stocksensei 0.1.2

AI-powered terminal stock analyzer with pluggable LLM providers

StockSensei 📈

StockSensei is an intelligent, AI-powered terminal CLI application that acts as your personal expert financial analyst — running entirely inside your terminal. Built with Python, LangChain, and yfinance, it lets you query real-time stock data, compare companies, render structured terminal visuals, and read market news using plain natural language. No browser. No dashboard. Just your terminal.

---

✨ Features

• 🖥️ Runs Entirely in Your Terminal — A first-class command-line experience built for developers and traders who live in the terminal. Launch it globally with one command from any directory.
• 💬 Natural Language Interaction — Ask questions the way you think: "Is NVDA a better buy than AAPL right now?" or "Show me Tesla's trend over the last 3 months."
• 📊 Structured Visual Blocks — Cards, bars, tables, sparklines, and news lists are rendered from a structured JSON output contract — not fragile model-authored markdown.
• 📋 Deterministic Terminal Visuals — Comparison tables and stock snapshots render consistently with rich — no more ugly pipe characters or misaligned columns.
• 📡 Real-Time Market Data — Live prices, daily % changes, market caps, P/E ratios, 52-week highs/lows, and more via yfinance.
• 📰 News Integration — Fetch the latest headlines for any stock or company.
• 🧠 Conversational Memory — The agent remembers context within your session, so follow-up questions just work.
• 🧱 JSON Output Contract — The AI returns a structured response schema with a message plus ordered UI blocks, making rendering and fallback behavior safe and predictable.
• ⚡ Animated Status Feedback — StockSensei shows a single animated status line (with tool-specific messages) while it works — no cluttered logs.
• ⌨️ Slash-Command Completion — In interactive terminals, typing / opens a styled autocomplete dropdown for /help, /models, /clear, and /quit, powered by prompt-toolkit.
• 🤖 Multi-Provider AI Support — Choose from OpenAI, Anthropic, Gemini, Groq, DeepSeek, OpenRouter, Ollama, or any custom OpenAI-compatible endpoint. Switch providers mid-session with /models.
• 🔐 Zero-Config Setup — On first launch, StockSensei walks you through picking a provider and saving your API key globally — so you never have to set it again.

---

📸 Visual Examples

Here are a few examples of StockSensei's terminal UI in action:

---

🛠️ Tech Stack

Layer	Technology	Purpose
Language	Python >= 3.13	Core runtime
Financial Data	yfinance	Real-time prices, OHLC history, news, company info
AI Framework	LangChain	Tool-calling agent and structured output
LLM Providers	OpenAI, Anthropic, Gemini, Groq, DeepSeek, OpenRouter, Ollama	Swappable AI backends via /models
Agent State	LangGraph	Conversation memory and session checkpointing
Terminal UI	Rich	Cards, tables, panels, bars, and live status
Interactive Input	prompt-toolkit	Slash-command completion and terminal prompt UX
Environment	python-dotenv	Secure .env loading for local development
Package Manager	uv	Fast dependency resolution and global CLI installation

---

⚙️ Prerequisites

• Python 3.13+ — required regardless of install method.
• pip or uv — either works for installation (see below).

---

🚀 Installation & Setup

Using pip (standard)

pip install stocksensei

Using uv (faster)

If you prefer uv, install it first if you don't have it:

Mac/Linux:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows (PowerShell):

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Then install as a global tool:

uv tool install stocksensei

---

Then run it from anywhere:

stocksensei

On first launch, StockSensei will walk you through selecting an AI provider and entering your API key — saved permanently so you only ever do this once.

---

Updating to the Latest Version

With pip:

pip install --upgrade stocksensei

With uv:

uv tool upgrade stocksensei

(For the absolute latest before a PyPI release: uv tool install git+https://github.com/lenminh002/StockSensei.git)

---

Developer Setup (Local Cloning) To modify the code or contribute:

git clone https://github.com/lenminh002/StockSensei.git
cd StockSensei
uv sync       # or: pip install -e .
uv run main.py

---

💡 Usage

Once running, just type your questions in plain English:

You: nvidia vs apple
StockSensei: [Renders a comparison table with price, P/E, market cap, 52w range]

You: show me nvda's chart for the last 3 months
StockSensei: [Draws a sparkline trend chart with annotated date range]

You: what's the latest news on tesla?
StockSensei: [Lists the 10 most recent headlines]

You: price of nvda
StockSensei: [Shows a snapshot card with live price and daily change]

Type exit, quit, q, or /quit to close the app.

