OpenTerminalUI

The open-source financial terminal for traders, researchers, and quant teams.

OpenTerminalUI is a self-hosted, full-stack financial terminal that combines real-time market data, institutional-grade charting, derivatives analytics, portfolio management, and quant research into a single platform. Built with a terminal-style shell interface inspired by Bloomberg and Refinitiv, it delivers professional-grade workflows to anyone with a browser.

Multi-market coverage across NSE, BSE, NYSE, NASDAQ, crypto, commodities, forex, bonds, ETFs, and mutual funds. 70+ technical indicators, multi-panel chart workstations, F&O option chains with live Greeks, backtesting with Model Lab, statistical arbitrage with Pair Trading Lab, Portfolio Lab and optimizer workflows, paper trading and trade journal, OMS / ops / data-quality consoles, saved views and launchpad workspaces, a tool-using AI research agent with multi-agent debate and Strategy Lab, and an extensible plugin system — all running on your own hardware.

Screenshots

Captured from the rebuilt Docker image running at http://localhost:8000. Account-backed screens are seeded before capture so portfolio, watchlist, paper trading, and journal views show populated data.

Complete Feature Gallery

Mission Control & Navigation

Mission Control	Launchpad

Market Dashboard	Account

Equity Research & Markets

Market View	Security Hub

India Security Hub	Financial Analysis

Chart Workstation	Multi-Timeframe

DOM	Time & Sales

Split Compare	Market Heatmap

Hotlists	Watchlist

Screener	Factor Dashboard

Relative Strength	Sector Rotation

Dividends	Insider Activity

Portfolio, Risk & Trading

Portfolio	Portfolio Lab

Portfolio Optimizer	Risk Dashboard

Correlation Dashboard	Cockpit

Paper Trading	Position Sizer

Trade Journal	Alerts

Quant Research & Backtesting

Backtesting	Model Lab

Model Governance	Algorithm Framework

Statistical Lab	Pair Trading

Futures & Options

Option Chain	Greeks

Futures	OI Analysis

Strategy Builder	PCR

Options Flow	F&O Heatmap

Expiry Calendar	Option Greeks Calculator

Cross-Asset & Macro

Commodities	Forex

Crypto	ETF Analytics

Mutual Funds	Bonds

Yield Curve	Bond Analytics

Economic Terminal	Data Quality

Intelligence, AI & Platform

News & Sentiment	Intelligence Timeline

AI Research Agent	Multi-Agent Debate

Strategy Lab Agent	Research Library

OMS Compliance	Ops Dashboard

Plugins	Settings

Saved Views

Workspace & Markets

Home / Mission Control — market context, AI Market Outlook, portfolio hub, system health, and the full feature launch grid.

Multi-panel chart workstation — a 6-chart grid with synchronized crosshairs, 70+ technical indicators, and drawing tools.

Full-screen market view (AAPL, NASDAQ) — candlestick price action with volume, multi-timeframe, and indicator overlays.

Security Hub for a US name (AAPL) — quotes, fundamentals, price chart, analysis tabs, and the AI Catalyst & Conviction panel.

Security Hub for an Indian name (RELIANCE, NSE) — the same workflow across markets, with India fundamentals and sector context.

Financial analysis — income statement, balance sheet, and cash-flow statements with multi-period trends.

Futures & Options (AAPL, US) — live option chain with Greeks, OI build-up, and PCR signals; the same workflow covers NSE F&O.

Cross-asset coverage — commodities, forex, crypto, bonds, ETFs, and mutual funds.

Research & Stock Picking

Advanced screener with query builder, custom formula engine, composite factor scores, and "why ranked" insights.

Factor Dashboard — multi-factor (Value / Momentum / Quality / Low-Vol) idea lists and ranked picks for US & Indian markets.

News & Sentiment with the AI Emotion Indicator powered by a local Gemma model via LM Studio.

Unified Intelligence Timeline — news, alerts, events, insider activity, earnings, and model signals in one feed.

Portfolio, Risk & Backtesting

Portfolio monitoring — holdings, movement & historical return, risk metrics, and AI Risk Assessment.

Cockpit Priority Stack — a ranked daily brief across portfolio risk, alerts, catalysts, movers, and model signals.

Risk dashboard with statistical risk metrics, factor/exposure heatmaps, and AI Risk Insights powered by Gemma.

