->|"Central Debounce Lock\n(platformLastRecordTime)"| F["Rate Limit & Deduplication Engine"]
G["chrome.alarms (Periodic Refresh)"] -->|"Alarm Trigger"| E
end
subgraph StorageLayer["Local-First Persistence"]
F -->|"Read / Write Operations"| H["chrome.storage.local\n(StatsStorage Interface)"]
end
subgraph PresentationLayer["User Interface Layer"]
H -->|"Summary Stats Aggregation"| I["Action Popup\n(popup.html / popup.js)"]
H -->|"Historical Trends & Export"| J["Analytics Dashboard\n(dashboard.html / dashboard.js)"]
E -->|"chrome.action.setBadgeText"| K["Toolbar Action Badge"]
end
- Dual-Engine Submission Detection: Combines network API interception in the main DOM execution world with targeted DOM event listeners in the isolated world to ensure 100% prompt capture accuracy across single-page applications.
- Cross-Frame Event Deduplication: Employs timestamp hashing and per-platform lock intervals to prevent duplicate counting during multi-frame page loads and rapid-fire network retries.
- 24-Hour Activity Density Analysis: Aggregates interaction timestamps into hourly distribution histograms to identify peak productivity windows.
- Multi-Window Rate Limiting Support: Computes rolling hourly and daily message budgets to assist users managing tiered API quotas and subscription limits.
- Granular Data Portability: Complete export pipeline supporting structured JSON backups and flattened RFC-4180 compliant CSV exports.
- 100% On-Device Isolation: Zero remote servers, zero external analytics telemetry, zero third-party script tags.
The sequence diagram below illustrates the complete lifecycle of a prompt capture event, from DOM trigger to local persistence and UI synchronization.
sequenceDiagram
autonumber
actor User
participant Page as Web Application (DOM)
participant Interceptor as Main World Interceptor
participant Tracker as Isolated Content Script
participant SW as Background Service Worker
participant Storage as chrome.storage.local
participant UI as Popup / Dashboard
User->>Page: Submits prompt (Enter key or click)
Page->>Interceptor: Dispatches fetch() / XHR POST
Note over Interceptor: Evaluates endpoint regex against platform rules
Interceptor->>Tracker: window.postMessage("AISTAT_NET_INTERCEPT")
Tracker->>SW: chrome.runtime.sendMessage({ type: "RECORD_PROMPT", platform })
SW->>SW: Check platformLastRecordTime lock (4000ms debounce)
SW->>Storage: StatsStorage.recordPrompt(platform, timestamp)
Storage-->>SW: Updated dailyLogs schema
SW->>SW: Update extension action badge count
opt Active Popup or Dashboard
UI->>Storage: StatsStorage.getSummaryStats()
Storage-->>UI: Real-time aggregated metrics
UI->>UI: Re-render charts and platform meters
end
All telemetry records are structured inside chrome.storage.local under a standardized hierarchical schema managed through the StatsStorage abstraction layer.
erDiagram
SETTINGS {
string badgeDisplay "message_count | none"
}
DAILY_LOG {
string dateKey PK "YYYY-MM-DD"
int messagesCount "Total daily count"
}
PLATFORM_COUNTS {
int chatgpt "ChatGPT count"
int claude "Claude count"
int gemini "Google Gemini count"
int deepseek "DeepSeek count"
int perplexity "Perplexity count"
}
HOURLY_DISTRIBUTION {
int hour_00 "00:00 - 00:59"
int hour_12 "12:00 - 12:59"
int hour_23 "23:00 - 23:59"
}
DAILY_LOG ||--|| PLATFORM_COUNTS : "contains platform breakdown"
DAILY_LOG ||--|| HOURLY_DISTRIBUTION : "contains hourly breakdown"
{
"settings": {
"badgeDisplay": "message_count"
},
"dailyLogs": {
"2026-08-18": {
"date": "2026-08-18",
"messagesCount": 42,
"platforms": {
"chatgpt": 18,
"claude": 14,
"gemini": 6,
"deepseek": 2,
"perplexity": 2
},
"hours": {
"9": 5,
"10": 12,
"14": 15,
"16": 10
}
}
}
}flowchart LR
subgraph UntrustedZone["Remote Network (Internet)"]
CloudServer["External Analytics / Ad Servers"]
LLMServers["AI API Endpoints\n(OpenAI, Anthropic, Google)"]
end
subgraph SecurityBoundary["Browser Client Boundary (Strictly Local)"]
direction TB
ContentScript["Content Script Sandbox"]
BackgroundWorker["Background Service Worker"]
LocalStorage["chrome.storage.local"]
end
LLMServers <-->|"User Chat Traffic (Untouched)"| ContentScript
ContentScript -->|"Metadata Only (Platform, Timestamp)"| BackgroundWorker
BackgroundWorker -->|"Local Persistence"| LocalStorage
LocalStorage -.->|"BLOCKED / ZERO ACCESS"| CloudServer
AIStat enforces an immutable security posture:
- No Payload Extraction: Prompt content, conversation transcripts, source code, and API tokens are never read, stored, or processed. Only platform identity and submission timestamps are recorded.
- No Remote Egress: The extension does not declare
webRequestor open external network sockets. - Strict Host Scoping: Host permissions are limited exclusively to official AI conversational endpoints.
| Platform | Domain Filter | Primary Detection Vector | Fallback Vector |
|---|---|---|---|
| ChatGPT | chatgpt.com, chat.openai.com |
POST /backend-api/conversation |
DOM Textarea keydown listener |
| Claude | claude.ai |
POST /api/organizations/.../chat_messages |
DOM Contenteditable submission button |
| Google Gemini | gemini.google.com |
POST .../_/BardChatUi/data/assistant.lamda.BardFrontendService/StreamGenerate |
Enter key on rich text editor |
| DeepSeek | chat.deepseek.com |
POST /api/v0/chat/completion |
DOM Textarea input handler |
| Perplexity | perplexity.ai, www.perplexity.ai |
POST /rest/queries, POST /rest/ask |
Search bar query interceptor |
ai_stat/
├── .github/
│ ├── ISSUE_TEMPLATE/
│ │ ├── bug_report.md
│ │ └── feature_request.md
│ ├── workflows/
│ │ └── validate.yml
│ └── pull_request_template.md
├── background/
│ └── service-worker.js # Manifest V3 background worker and badge controller
├── content-scripts/
│ ├── network-interceptor.js # Main-world fetch/XHR hook script
│ └── tracker.js # Isolated-world DOM and event bridge script
├── dashboard/
│ ├── dashboard.html # Full-screen analytics interface
│ ├── dashboard.css # Modern design system and CSS grid layout
│ └── dashboard.js # Analytics controller, SVG chart generators, exporters
├── popup/
│ ├── popup.html # Action popup user interface
│ ├── popup.css # Compact popup styling
│ └── popup.js # Popup live data binding controller
├── shared/
│ ├── constants.js # Platform definitions, color taxonomy, domain rules
│ ├── storage.js # Chrome storage abstraction layer and calculations
│ └── lucide.min.js # Lightweight vector iconography engine
├── icons/ # Extension asset icons (16px, 32px, 48px, 128px, 1024px)
├── .gitignore # Git exclusion rules
├── CHROMEWEBSTORE.md # Chrome Web Store technical specification
├── CONTRIBUTING.md # Developer contribution guidelines
├── LICENSE # MIT Open Source License
├── manifest.json # Manifest V3 extension configuration
├── README.md # Technical documentation and architecture guide
└── SECURITY.md # Security model and vulnerability disclosure policy
- Google Chrome version 114+ or any Chromium derivative (Brave, Microsoft Edge, Opera, Arc).
- Node.js 18+ (optional, for running local CI validation scripts).
-
Clone the repository to your local workstation:
git clone https://github.com/Sreeram5678/ai_stat.git cd ai_stat -
Open Google Chrome and navigate to the Extensions management console:
chrome://extensions -
Enable Developer mode via the switch located in the upper-right corner.
-
Click the Load unpacked button in the upper-left toolbar.
-
Select the
ai_statproject directory. -
The extension is now active. Open any supported AI platform (e.g., ChatGPT or Claude) and submit a prompt to verify counter increments.
AIStat provides two native export formats directly from the Analytics Dashboard:
- Structured JSON: Complete dump of all raw day-by-day telemetry, hourly distributions, and user preferences for backup and migration.
- RFC-4180 CSV: Normalized tabular export suitable for immediate ingestion into business intelligence tools, pandas, R, or spreadsheet applications.
Date,Total Messages,ChatGPT,Claude,Gemini,DeepSeek,Perplexity
2026-08-14,35,12,15,5,2,1
2026-08-15,48,20,18,6,3,1
2026-08-16,29,10,12,4,2,1
2026-08-17,54,22,20,8,3,1
2026-08-18,42,18,14,6,2,2Contributions to AIStat are welcome. Whether you want to add support for new AI platforms, build visual themes, or enhance data exporters:
- Quick Start Guide: Read How to Add a New AI Platform for a 3-step walkthrough.
- Contribution Workflow: Check CONTRIBUTING.md for full branch naming, code standards, and verification guidelines.
- Good First Issues: Explore open beginner-friendly issues on the Issue Tracker.
This project is licensed under the terms of the MIT License.