Power-curve tuning tool for a wind turbine driving a Morningstar TriStar 600V MPPT charge controller. Built with Deno + Fresh, SQLite, and Canvas-2D scatter plots. Companion to wendy, the data-collection dashboard that feeds it live samples.

Read-only. wendy-curves observes and analyses — it does not write to the TriStar controller. All curve changes are still applied manually. A future optional Modbus write path is gated behind WENDY_CURVES_ALLOW_WRITE=1 and is not implemented yet.

wendy-curves consumes wendy's live data via SSE, archives samples in its own SQLite, lets you record which voltage→watts curve was loaded between which timestamps, and helps find a better curve by:

• Plotting all observed (array V, array W) samples with the active curve, the empirical p90 envelope, and a suggested curve overlaid.
• Scoring the active curve against the previous one on energy yield and stall count.
• Generating a 16-setpoint suggested curve under monotonicity, hardware cap, stall hotspot back-off, and small-step constraints.

Phase 1 implemented:

• Bootstrap consumer: SSE + history fetch from wendy
• SQLite long-term archive (curves, activations, samples, metrics cache)
• Pure analyzer: envelope, stalls, suggestion algorithm, verdict label
• Fresh UI: scatter plot, setpoint table, curve drawer, activation timeline, live footer
• Docker + deploy script

Phase 2 (direct Modbus curve write to TriStar) is out of scope; tracked as a follow-up. Set WENDY_CURVES_ALLOW_WRITE=1 will be required to enable it once implemented.

wendy ──SSE/HTTP──► wendy-curves ──► wendy-curves.db (long-term archive)
                          │
                          ▼
                       Analyzer (pure)
                          │
                          ▼
                    Fresh app (scatter UI)
                          :
                          └──► CurveWriter (phase 2: Modbus)

wendy-curves is conceptually "another display consumer" of wendy — it talks to wendy over HTTP/SSE rather than reading wendy's database directly. Wendy keeps its own 7-day hot window plus a compressed 365-day cold archive; wendy-curves keeps its own independent long-term archive tuned for power-curve analysis.

• wendy reachable via HTTP/SSE (exposes tristarPower on /api/events and ?from=&to=&source= query parameters on /api/history)
• Deno 2.x (or Docker via the bundled Dockerfile + docker-compose.yml)

# Copy the example env file and point at a running wendy instance
cp .env.example .env
# edit .env — at minimum set WENDY_CURVES_WENDY_URL to your wendy host

# Run locally
deno task dev

# Or via Docker
docker compose up

The dashboard is served on WENDY_CURVES_PORT (default 8087). See .env.example for the full list of configuration options.

• Deno 2.x / Fresh 2 / TypeScript — server and UI
• SQLite (via @db/sqlite) — long-term sample archive
• Server-Sent Events — live ingest from wendy
• Canvas 2D — scatter plot rendering (no charting library)
• Docker — optional containerised deployment

• wendy — the energy-dashboard sibling that polls Morningstar TriStar and Victron hardware via Modbus TCP and serves the live SSE stream wendy-curves consumes.

MIT