Backtesting workspace with strategy presets, execution-profile modeling, performance summary, and AI analysis.

Model Lab — parameter sweeps, walk-forward validation, Monte Carlo robustness, and run leaderboards.

Portfolio Lab — multi-asset portfolio backtests, strategy blends, and correlation analysis.

Watchlists with live quotes, heatmap view, and one-click routing to charts, screener, and backtests.

AI Research Agent

The tool-using research agent (Ctrl/Cmd + J) — a screen-aware verdict on AAPL backed by a live snapshot card it fetched itself.

Multi-agent debate — an analyst team (fundamental / sentiment / technical) each returns an evidence-backed verdict, which feeds a bull-vs-bear debate and a portfolio-manager decision.

Strategy Lab — the agent proposes a strategy, backtests it, changes one variable to iterate, then runs out-of-sample validation and reports an honest verdict (here, "not a validated edge" at p = 0.25).

Features

Terminal Shell

GO Bar (Ctrl+G) — Bloomberg-style command bar with symbol lookup and route navigation
Command Palette (Ctrl+K) — fuzzy search across 25+ functions, tickers, and natural language queries
Function Keys (F1-F9) — rapid workspace switching with Bloomberg-style hotkeys
Ticker Tape — rolling market pulse with live quotes across exchanges
Theme Engine — Terminal Noir (default), classic, and light themes with custom accent support
Desktop & Mobile Layouts — responsive design with persistent workspace framing

Charting & Technical Analysis

Multi-Panel Workstation — up to 9 synchronized chart panels with crosshair linking
70+ Technical Indicators — SMA, EMA, RSI, MACD, Bollinger Bands, Keltner, Supertrend, ATR, VWAP, OBV, CMF, Stochastic, CCI, ADX, Donchian, and many more
Multi-Timeframe — 1m, 2m, 5m, 15m, 30m, 1h, 4h, 1D, 1W, 1M with extended hours toggle
Drawing Tools — persistent annotations with templates, save/restore
Volume Profile — VPOC + 70% value area overlay
Replay Mode — step through historical price action bar by bar
Comparison Overlays — multi-symbol normalized or raw price comparison
Alternative Charts — Renko, Kagi, Point & Figure, Line Break
Chart Export — PNG, SVG, and CSV data export
OpenScript — custom indicator scripting with script library

Equity Research & Security Hub

8-Tab Security Analysis — overview, financials, chart, news/sentiment, ownership, estimates, peers, ESG
Fundamental Metrics — P/E, P/B, ROE, ROA, dividend yield, earnings growth, debt ratios
Earnings Calendar — historical surprises, upcoming events, guidance tracking
Shareholding History — promoter/FII/DII/public breakdown with trend visualization
Analyst Estimates — consensus tracking, revisions, and target prices
Corporate Actions — splits, dividends, rights, bonuses timeline
Peer Comparison — relative valuation matrices across comparable companies
Insider Trading Monitor — recent insider trades, per-stock insider activity, top buyers/sellers leaderboard, and cluster-buy detection with minimum insider thresholds
Trade Journal — trade logging with equity curve, calendar heatmap, and performance statistics

Advanced Screener

Query Builder — custom filters with preset formulas and arithmetic operations
Custom Formula Engine — write, save, and share custom formulas with server-side evaluation, formula library with descriptions and categories
15+ Visualization Modes — tables with sparklines, sector treemaps, heatmaps, scatter plots, radar charts, box plots, bubble charts, waterfall charts, RRG quadrants, gauge dials, distribution histograms, stacked area, and comparison bars
Multi-Market Scanning — NSE, BSE, NYSE, NASDAQ with technical and fundamental overlays
Preset Management — save, load, share, and browse community screens
Score-Based Ranking — deterministic scoring with stable ordering and explainable setup detection

Insight-Driven Stock Picking

Multi-Factor Composite Scoring — cross-sectional, sector-relative Value / Momentum / Quality / Low-Volatility z-scores combined into a weighted composite rank
Ranked Idea Lists — top-quintile picks per market and sector for both US (NYSE/NASDAQ) and Indian (NSE/BSE) universes
Factor Dashboard — per-symbol factor radar, factor chips, and conviction scoring with a US/India market toggle
Catalyst & Conviction Engine — LLM-extracted sentiment and upcoming catalysts from NSE/BSE and SEC filings, surfaced in the Security Hub
Point-in-Time Fundamentals — as-reported fundamental history that removes look-ahead bias from factor and fundamental backtests
Why-Ranked Explanations — composite scores, factor chips, and plain-language rationale on screener rows, with one-click routing to chart and backtest

