A late correction.
A revised estimate.
Feed timestamped observations into Backfold. Correct or withdraw an old value, then query the updated statistics. Fork the stream to keep both histories.
| Time | Observed | Fork |
|---|---|---|
| 10 | 4 | 4 |
| 20 | 40 | 7 |
| 30 | 10 | 10 |
| Mean | 18.0 | 7.0 |
Running real compiled Rust backfold-core (backfold_wasm.wasm, zero backend server). Click anywhere on the top plot to inject a late-arriving sensor reading; Shift+Click to retract that tick into a missing-observation transition step.
Use Backfold in your stack
python -m pip install backfold
Add timestamped observations in the event consumer, then query the revised statistics. Run the Python example.
Build the CLI, place it on your PATH and use this client configuration.
{
"mcpServers": {
"backfold": {
"command": "backfold",
"args": ["mcp"]
}
}
}
Built-in MCP tools (sub-1 MB native binary, zero Python/Node runtime required): backfold_digest_csv, backfold_drill_csv, backfold_kalman_rts_csv (with automatic changepoint shock localization), backfold_covariance_pca_csv, and backfold_ebh_sift_csv.
Verified Live Benchmarks (v0.3.0, max |diff| ≤ 1e-12 vs. Sequential Oracles)
| Operator (T = 100,000 late edit) | backfold Py | Ecosystem Baseline | Python Speedup | Rust vs -O3 Seq Rust |
|---|---|---|---|---|
| 2D Kalman RTS Smooth + log p(y) | 11.12 µs | 6,437,010 µs (NumPy RTS) | 578,868× | 760× (3,694× @500k) |
| 2D Constant-Velocity Kalman | 8.82 µs | 4,563,614 µs (NumPy seq) | 517,710× | 760× (3,538× @500k) |
| Page CUSUM + O(log T) 1st Alarm | 3.83 µs | 137,450 µs (Py seq) | 35,897× | 454× (2,177× @500k) |
| CT Irregular-Δt EWMA | 3.92 µs | 4,832 µs Pandas / 1,941 µs NumPy | 1,231× / 495× | 1,736× (8,152× @500k) |
| 1D Random-Walk Kalman | — | — | — | 1,338× (6,185× @500k) |
| 8D Compensated Covariance & PCA | 15.60 µs | 5,589 µs (NumPy cov) | 358× | 262× (5.68 µs @100k) |
| 8D Discounted Ridge (delayed label) | 16.95 µs | 5,921 µs SciPy / 103,981 µs River@10k | 349× / 7,092× | 659× (6.11 µs @100k) |
| 106-bit Compensated Moments | 6.06 µs | 1,293 µs Pandas / 730 µs NumPy / 273 µs Polars | 213× / 121× / 45× | 259× (1,288× @500k) |
| Arbitrary 60% Range Query (Moments) | 8.57 µs | 35.54 µs (NumPy slice) | 4.1× | 154× (664× @500k) |
Honest Crossover
For strictly append-only, single-pass, in-order aggregation with zero retractions, zero late arrivals, zero range queries, and zero branching, a plain sequential loop wins on initial ingest. On MomentsMonoid, bulk tree build (9.99 ms @ 100k, 51.07 ms @ 500k) breaks even against a single-pass -O3 Rust loop after 31–33 late edits or range queries.