A research dashboard for global technology equities, focused on pair trading, earnings tracking, and thematic exposure analysis. Designed for an equity long/short PM workflow — it answers the questions you ask in the hour before a print: what's the right hedge for this name, where is the IV cheap, who reports next week, what did the stock do last quarter.
┌────────────────────────┐
│ Render (FastAPI) │
│ serves dashboard │
└───────────▲────────────┘
│
│ deploy on push
│
┌────────────────┐ ┌────────────┴────────────┐
│ yfinance │───▶│ Local ETL (Python) │───▶ git push
│ Finnhub │ │ → dashboard.db (SQLite)│
└────────────────┘ └─────────────────────────┘
The application is a single FastAPI process with a server-rendered Alpine.js frontend. The ETL pipeline runs locally on a schedule (Windows Task Scheduler in our setup), updates a SQLite snapshot, commits it to the repo, and Render redeploys with the new data.
| Heatmap (themes × subsectors) | Per-company page |
|---|---|
 |
 |
| Pair workspace with 12m chart | Upcoming earnings feed |
 |
 |
- Thematic heatmap — a matrix of
theme × subsectorbullishness ratings, with per-cell rationale stored in markdown. Click any theme row to drill into the exposed tickers and a suggested long/short basket. - Per-company page — valuation snapshot (P/E forward & TTM, P/B, P/S, PEG, growth metrics), consensus estimates, next earnings event, business exposure weights by category, and historical day-after earnings reactions for the last four quarters.
- Pair trading workspace — cosine similarity between any two tickers based on business exposure weights, side-by-side multiples & volatility stats, IV-neutral and vol-neutral sizing ratios, an exposure overlap visualization, and a 12-month rebased performance chart with an interactive crosshair.
- Upcoming earnings feed — chronological cards for the next 1/2/4/8 weeks, each showing mkt cap, forward P/E, EPS YoY growth, and the last four earnings reactions inline.
- Focal ticker overlay — pick a ticker and the heatmap highlights its subsector column and dots the cells where that ticker has a theme score above 40 (strong dot above 70).
| Layer | Tool |
|---|---|
| Backend | FastAPI, Uvicorn |
| Database | SQLite (backend/db/dashboard.db, committed to the repo) |
| Frontend | Alpine.js + Tailwind CSS via CDN — no build step |
| Charts | Inline SVG (no chart library dependency) |
| Data ingestion | yfinance, finnhub-python, lxml (for earnings dates) |
| Deploy | Render (free tier) |
No Webpack, no React, no npm, no Docker. The whole frontend is three self-contained HTML files served by FastAPI.
git clone https://github.com/figoux/solana-capital-global-tech-equities.git
cd solana-capital-global-tech-equities
pip install -r requirements.txt
# The DB ships populated in the repo — no setup needed to view existing data
uvicorn backend.api.server:app --host 127.0.0.1 --port 8000
# Open http://127.0.0.1:8000That's it. The dashboard works against the snapshot of dashboard.db in the
repo. No API keys, no env vars, no auth required for read-only viewing.
The included render.yaml is a Render Blueprint. Connect the repo to Render
and it provisions a free web service automatically.
https://dashboard.render.com → New + → Blueprint → connect this repo → Apply
Build takes ~5 minutes. After that, every git push triggers a redeploy in
~2 minutes.
If you want to gate your deployment behind HTTP Basic auth (e.g. for an
internal-only mirror), set DASHBOARD_PASSWORD as an env var on the service
— the server picks it up automatically and requires login. Leave it unset
for an open, public dashboard.
The ETL is a sequence of Python modules. Each can be run independently or as part of the daily refresh script.
# 1. Universe & taxonomy (one-time or after CSV edits)
python -m backend.etl.universe # loads universe.csv → companies table
python -m backend.etl.business_exposures_seed # taxonomy of 27 exposure buckets
python -m backend.etl.exposures_seed # ticker → bucket weights
python -m backend.etl.themes_seed # theme definitions
python -m backend.etl.theme_mapping_seed # theme → bucket weights
# 2. Market data refresh (run daily)
python -m backend.etl.prices_yf # 2y of daily OHLCV
python -m backend.etl.fundamentals_yf # multiples, EPS/revenue estimates
python -m backend.etl.vol_yf # realized & implied vol from options
python -m backend.etl.earnings_cal # upcoming earnings via Finnhub + yfinance
python -m backend.etl.earnings_history_yf # historical day-after reactions
# 3. Recompute derivatives
python -m backend.etl.pairs_compute # cosine similarity matrixscripts/daily_update.ps1 runs the full sequence and commits the DB. We have
it scheduled at 07:00 daily via Windows Task Scheduler. A cron equivalent
on Linux is straightforward.
earnings_cal.py uses Finnhub for forward-looking earnings calendar coverage
that yfinance misses. The free tier is sufficient.
cp .env.example .env
# Edit .env and set FINNHUB_API_KEYWithout a key, the script falls back to yfinance.calendar (works but covers
fewer names).