AI Research Agent

Conversational Console — a slide-over agent panel (toggle with Ctrl/Cmd + J) that researches and analyzes stocks on demand
Tool-Using Agentic Loop — the agent autonomously calls read-only tools — screener, full stock snapshot, multi-ticker compare, and research-knowledge-base search (RAG) — and reasons over the results
Screen-Aware Context — defaults to the stock you currently have open and your selected market, so "tell me about this stock" resolves to the right ticker/exchange without re-typing it
Multi-Agent Debate Mode — an analyst team (fundamental / sentiment / technical) feeds a bull-vs-bear debate that a portfolio manager resolves into a BUY / HOLD / SELL decision with a conviction score
Strategy Lab (idea → tested result loop) — a bounded, read-only research loop that proposes a strategy, backtests it, changes one variable to iterate toward a target metric, then runs mandatory out-of-sample validation (permutation + multi-window robustness) and reports an honest verdict — refusing to call a curve-fit result an edge. Flag-gated and capped on rounds and wall-clock
Beautifully Rendered Output — answers render as styled markdown (headings, tables, lists), stock snapshots as crafted cards (logo, price, valuation/quality/growth metrics), and debates as a phase stepper with bull/bear cards and a decision banner with conviction meter
Provider-Flexible — runs against OpenRouter, OpenAI, or a local LM Studio model, with an automatic free-model fallback chain and per-phase model routing
MCP Server — the read-only agent tools (screener, snapshot, compare, technicals, backtests, research search) are also exposed over the Model Context Protocol for use by external MCP clients
Read-Only & Resilient — the agent never places orders or mutates data, and degrades gracefully on rate limits, empty completions, or unavailable data sources

Futures & Options (F&O)

Option Chain — full contract listing with live Greeks (Delta, Gamma, Theta, Vega, Rho)
IV Analysis — historical and implied volatility tracking, term structure visualization
Strategy Builder — multi-leg construction for spreads, butterflies, straddles, strangles
OI Analysis — open interest trends, buildup patterns, strike-level concentration
PCR Tracking — put-call ratio monitoring with overbought/oversold signals
Heatmaps — IV/volume/OI heatmaps across the strike grid
Options Flow — unusual activity scanner with volume/OI ratios, premium tracking, heat scores, and bullish/bearish sentiment classification
Futures Analytics — term structure, basis analysis, contract specifications
Expiry Calendar — contract schedules with roll suggestions

Portfolio & Risk Management

Multi-Portfolio CRUD — holdings management with cost basis and transaction tracking
Allocation & Attribution — sector allocation charts, contributor/detractor analysis
Benchmark Overlay — compare against indices with relative performance metrics
Risk Engine — VaR (95%), CVaR, EWMA volatility, rolling correlation, PCA factor exposures
Factor Analytics — multi-factor exposure radar, attribution waterfall, rolling factor history, and factor return comparison across market, size, value, momentum, quality, and low-volatility factors
Stress Testing — 6 predefined macro scenarios (GFC 2008, COVID 2020, rate shock, INR depreciation, tech rotation, commodity spike), custom shock builder, Monte Carlo simulation, and historical event replay
Correlation Deep Dive — correlation matrix, rolling correlation with regime detection, hierarchical clustering with dendrogram, and cross-asset dependency visualization
Tax Lot Manager — cost basis tracking across tax lots
Dividend Tracker — income tracking with ex-date calendar
Paper Trading — virtual trading engine with realistic order fills, slippage modeling, and TCA analytics

