Skip to content
BranchesTags

Repository files navigation

🏛️ Atlas — AI Financial Assistant & Executive Analyst

Autonomous Financial Intelligence Delivered Exclusively via Telegram (@Atlas39bot)

Telegram Bot Demo Video Python 3.9+ FastAPI Groq Llama 3.3 Groq Whisper SEC EDGAR License: MIT

📌 What is Atlas?

Atlas is an autonomous AI financial assistant and executive analyst built for equity researchers, venture investors, founders, and finance professionals. Operating exclusively on Telegram (@Atlas39bot), Atlas communicates like an experienced senior colleague—eliminating rigid slash commands, buttons, and forms in favor of natural human conversation.

                           User on Telegram (@Atlas39bot)
                                         │
        ┌────────────────────────────────┼────────────────────────────────┐
        ▼                                ▼                                ▼
   [Text Query]                   [Voice Message]                  [Image / PDF]
"Compare MSFT vs GOOGL"         🎙️ Transcribed with             🖼️ Vision OCR &
"Explain why NVDA moved"          Groq Whisper v3                 10-K Summaries
        │                                │                                │
        └────────────────────────────────┼────────────────────────────────┘
                                         │
                                         ▼
                      [Atlas Senior Analyst Engine]
                     • Live Finnhub institutional data
                     • "Why It Matters" thesis generation
                     • Continuous SQLite + Redis memory
                                         │
                                         ▼
                        Direct, High-Signal Telegram Reply

🌟 Core Highlights

1. 🗣️ 100% Natural Conversation (Zero Commands)

  • No Slash Commands, Buttons, or Menus: Ask questions naturally ("What are the biggest market-moving events today?", "Compare Microsoft and Google on valuation and cloud margins", "Summarize Apple's latest earnings in 5 points").
  • Smart Ambiguity Clarification: When asked open-ended queries like "Tell me about Apple", Atlas shares a 2-bullet orientation and proactively asks clarifying questions before assuming intent.

2. ☀️ Proactive Intelligence with Noise Suppression

  • Morning Brief (08:00 Local Time): Macro rate backdrop, sector rotation, watchlist movers, and "Why It Matters" catalyst breakdown.
  • Evening Summary (18:00 Local Time): Session scorecard, breadth, and next-day setup.
  • Intelligent Noise Suppression: Stays completely silent during flat sessions without material news (Quality > Frequency).

3. 🎙️ Multi-Modal Intelligence (Voice, Vision & 10-K Reports)

  • Voice Notes: Speak naturally; audio is transcribed via Groq Whisper v3 and answered conversationally.
  • Chart Vision OCR: Send chart screenshots or balance sheet tables for support/resistance and margin analysis.
  • 10-K & Report Synthesis: Drop PDF annual reports or earnings decks for 5-point executive summaries and categorized risk factor isolation.

4. 📈 Real-Time Multi-Asset Market Data

  • US Equities: Real-time quotes, P/E ratios, gross margins, and market caps (NVDA, AAPL, MSFT, TSLA, AMD).
  • Indian Equities & Indices: Live quotes in Indian Rupees (₹) for RELIANCE, TCS, NIFTY 50, SENSEX.
  • Crypto, Commodities & Macro: Live 24/7 Bitcoin, Ethereum, Gold, Crude Oil, DXY, and US 10Y Yield.

5. ⚡ Conversational Action Execution & Persistent Memory

  • Background Actions: Saying "Track Tesla and alert me of SEC filings" or "Schedule a meeting tomorrow at 10 AM on this report" automatically updates your watchlist and calendar reminders.
  • Two-Tier Memory: Caches active dialogue in Redis L1 for instant recall and persists profile facts and holdings in SQLite WAL L2.

📌 Welcoming Conversational Onboarding

Atlas begins with a natural conversation—not a lengthy registration form.

  1. First Interaction: Greets the user with a live market backdrop and asks 2-3 simple questions to understand their role (Investor, Equity Analyst, Founder, Finance Pro, Student) and focus sectors (AI, Crypto, Biotech, Macro).
  2. Natural Integrations: Organically introduces Google Sheets, Google Calendar, and Gmail when relevant to the workflow.
  3. 100% Skippable: Users can ask any direct market question immediately to bypass onboarding.
  4. Progressive Learning: Silently remembers preferences, watchlist tickers, and briefing times across sessions.