The default universe is 130 global tech tickers (US, EU, Asia). To adapt to your own coverage:
-
Edit
universe.csv— one row per ticker. Required fields:ticker,yahoo_ticker,name,subsector,country,currency. Setis_private=1for unlisted names you want tagged but not fetched. -
Edit
backend/etl/exposures_seed.py— map each ticker to one or more business exposure buckets (weights sum to 100). The 27 buckets cover Consumer, Cloud/Data, Hardware/Semis, Fintech, and Frontier categories. Add new buckets inbusiness_exposures_seed.pyif your domain needs them. -
Edit
backend/etl/themes_seed.py— define investment themes and rate each(theme × subsector)cell from −2 (very bearish) to +2 (very bullish) with an optional markdown rationale. -
Re-run the seed scripts in the order shown above, then
pairs_computeto recompute the similarity matrix.
The dashboard reflects changes automatically on the next request.
.
├── backend/
│ ├── api/
│ │ └── server.py # FastAPI app + all REST endpoints
│ ├── db/
│ │ ├── dashboard.db # SQLite snapshot (committed)
│ │ └── schema.sql # canonical schema
│ └── etl/ # one module per pipeline stage
├── frontend/
│ ├── index.html # dashboard home — heatmap + upcoming earnings
│ ├── company.html # per-ticker drill-down
│ ├── pair.html # pair workspace with 12m chart
│ └── theme.html # theme drill-down with suggested basket
├── scripts/
│ ├── daily_update.ps1 # full ETL + git push (Windows)
│ ├── migrate.py # idempotent schema migrations
│ └── db_info.py # sanity-check counts and freshness
├── universe.csv # ticker universe (editable)
├── render.yaml # Render Blueprint
└── requirements.txt
Cosine similarity for pairs. Each ticker is a vector in 27-dimensional
exposure space (one dimension per business bucket, weight 0–100). Pair
similarity is the cosine of the angle between two such vectors. The matrix
is stored in pairs_similarity and the top-N peers per ticker are surfaced
in the UI. Values above 0.7 usually mean "trades as a pair," below 0.4 means
"different businesses despite shared subsector."
Theme scores. A (ticker × theme) score is the weighted sum of the
ticker's exposure weights times the theme's bucket weights, normalized to
0–100. A manual direction_override can pin a ticker's direction (long/short)
inside a theme regardless of its subsector-level bullishness rating.
Day-after reactions. For each historical earnings date, the reaction
is close_after / close_before − 1, where the choice of before/after
depends on whether the company reports BMO (before market open) or AMC
(after market close). When the timing is unknown, the fallback compares
close[D+1] to close[D−1] which captures the event either way.
IV / RV / skew. Implied vol is the median of ATM puts/calls in the nearest two expiries weighted to 30-day and 60-day buckets. Skew is the spread between the 25-delta put IV and the 25-delta call IV. Realized vol is the annualized close-to-close standard deviation over the trailing 30 and 60 sessions.
- SQLite single-file storage is fine for read-heavy single-process workloads, not for concurrent writes. Don't run the ETL while uvicorn is hot.
- yfinance options chains are unreliable for some non-US listings (notably Taiwan and Hong Kong). Those tickers show RV but no IV/skew.
- The free Render tier spins down after 15 minutes of inactivity. First request after spin-down takes ~30 seconds. Upgrade tier to keep warm.
- Earnings history depth depends on
yfinance.get_earnings_dates(), which typically returns 4–8 past quarters. Recently-IPO'd names will have shorter histories.
PRs welcome, especially:
- New universe entries with reasoned exposure splits
- New themes with subsector ratings and rationale
- Frontend additions that respect the "no build step" constraint
- ETL connectors for non-yfinance data sources (Polygon, Refinitiv, etc.)
Please open an issue first for anything that changes the schema or the API surface.
MIT — see LICENSE.