Backtesting & Model Lab

16+ Strategy Templates — SMA/EMA crossover, mean reversion, breakout, RSI, MACD, Bollinger Bands, dual momentum, VWAP reversion, Awesome Oscillator, Heikin-Ashi, Parabolic SAR, Dual Thrust, shooting star reversal, and Bollinger W/M patterns
Pair Trading Lab — cointegration screening, hedge-ratio estimation, spread z-score diagnostics, half-life analysis, and mean-reversion trade simulations for statistical arbitrage workflows
Intraday & Daily Testing — 1m to monthly resolution with session-aware logic
Vectorized Engine — NumPy-based computation for fast large-dataset backtests
Realistic Execution — slippage, commission, partial fills, latency, and market impact simulation
Result Visualization — equity curves, drawdown charts, monthly return heatmaps, rolling Sharpe, 3D parameter surfaces, Monte Carlo paths, trade analysis
Walk-Forward Analysis — out-of-sample validation with sliding windows
Parameter Sweep — sensitivity analysis across hyperparameter ranges
Experiment Tracking — create, run, compare, and promote models through the Model Lab
Model Governance — version tracking with code/data hashing, promotion to paper trading
Monte Carlo Robustness — trade/return resampling with confidence cones, terminal-wealth distribution, and probability-of-profit
Liquidity-Aware Execution — fixed-bps, volume-weighted, and square-root market-impact slippage models with percent-of-volume caps
Strategy Tear-Sheets — standardized HTML reports with equity, drawdown, rolling Sharpe, monthly returns, and benchmark overlay
Run Leaderboards — sortable Model Lab / Portfolio Lab run comparison by Sharpe, CAGR, max drawdown, turnover, and stability

Portfolio Lab

Multi-Asset Backtesting — portfolio-level backtests with up to 200 assets
Weighting Modes — equal weight, volatility target, risk parity, momentum, market cap
Strategy Blends — combine up to 10 strategies with weighted sum returns
Rebalance Scheduling — weekly, monthly, quarterly, or custom frequency
Attribution Analysis — top contributors/detractors, worst drawdowns, rebalance log
Correlation Matrices — cross-asset cluster analysis

Cockpit, Workspaces & Intelligence

Cockpit Priority Stack — a ranked daily brief across portfolio risk, alerts, catalysts, news shocks, top movers, and model signals
Unified Intelligence Timeline — news, alerts, events, insider activity, earnings, corporate actions, model signals, and backtest runs in one chronological feed
Exposure Heatmaps — sector, factor, currency, and correlation exposure maps across Home, Cockpit, and Risk
Workspace Presets — Trader / Quant / PM / Risk / Ops presets that reconfigure dashboards, panels, and quick links
Saved Views — capture and restore page, filters, ticker, tabs, columns, and chart layout across major workflows
AI Insight Cards — Gemma-powered insights embedded consistently across Home, Cockpit, Screener, Portfolio, and Security Hub, with graceful offline fallback

Cross-Asset & Macro

Commodities — energy, metals, agriculture with futures term structure and seasonal analysis
Forex — major pairs, cross rates matrix, central bank monitor (Fed, ECB, BoE, BoJ, RBI, and more)
Cryptocurrency — full workspace with markets, movers, sectors, DeFi, derivatives, heatmaps, and correlation
ETF Analytics — holdings viewer, flow tracker, multi-ETF overlap analysis
Mutual Funds — search, comparison, rolling returns, SIP calculator, category rankings, fund overlap
Bonds — fixed income yields, spreads, and duration analytics
Yield Curve — interactive US Treasury curve with historical comparison and 2s10s inversion detection
Economics — global event calendar with impact coding, macro indicators dashboard
Sector Rotation — Relative Rotation Graph (RRG) with 12-week trailing momentum paths

Alerts & Breakout Scanner

Multi-Condition Alert Builder — compound rules with AND/OR logic, multi-field conditions (price, volume, RSI, MACD, moving averages), and natural-language summary
Multi-Channel Delivery — in-app, email, webhook, Slack, and Telegram with per-channel configuration and delivery testing
Alert Lifecycle — cooldown periods, expiry dates, max trigger limits, trigger history with deduplication
WebSocket Push — real-time desktop notifications on alert trigger
Breakout Scanner — automated pattern detection with confidence scoring
Alert History — full timeline with delivery status and re-trigger tracking

Operations & Compliance

OMS — order management with restricted list enforcement and audit trail
Ops Dashboard — feed health monitoring, kill switches, data quality panels
Model Governance — model registry, approval workflows, risk limit monitoring
Cockpit — executive dashboard aggregating portfolio, signals, risk, and events

