Overview
HashEyes (formerly ChainTelescope) is a modern command center for crypto intelligence — built to transform noisy market data into practical decisions. What started as the Jupyter-Crypto-Wizard notebook suite has evolved into a production Streamlit application with modular architecture, multi-provider fallback ingestion, and an in-app AI assistant grounded in live dashboard context.
Stack: Python 3.11 · Streamlit · Plotly · pandas · feedparser · loguru · tenacity · MIT License
Links: hasheyes.flabs.tech · GitHub
Features
- KPI Dashboard — Real-time price, volume, market cap, and dominance metrics with filter-aware time windows and watchlist selection
- Price Trend Charts — Interactive Plotly charts with configurable market sources (Binance, CoinGecko, Coinbase) and automatic mock fallback
- Risk Panels — Bar chart visualization of volatility and risk scores, adjustable by time window
- Alert System — Rule-driven alerts with drawdown and momentum threshold evaluation against live price series
- News Aggregation — RSS/Atom feed ingestion with normalized feed models and safe fallback content
- AI Assistant — GPT-powered chat panel grounded in the current dashboard snapshot (watchlist, KPIs, alerts, news) with safe local fallback when providers are unavailable
- Newsletter Flows — Local subscription persistence with email validation and stub delivery architecture ready for scheduled generation
Architecture
The application follows a clean layered design: app.py (22 lines) delegates to routed pages/ directory, which call view renderers in src/views/, which consume a DashboardSnapshot from the query layer in src/data/. Market adapters, news ingestion, alert rules, and newsletter persistence each live in their own module under src/data/ with provider fallback chains and structured logging.
Shared sidebar filters (time_window, watchlist, market_source, trend_filter) drive the entire snapshot assembly — one filter change propagates through all panels consistently.
AI Assistant (MVP)
The in-app assistant provides short, context-grounded Q&A about the current watchlist, KPIs, alerts, news, and trending highlights. It connects to any OpenAI-compatible endpoint, making provider A/B comparisons straightforward via OPENAI_BASE_URL and OPENAI_MODEL configuration. When credentials are missing or calls fail, it returns a safe fallback summary instead of crashing.
By the Numbers (v0.3.0)
| Metric | Value |
|---|---|
| Tests | 43 (unittest, mock data, assistant wiring, doc contracts, AppTest smoke) |
| CI Jobs | test · build · lint · maintainability |
| Market Providers | Binance → CoinGecko → Coinbase → mock fallback |
| Source Files | 30+ across src/, pages/, scripts/ |
| Dependencies Removed | 8 unused packages trimmed |
Line Count (app.py) | 160 → 22 (modular rewrite) |
Design Philosophy
- Graceful degradation — Every external provider has a fallback chain; the app never hard-fails on remote unavailability
- Observability by default —
logurustructured logging across all ingestion, query, and alert modules - CI-enforced quality — Ruff lint/format,
py_compileon every source file, and unit tests run on every push and PR - Agent-ready — Repository includes AI agent automation playbooks, a PR skill script, and OpenCode planning documents for AI-assisted development workflows
Future Direction
Near-term focus includes expanding market provider coverage, richer alert scoring with investor/developer signal inputs, scheduled newsletter generation and alert evaluation workers, and deeper AI assistant capabilities with tool use and retrieval over ingested datasets.
