SunQueue: Same Work, Less Grid

Less grid energy only counts when the work gets done

Open SunQueue Shift Sheet · Watch the working demonstration · Source and setup

A solar-powered lab or workshop can have enough energy across a day and still run its flexible work at the wrong hour. SunQueue helps the operator plan around daylight, take a readable schedule into the workday, and check the reported outcome afterwards. It keeps stopped jobs separate from completed work instead of counting them as an energy-saving success.

Enter the work, then take the schedule with you

Enter each job's operating power, duration and allowed time window. The new job builder shows the energy conversion immediately: 650 W for two hours is 1.30 kWh. Background demand is separate from job demand, so it is not counted twice. Battery and supply inputs have unit explanations; they remain operator-supplied assumptions, not measurements inferred from a battery label.

The existing planner keeps every requested job and tests both forecast solar and a selected reduced-solar case. It respects one shared machine, deadlines, reserves, charging/discharging limits, outages, an AC load limit and an end-of-day battery minimum.

Shift Sheet maps the chosen slots to an explicit date, start hour and site UTC offset. Each work card gives clock-time start and finish, operating power, job energy and deadline. Download a readable work sheet or tentative calendar file. Nothing is added to a calendar account and no equipment starts automatically.

Record a stopped job without hiding its energy

Each job starts as not yet reported. The operator records a completed run, stopped attempt or job that did not run, together with its actual whole-hour interval. The application refuses to silently treat an unknown outcome as zero energy.

In the demonstration, the same three synthetic jobs require 4.25 kWh of modeled job energy. Then the three-hour training job stops after one hour. Its 850 Wh attempt still contributes to demand; the report shows two of three jobs complete and 2.55 kWh of job load. The equivalent-work savings claim is blocked rather than celebrating the lower workload.

Overlapping runs, missing outcomes and malformed profiles cannot produce a valid comparison. Late finishes and runs before their release window remain visible. A shorter attempt cannot simply be renamed completed under the unchanged duration model.

Compare the same conditions and stored energy

Paste the matching hourly conditions CSV. The same energy simulator compares the reported run, the frozen proposed schedule and an early-start baseline under that profile. It does not optimize again after seeing the reported conditions.

A comparable grid-energy difference appears only when all work is reported completed within its windows, both the reported run and baseline satisfy the energy constraints, and the reported run ends with at least as much stored energy as the baseline. Spending extra battery energy or skipping work is not passed off as a like-for-like saving.

The original synthetic 65%-solar example remains reproducible: the selected schedule models 0 Wh of grid use versus 1,815 Wh for early starts while completing the same three jobs. These are calculated comparisons, not observed workshop savings or carbon reductions. The new demonstration explicitly labels its forecast-derived conditions as an example, not a measurement.

Save the fixed plan, clock mapping, reported intervals and supplied profile together. Reopening checks the inputs and requires recomputation; saved outcome totals are not trusted. Edits invalidate stale reports. The earlier fixed-plan Replay Desk is retained.

Small, local and reproducible

Plain JavaScript runs a bounded scheduling worker and a single hourly energy simulator. The new interval ledger separately checks every run's contribution to demand. No account, API key, cloud inference or upload of user inputs is required. The page continues computing after disconnection while open; the source includes a single-file app for local serving. Direct file-opening behavior was not validated in this environment.

The original scheduler and simulator are unchanged. The new contribution is the complete operator handoff and reported-run comparison, not a newly invented scheduling algorithm. A static site can serve more independent operators without central computation, but the scheduling model remains limited to four jobs, one machine, 24 whole-hour slots and one continuous attempt per job. It does not model changing job power, electrical surges, battery aging, strategic grid charging or concurrent machines. UTC offsets are explicit and do not automatically follow daylight-saving changes.

Evidence and next validation

The release passed 187 Node tests, 41 comparisons with the separately implemented Python solver, 59 browser workflows and 54 calendar-event checks against independent datetime calculations. The 59 browser workflows passed again on the anonymous public deployment with the original document security policy enforced. The new app, source archive and 250.35-second demonstration matched the verified release; only a known link rewrite on the separate judge page differed. Calendar timestamps and file structure were checked, but external calendar-client imports were not.

The four-minute-ten-second demonstration was recorded from the actual app. Narration uses a disclosed stock synthetic voice at approximately 124 words per minute. All cases use synthetic jobs or user-supplied assumptions. These checks do not establish observed workshop benefits.

Executed release · Anonymous public verification · Download source and tests

The target buyer is an operator of a small solar-powered lab or workshop with a few movable jobs. The commercial hypothesis is a free local planner with paid setup and data-import support. Illustrative $25 setup and $5 monthly support pricing is not validated revenue or willingness to pay. A $2 monthly support allowance would leave $3 per site before overhead, acquisition, fees and taxes; actual support burden must be measured before charging.

The next step is an authorized site pilot: freeze schedules beforehand, record completed work and usable energy data, then compare grid use, terminal battery energy and operator effort against the site's normal schedule. No pilot, paying user or physical energy saving is claimed. Completion and profile provenance are self-reported, not authenticated by a fingerprint.

Project history

SunQueue's planner and Replay Desk were created September 7, 2026. Shift Sheet, the job builder, reported-run comparison and this demonstration were added September 8, during the hackathon periods, with substantial AI assistance. This improves the same project entered in NextStep and Next Founders, not two separately invented products.

Original code is MIT licensed; dependencies retain their licenses. The demonstration records actual application actions with synthetic jobs and disclosed stock Kokoro af_heart narration, not a cloned person's voice. Prior immutable releases and the unrelated production portfolio remain unchanged.

The downloadable archive preserves the tested release. The repository branch additionally includes the later public-origin verification report.

Built With

• css3
• html5
• javascript
• netlify
• playwright
• python
• web-workers