News & Sentiment

Ticker-Specific News — per-symbol news feed with multi-period filtering, scoped strictly to the selected ticker
Sentiment Analysis — bullish/bearish/neutral classification with confidence scores
Market-Wide Feed — latest headlines with source attribution and sentiment trends
AI Emotion Indicator — per-stock fear/greed gauge powered by a locally hosted Gemma model via LM Studio, surfacing a 0–100 emotion index, dominant emotion (panic → euphoria), emotion mix, and per-article bullish/bearish breakdown
Local & Private — LLM sentiment runs entirely on your own machine; gracefully falls back to the lexical/FinBERT engine when LM Studio is offline

Plugin System & Scripting

Plugin API — extensible architecture for custom analysis modules
Included Plugins — RSI Divergence Scanner, Sector Rotation Monitor, Unusual Volume Detector
Python Scripting — sandboxed execution with security-hardened imports
OpenScript — chart-based indicator scripting with library and sharing

Real-Time Data

Multi-Provider WebSocket — Zerodha Kite (India) and Finnhub (US) real-time ticks
Provider Waterfall — automatic failover chain: primary → fallback → error
Multi-Level Caching — L1 SQLite + L2 Redis with TTL-based invalidation
Candle Aggregation — tick-by-tick to any interval with distributed bar construction
Redis Pub/Sub — horizontal scaling for multi-client quote fan-out

Architecture

+---------------------------------------------------+
|                   CLIENT TIER                     |
|   React 18 + TypeScript + Vite + Tailwind CSS    |
|   TanStack Query + Zustand + Lightweight Charts   |
|   Recharts + Three.js + Playwright + Vitest       |
+--------------------------+------------------------+
                           | REST API + WebSocket
+--------------------------+------------------------+
|                   API GATEWAY                     |
|   FastAPI + Uvicorn + JWT Auth + CORS Middleware  |
|   53 Route Modules (Equity, F&O, Backtest, Risk) |
+--------------------------+------------------------+
                           |
+--------------------------+------------------------+
|                  SERVICE LAYER                    |
|   Unified Fetcher + Screener Engine + Model Lab  |
|   Risk Engine + Alert Scheduler + Quote Hub      |
|   Provider Registry + Failover Chain             |
+--------------------------+------------------------+
                           |
+--------------------------+------------------------+
|                 DATA PROVIDERS                    |
|   Zerodha Kite | Finnhub | FMP | Yahoo Finance  |
|   NSEPython (F&O, Corporate Actions)             |
+--------------------------+------------------------+
                           |
+--------------------------+------------------------+
|                  PERSISTENCE                      |
|   SQLite (default) | PostgreSQL 16 (production)  |
|   Redis (cache + pub/sub + sessions)             |
+---------------------------------------------------+

Data Flow

Market data flows through a unified pipeline:

Exchange ticks arrive via WebSocket adapters (Kite, Finnhub)
Quote Hub fans out ticks to connected clients via /api/ws/quotes
Bar Aggregator constructs OHLCV candles at all supported intervals
OHLCV Cache persists bars in SQLite (L1) and Redis (L2)
Unified Fetcher serves chart requests with cache-first, provider-fallback semantics
Chart Engine renders via Lightweight Charts v5 with indicator overlays

Provider Waterfall

Request → L1 Cache (SQLite) → L2 Cache (Redis) → Primary Provider → Fallback Provider → 503
             HIT → return         HIT → return       OK → cache+return    OK → cache+return

System Requirements

Component	Minimum	Recommended
OS	Linux, macOS, Windows 10+	Ubuntu 22.04+ / macOS 13+
CPU	2 cores	4+ cores
RAM	4 GB	8 GB+
Disk	2 GB	10 GB+ (historical data cache)
Display	1280 x 720	1920 x 1080+
Browser	Chrome 90+, Firefox 90+, Safari 15+, Edge 90+	Latest Chrome or Firefox

Software Dependencies

Software	Version	Notes
Docker	20.10+	Required for containerized deployment
Docker Compose	v2.0+	Included with Docker Desktop
Python	3.11+	Local development only
Node.js	22+	Local frontend development only
Git	2.30+	For cloning the repository

Quick Start

One command (recommended)

git clone https://github.com/Hitheshkaranth/OpenTerminalUI.git
cd OpenTerminalUI
./install.sh          # macOS / Linux / WSL   (Windows: ./install.ps1)

That's it. The installer detects your host OS (macOS, Linux, WSL, or Windows) and adapts, then:

creates a single .env from .env.example,
auto-generates strong JWT_SECRET_KEY and CACHE_SIGNING_KEY (no secret errors),
auto-generates a unique admin password and seeds an admin account, so there are no login errors on first launch,
uses Docker if available, otherwise a local Python + Node setup (auto-detected),
launches the app at http://localhost:8000 and prints your login credentials.

