pulseon 0.1.0a2

Local-first training metrics tracker backed by Rust, PyO3, DuckDB, and DuckLake.

PulseOn

PulseOn is a local-first training metrics tracker backed by Rust, PyO3, DuckDB, and DuckLake.

Current focus: plan and implement the 0.1.0a2 native loop:

• create a project
• start or resume a run
• log explicit-step numeric metrics through a bounded async queue
• query metric series and summaries locally
• support local DuckDB DuckLake catalog storage
• defer SQLite catalog support until real DuckLake-backed parity tests pass
• keep the v2 public data path local-filesystem only
• keep Parquet as the long-term compatibility boundary

Python API shape:

import pulseon

client = pulseon.init(
    "runs",
    data_path=None,
    catalog_backend="duckdb",
    catalog_path=None,
    metric_queue_capacity=65536,
)
project = client.create_project("local training")
run = client.create_run(project.project_id, "baseline")
run.log("train/loss", 0, 0.25)

For bounded teardown, stop active logging threads before calling client.shutdown(timeout=...); PulseOn keeps admission open while bounded shutdown is draining, so concurrent run.log(...) calls can prevent that drain from completing before the timeout.

Architecture entry points:

• Catalog/data boundary
• V1 native architecture
• Glossary
• Roadmap
• ADRs

Runtime extensions:

• DuckLake is installed and loaded by the native engine because it is required for native storage.
• DuckDB LTTB is not bundled into the SDK. PulseOn loads an already installed lttb extension when downsampling is requested with max_points. To allow PulseOn to download it from the DuckDB community repository at that point, set PULSEON_LTTB_AUTO_INSTALL=1. To use a local build instead, set PULSEON_LTTB_EXTENSION_PATH=/path/to/lttb.duckdb_extension.