Commands

/models   — switch AI provider or model
/clear    — clear conversation history and reset the terminal
/help     — show available commands and example prompts
/quit     — exit StockSensei

In an interactive terminal, typing / opens an autocomplete dropdown with all available commands and their descriptions.

Switching Providers & Models

Type /models at any time during a session to switch your AI provider or model:

You: /models

Current: openai / gpt-4.1-mini
Select provider:
  1. openai  (default: gpt-4.1-mini)
  2. anthropic  (default: claude-sonnet-4-6)
  3. groq  (default: llama-3.3-70b-versatile)
  4. + Add new provider
Your choice: 2

Select model:
  1. claude-opus-4-7
  2. claude-sonnet-4-6
  3. claude-haiku-4-5-20251001
Your choice: 1

✓ Switched to anthropic / claude-opus-4-7

Your selection is saved to ~/.stocksensei_config.json and remembered across sessions.

---

🧱 Structured Output Contract

StockSensei uses a strict JSON output contract. The AI always returns a structured response shaped like:

{
  "message": "Short explanation for the user.",
  "blocks": [
    {
      "type": "metric_card",
      "title": "AAPL snapshot",
      "subtitle": "Apple Inc.",
      "items": [
        {"label": "Price", "value": "$201.68", "tone": "bright_cyan"},
        {"label": "Daily change", "value": "+1.23%", "tone": "bright_green"},
        {"label": "Market cap", "value": "$3.01T", "tone": "bright_magenta"},
        {"label": "P/E", "value": "33.42", "tone": "bright_yellow"}
      ]
    },
    {
      "type": "range_bar",
      "title": "AAPL 52-week range",
      "minimum_label": "Low $170",
      "maximum_label": "High $220",
      "current_label": "Current $201.68",
      "position": 0.73
    }
  ]
}

Supported block types: text, metric_card, table, barchart, range_bar, sparkline, news.

This contract means output is deterministic, safe to render, and consistent regardless of which AI provider is active.

---

🤖 Supported Providers

Provider	Models
OpenAI	gpt-5.4, gpt-5.4-mini, o3, o4-mini, gpt-4o, gpt-4o-mini
Anthropic	claude-opus-4-7, claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5-20251001
Gemini	gemini-3.1-pro-preview, gemini-3-flash-preview, gemini-3.1-flash-lite-preview, gemini-3-deep-think, gemini-2.5-pro, gemini-2.5-flash
Groq	llama-3.3-70b-versatile, llama-3.1-8b-instant, mixtral-8x7b-32768
DeepSeek	deepseek-chat, deepseek-reasoner
OpenRouter	openai/gpt-4.1-mini, anthropic/claude-opus-4-7, google/gemini-2.5-pro, meta-llama/llama-3.3-70b-instruct
Ollama	llama3.2, qwen2.5, mistral
Custom	Any OpenAI-compatible endpoint

---

📝 Notes

• Config file: Provider settings and API keys are stored in ~/.stocksensei_config.json.
• Tool architecture: Market-data tools return clean structured data; visual builder tools return render-ready block payloads; terminal rendering is handled in the CLI layer.
• Structured rendering: The CLI validates the JSON response schema and renders supported block types — text, metric cards, tables, bar charts, range bars, sparklines, and news lists.
• Fallback handling: If the AI output cannot be parsed as a structured response, StockSensei safely falls back to a plain text block rather than crashing.
• Cross-Platform: Works on macOS, Linux, and Windows (PowerShell).

---

📄 License

This project is licensed under the MIT License.

© 2025 Minh Nguyen Le, Minh Le Hoang

---

🗑️ Uninstall

1. Uninstall the CLI tool:

With pip:

pip uninstall stocksensei

With uv:

uv tool uninstall stocksensei

2. Remove saved config (optional — only if you want a full clean removal):

Mac/Linux:

rm ~/.stocksensei_config.json

Windows (PowerShell):

Remove-Item "$HOME\.stocksensei_config.json"