Prerequisites: either Docker (Desktop/Engine with the daemon running) or, for the local path, Python 3.11+ and Node 20+. Nothing else to configure.

First login: when the installer finishes it prints something like:

 OpenTerminalUI is ready  ->  http://localhost:8000
   email:    admin@openterminal.local
   password: <generated unique password>

The same credentials are saved in your .env (BOOTSTRAP_ADMIN_EMAIL / BOOTSTRAP_ADMIN_PASSWORD). Change the password after first login. Seeding is skipped automatically once any user exists, so re-running never clobbers data.

Force a mode if you prefer: OTUI_MODE=docker ./install.sh or OTUI_MODE=local ./install.sh.

Stopping / restarting (Docker):

docker compose down        # stop (keeps your data + seeded admin)
docker compose down -v      # stop and wipe the database (fresh start next time)
./install.sh                # start again

Adding API keys (one place, guided)

All keys live in the single repo-root .env. The easiest way to add or update them is the interactive wizard, which shows what each key unlocks:

make keys          # or: ./scripts/setup-keys.sh

All keys are optional — the platform runs on built-in fallback data without them.

Manual alternatives

Docker by hand

cp .env.example .env      # add API keys if you have them
docker compose up --build            # Backend + Frontend + Redis (SQLite)
docker compose --profile postgres up --build   # with PostgreSQL

Local development (hot reload)

# Backend
python3.11 -m venv .venv && source .venv/bin/activate
pip install -r backend/requirements.txt
PYTHONPATH=. uvicorn backend.main:app --reload --host 127.0.0.1 --port 8000

# Frontend (separate terminal)
cd frontend && npm ci && npm run dev

Backend API: http://127.0.0.1:8000
Frontend dev server: http://127.0.0.1:5173

Environment Variables

The platform runs without API keys using fallback providers. Add keys to unlock full data access:

Variable	Purpose
`FMP_API_KEY`	Financial Modeling Prep — US equities, fundamentals, earnings
`FINNHUB_API_KEY`	Finnhub — US real-time WebSocket ticks
`KITE_API_KEY`	Zerodha Kite — India NSE/BSE real-time + historical
`KITE_API_SECRET`	Zerodha Kite secret
`KITE_ACCESS_TOKEN`	Zerodha Kite session token
`JWT_SECRET_KEY`	JWT signing key for authentication (auto-generated by `install.sh`)
`CACHE_SIGNING_KEY`	Cache integrity signing key (auto-generated by `install.sh`)
`BOOTSTRAP_ADMIN_EMAIL`	Email for the first-run admin account (default `admin@openterminal.local`)
`BOOTSTRAP_ADMIN_PASSWORD`	Password for the first-run admin (auto-generated by `install.sh`; seeding is skipped once any user exists)
`DATABASE_URL`	Database connection (default: SQLite)
`REDIS_URL`	Redis connection for caching and pub/sub
`OPENTERMINALUI_CORS_ORIGINS`	Allowed CORS origins
`OPENTERMINALUI_PREFETCH_ENABLED`	Enable background data prefetch
`LM_STUDIO_BASE_URL`	LM Studio OpenAI-compatible endpoint (default `http://localhost:1234/v1`; use `http://host.docker.internal:1234/v1` from Docker)
`LM_STUDIO_MODEL`	Gemma model id loaded in LM Studio (default `google/gemma-4-26b-a4b`)
`LM_STUDIO_ENABLED`	Toggle the LLM emotion analysis (default `true`; falls back to lexical sentiment when off)
`OPENROUTER_API_KEY`	OpenRouter key powering the AI research agent (free `:free` models work)
`AGENT_PROVIDER`	Agent LLM provider: `openrouter` \| `openai` \| `lmstudio` (default `openrouter`)
`AGENT_MODEL`	Primary agent model id (default `openai/gpt-oss-20b:free`)
`AGENT_FALLBACK_MODELS`	Comma-separated models tried when the primary is rate-limited (429) or unavailable (404)
`AGENT_DEBATE_ENABLED`	Enable multi-agent debate mode in the agent console (default `true`)

AI News Sentiment with Gemma 4 (LM Studio)

OpenTerminalUI integrates a locally hosted Google Gemma 4 model, served through LM Studio, to power the per-stock AI Emotion Indicator on the News workspace. The model reads recent headlines for a ticker and returns a structured judgement — sentiment, confidence, and a market emotion — which the backend aggregates into a 0–100 fear/greed index, a dominant emotion, an emotion mix, and per-article bullish/bearish signals. All inference runs on your own machine; no news or prompt data leaves your hardware.