🤖 Multi-Model Architecture

Modality / Task Primary Model Provider Key Purpose
Deep Financial Reasoning llama-3.3-70b-versatile Groq Valuation matrices, DCF models, catalysts & 10-K risks
Fast Conversational Q&A llama-3.1-8b-instant Groq Sub-second low-latency financial dialogue
Voice Audio Transcription whisper-large-v3 Groq High-accuracy voice note transcription
Chart Vision & OCR gemma-4-26b-a4b-it:free OpenRouter Technical levels, chart patterns & table OCR
Failover Fallback gpt-oss-120b Cerebras Automatic failover during rate limits

📁 Clean & Modular Architecture

atlas/
├── backend/
│   ├── agent/
│   │   ├── core.py              # Conversational analyst loop & ambiguity handler
│   │   └── prompts.py           # Senior analyst persona & formatting guidelines
│   ├── bot/
│   │   └── handlers.py          # Telegram handlers (Text, Voice, Photos, Documents)
│   ├── db/
│   │   ├── models.py            # SQLAlchemy models (Users, Holdings, Documents, Alerts)
│   │   └── repo.py              # SQLite repository & user memory fact store
│   ├── scheduler/
│   │   └── jobs.py              # Proactive APScheduler (08:00 Briefs, SEC Watchers)
│   ├── services/
│   │   ├── doc_intelligence.py  # 10-K 5-point summaries & comparative delta matrices
│   │   ├── edgar.py             # Official SEC EDGAR Form 10-K, 10-Q & 8-K fetcher
│   │   ├── llm_router.py        # Multi-model LLM router (Groq, Cerebras, OpenRouter)
│   │   ├── market.py            # Live quotes, ASCII sparklines & valuation multiples
│   │   ├── media.py             # Voice transcription (Whisper v3) & Vision OCR
│   │   ├── proactive.py         # Briefing generator & noise suppression engine
│   │   └── redis_service.py     # L1 Redis cache & rate-limiting service
│   └── main.py                  # FastAPI server & Telegram Bot long-polling daemon
├── requirements.txt             # Python dependencies
├── Procfile                     # Deployment entrypoint
└── README.md                    # System documentation

🚀 Quickstart Guide

1. Clone & Setup

git clone https://github.com/Aj2280/Atlas-AI-Financial-Assistant-.git
cd Atlas-AI-Financial-Assistant-

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

2. Set Environment Variables

export TELEGRAM_BOT_TOKEN="your_telegram_bot_token"
export GROQ_API_KEY="your_groq_api_key"
export FINNHUB_API_KEY="your_finnhub_api_key"
export OPENROUTER_API_KEY="your_openrouter_api_key"  # (Optional, for Vision)
export CEREBRAS_API_KEY="your_cerebras_api_key"      # (Optional, for fallback)

3. Run Atlas

PYTHONPATH=. python3 backend/main.py

Atlas will start long-polling Telegram immediately on @Atlas39bot with all proactive schedulers active.

💬 Real-World Interaction Examples

Scenario Example User Prompt
Macro Outlook "What are the biggest market-moving events I should know about today?"
Peer Comparison "Compare Microsoft and Google from an investment perspective on valuation and cloud margins."
Earnings Breakdown "Summarize Apple's latest earnings call in five key points."
10-K Risk Analysis "Analyze this annual report and highlight the biggest regulatory and supply chain risks."
Price Movement Driver "Explain why Nvidia moved today and what the transmission mechanism is."
SEC Watcher "Track Tesla and notify me whenever there's a major announcement or Form 8-K filing."
Custom Briefing "Create a daily 8 AM briefing covering AI and semiconductor stocks."
Meeting Scheduling "Schedule a meeting with my team tomorrow at 10 AM to discuss this earnings report."

📜 License

Distributed under the MIT License. See LICENSE for details.

Atlas: Institutional AI Financial Intelligence 🏛️

About

Atlas — Autonomous AI Financial Assistant on Telegram .

Resources

Readme
Activity

Stars

2 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages