Tastier

A local-only TastyTrade dashboard: live positions, payoff graphs, and AI position analysis. Runs on 127.0.0.1:8420 so your TT credentials never leave your machine.

What it does

• Live positions — pulls open positions (read-only), groups by underlying, streams live marks via DXLink.
• Payoff graph — server-side Black-Scholes T+0 curve using live IV plus expiration P/L, breakevens, and live P/L. Rendered in the browser with Recharts.
• AI analysis — asks an LLM to rate position health/risk/P-L and suggest actions. Supports Anthropic, OpenAI, Gemini, DeepSeek, and Kimi (Moonshot).
• History — every analysis is saved per symbol; switch between past runs from the dropdown.
• Editable prompt — the position-analysis system prompt lives in app/prompts/position_analysis.md. A copy in your OS user-data folder can override it without rebuilding the app.

Security model

• Read-only OAuth grant — the app cannot trade.
• Credentials live only in a .env file inside your OS user-data folder (%LOCALAPPDATA%\Tastier on Windows, ~/Library/Application Support/Tastier on macOS). Never in source control, never sent to the browser.
• Server binds 127.0.0.1 only.
• No order/trade endpoints in the codebase.

Quick start (release binary)

Download a zip from Releases:

• Windows: run Tastier.exe.
• macOS: right-click Tastier.app → Open.

The first-run setup page asks for TT_SECRET and TT_REFRESH. See OAuth setup below. Choose paper for the sandbox or live for your real account.

Developer setup (Python 3.11+)

pip install -r requirements.txt

# Copy the example .env to your user-data folder and edit it.
# Windows PowerShell:
$d = "$env:LOCALAPPDATA\Tastier"; New-Item -ItemType Directory -Force $d
Copy-Item .env.example "$d\.env"
# macOS / Linux:
mkdir -p ~/Library/Application\ Support/Tastier
cp .env.example ~/Library/Application\ Support/Tastier/.env
chmod 600 ~/Library/Application\ Support/Tastier/.env

make check   # should print ok:true with your accounts
make run     # open http://127.0.0.1:8420

OAuth setup

Get a read-only personal OAuth grant from tastytrade.

1. Log in to the right portal:
   • Paper: https://developer.tastytrade.com (sandbox login)
   • Live: https://my.tastytrade.com
2. Manage → My Profile → API → OAuth Applications.
3. Create an app named Tastier Local. Scope: read only.
4. Create a personal grant.
5. Copy Client Secret → TT_SECRET, Refresh Token → TT_REFRESH.

Paper and live credentials are separate. If auth fails, make sure TT_ENV matches the portal you used.

LLM analysis setup

Add one or more keys to your user-data .env:

ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...
GEMINI_API_KEY=...
DEEPSEEK_API_KEY=sk-...
# Kimi/Moonshot accepts either name:
KIMI_API_KEY=sk-...
MOONSHOT_API_KEY=sk-...

# Optional:
LLM_PROVIDER=anthropic          # anthropic | openai | gemini | deepseek | kimi
LLM_MODEL=claude-sonnet-5       # override the default model
KIMI_BASE_URL=https://api.moonshot.ai/v1  # proxy/OpenRouter support

The model dropdown is refreshed from each provider at startup. Optional pricing overrides go in <user-data>/Tastier/pricing/models_pricing.json.

Tests

Command	What it checks
make test	Offline: option pricing, P&L, breakevens, position grouping, mocked broker routes
make test-e2e	Live paper account: auth → positions → DXLink quote → analysis

Architecture

browser (React + Recharts, no build step)
   │  REST: /api/positions, /api/analyze, /api/analyses
   │  WS:  /ws/quotes
   ▼
FastAPI (127.0.0.1:8420)
   │  session + token refresh
   ├─► tastytrade REST  — accounts, positions, option details
   ├─► DXLink websocket — Quote/Greeks, single upstream, fan-out relay
   └─► LLM providers      — Anthropic, OpenAI, Gemini, DeepSeek, Kimi

Key files:

• app/main.py — FastAPI routes, DXLink relay, analysis endpoints.
• app/advisor.py — LLM provider selection, model enumeration, pricing.
• app/payoff.py — T+0/expiration curves and Greeks-based valuation.
• app/prompts/position_analysis.md — editable system prompt.
• run_app.py — PyInstaller entry point that locates bundled assets.

Release builds

Push a version tag to build Windows and macOS binaries via GitHub Actions:

git tag v0.1.0
git push origin v0.1.0

.github/workflows/release.yml attaches Tastier-Windows-x64.zip and Tastier-macOS.zip to a GitHub release. .github/workflows/nightly.yml builds a daily pre-release at 4 AM Pacific if the repo changed in the last 24 hours.

Build locally with:

make build

Output: dist/Tastier.exe (Windows) or dist/Tastier.app (macOS).