How it works

News (DB / Yahoo / Google RSS)
        │
        ▼
backend/services/stock_emotion.py ──▶ backend/services/lm_studio_client.py
   (batch prompt + JSON schema)          (OpenAI-compatible /v1/chat/completions)
        │                                          │
        │                                          ▼
        │                                   LM Studio  ·  Gemma 4
        ▼
GET /api/sentiment/emotion/{ticker}  ──▶  Emotion Indicator (News page)

All articles for a ticker are analyzed in a single batched request (large local models are slow — per-article calls would pay the latency N times over).
The request uses LM Studio structured output (json_schema) so the model is constrained to valid, parseable JSON.
If LM Studio is disabled or unreachable, the feature falls back to the built-in lexical / FinBERT sentiment engine, so the endpoint always returns a result.

Integration procedure

Install LM Studio — download from lmstudio.ai (macOS, Windows, Linux).
Download a Gemma model — in LM Studio's Discover tab, search for and download a Gemma model (e.g. google/gemma-4-26b-a4b, or a smaller Gemma variant for faster responses).
Load the model and start the server — load the model, open the Developer / Local Server tab, and click Start Server. It listens on http://localhost:1234 and exposes the OpenAI-compatible API at /v1.
Note the model id — copy the exact model id shown by LM Studio (visible at http://localhost:1234/v1/models); you will set it as LM_STUDIO_MODEL.
Configure OpenTerminalUI:
- Local development — add to .env (defaults already point at localhost):
```
LM_STUDIO_BASE_URL=http://localhost:1234/v1
LM_STUDIO_MODEL=google/gemma-4-26b-a4b
LM_STUDIO_ENABLED=true
```
- Docker — the container must reach LM Studio on the host. docker-compose.yml already defaults LM_STUDIO_BASE_URL to http://host.docker.internal:1234/v1 and maps host.docker.internal. Override LM_STUDIO_MODEL via .env if your model id differs.
Restart the backend (or docker compose up -d) so the new settings load.
Verify — open the News workspace, select any ticker, and check the Emotion Indicator badge:
- Gemma · <model id> — the model is live and analyzing.
- Lexical fallback — LM Studio was unreachable; the built-in engine was used.

Configuration

Variable	Default	Purpose
`LM_STUDIO_BASE_URL`	`http://localhost:1234/v1`	LM Studio OpenAI-compatible endpoint. Use `http://host.docker.internal:1234/v1` from Docker.
`LM_STUDIO_MODEL`	`google/gemma-4-26b-a4b`	Model id loaded in LM Studio. Must match exactly.
`LM_STUDIO_ENABLED`	`true`	Master toggle for LLM emotion analysis.
`LM_STUDIO_TIMEOUT_SECONDS`	`240`	Per-request timeout for the model call.

These can also be set under app: in config/settings.yaml.

Performance: large models such as gemma-4-26b-a4b are slow on consumer hardware — the first analysis for a ticker can take a minute or more (results are then cached). For a snappier experience, load a smaller Gemma / instruct model in LM Studio and point LM_STUDIO_MODEL at it.

Testing

Backend

PYTHONPATH=. python -m compileall backend
PYTHONPATH=. pytest backend/tests -q --cov=backend --cov-fail-under=45

Frontend

cd frontend
npm run build
npx vitest run

End-to-End

cd frontend
npx playwright install chromium
npm run test:e2e

Gate (all checks)

