Sentinel
Crypto trading bot with an Ollama filter and built-in backtesting
A crypto trading bot for Binance Spot USDC: rule-based entries, portfolio risk, optional local Ollama filter, and backtest - built for a crypto enthusiast group.
Situation
A group of crypto enthusiasts needed a shared, explainable spot-trading tool: the same decision stack, clear risk controls, and a local AI filter - not a cloud SaaS and not margin/short strategies.
Bottleneck
Cloud bots and black boxes hide decision reasons. Goal: deterministic rules, logs, paper mode before any live risk, and one Windows package the group can run locally.
What we built
Python/FastAPI + pywebview GUI: chart dashboard, readiness and decision stack, market context, AI BUY/SELL review, performance, settings, Ollama panel, backtest, journal/logs; SQLite persistence. ~76 pytest unit tests on the engine.
Outcome
The group received one local tool for following the same signals, testing the strategy against historical and live data, and documenting every decision. Paper mode is the default, while live trading is enabled separately.
Detail
AI review helps evaluate a BUY candidate, but it never takes control of exits. Every manual Buy/Sell remains subject to user confirmation.
Related work
ISTA+ Reader
A Windows workstation for a mechanic shop: reads VIN, DTCs, and ECU modules over ENET/HSFZ - without dealer software, without coding, and without write risk to the vehicle.
MJDD project assessment
When you started a conversation on our website, you used exactly this system. Instead of a contact form, a visitor describes their situation in a short conversation and gets an assessment within minutes - and we arrive at the first call prepared.
MAT Design platform
For the MAT design studio we built a bilingual React + Node platform: editorial case studies, an art-print shop, Stripe and bank-transfer checkout, and admin for products, orders, and gallery - without a generic CMS.