make gate

Repository Layout

backend/                 FastAPI app, adapters, services, routes, tests
  adapters/              Market data provider adapters
  agent/                 AI research agent: orchestrator, tools, debate roles
  api/routes/            53 route modules (equity, fno, backtest, risk, oms, ...)
  core/                  Unified fetcher, failover, service status
  services/              48 business logic modules
  db/                    SQLAlchemy ORM, migrations, caching
  auth/                  JWT authentication and middleware
  config/                Settings, environment, security
  tests/                 409+ backend tests
frontend/                React + Vite + TypeScript SPA
  src/agent/             AI agent console, SSE client, artifact + markdown UI
  src/pages/             51 page components
  src/components/        UI components, terminal design system
  src/fno/               F&O workspace modules
  src/store/             Zustand state management
  src/__tests__/         234+ unit tests
  tests/e2e/             Playwright E2E specs
plugins/                 Extensible plugin system with examples
docs/                    Wiki, architecture specs, and contributor docs
  site/                  GitHub Pages website
  wiki/                  Getting started, contributing guides
data/                    Local SQLite databases and test fixtures
docker-compose.yml       Container orchestration (backend + Redis + Postgres)
Dockerfile               Multi-stage build (Node builder + Python runtime)
Makefile                 Development commands (setup, test, gate)

Keyboard Shortcuts

Shortcut	Action
`Ctrl+G`	GO Bar — symbol lookup and navigation
`Ctrl+K`	Command Palette — fuzzy search across all features
`Ctrl+J`	AI Research Agent — toggle the agent console
`F1`-`F9`	Function keys for workspace switching
`1`-`7`	Timeframe hotkeys in chart views
`Esc`	Close active panel or dialog

Contributing

We welcome contributions. See CONTRIBUTING.md for the full guide.

Fork the repo and create a branch: feat/your-feature or fix/your-fix
Write tests first (TDD encouraged)
Run make gate to pass all checks
Open a PR with a clear description

License

MIT — free to use, modify, and distribute.

Name		Name	Last commit message	Last commit date
.github		.github
.vite/deps		.vite/deps
assets		assets
backend		backend
cli		cli
core		core
data		data
db		db
docs		docs
frontend		frontend
models		models
nlp		nlp
playwright/.auth		playwright/.auth
plugins		plugins
scripts		scripts
test-results		test-results
trade_screens		trade_screens
ui		ui
.dockerignore		.dockerignore
.env.example		.env.example
.gitattributes		.gitattributes
.gitignore		.gitignore
.pre-commit-config.yaml		.pre-commit-config.yaml
CONTRIBUTING.md		CONTRIBUTING.md
Dockerfile		Dockerfile
LICENSE		LICENSE
Makefile		Makefile
OpenTerminalUI.spec		OpenTerminalUI.spec
PRODUCT.md		PRODUCT.md
README.md		README.md
commit_msg.txt		commit_msg.txt
demo_runner.py		demo_runner.py
docker-compose.yml		docker-compose.yml
fundamental_screener_gui.py		fundamental_screener_gui.py
install.ps1		install.ps1
install.sh		install.sh
nse_stock_analyzer.py		nse_stock_analyzer.py
nse_stock_gui.py		nse_stock_gui.py
playwright.config.ts		playwright.config.ts
run.txt		run.txt
run_lan.cmd		run_lan.cmd
run_windows.py		run_windows.py
test_nse_events.py		test_nse_events.py
test_nse_events_v2.py		test_nse_events_v2.py

Latest commit

History

Folders and files

Repository files navigation

OpenTerminalUI

Screenshots

Complete Feature Gallery

Mission Control & Navigation

Equity Research & Markets

Portfolio, Risk & Trading

Quant Research & Backtesting

Futures & Options

Cross-Asset & Macro

Intelligence, AI & Platform

Workspace & Markets

Research & Stock Picking

Portfolio, Risk & Backtesting

AI Research Agent

Features

Terminal Shell

Charting & Technical Analysis

Equity Research & Security Hub

Advanced Screener

Insight-Driven Stock Picking

AI Research Agent

Futures & Options (F&O)

Portfolio & Risk Management

Backtesting & Model Lab

Portfolio Lab

Cockpit, Workspaces & Intelligence

Cross-Asset & Macro

Alerts & Breakout Scanner

Operations & Compliance

News & Sentiment

Plugin System & Scripting

Real-Time Data

Architecture

Data Flow

Provider Waterfall

System Requirements

Software Dependencies

Quick Start

One command (recommended)

Adding API keys (one place, guided)

Manual alternatives

Environment Variables

AI News Sentiment with Gemma 4 (LM Studio)

How it works

Integration procedure

Configuration

Testing

Backend

Frontend

End-to-End

Gate (all checks)

Repository Layout

Keyboard Shortcuts

Contributing

License

About

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages