v1.0.0

Added

• Added independent snapshot scheduler jobs, immutable-only emergency recovery, atomic fenced lifecycle transitions, immediate owned-stage retries, bounded source auditing, and separately reported scheduled/weekly backup health.
• Added pipx-installable bassoon CLI token usage history collection, persistence, querying, reporting, and exporting.
• Added the usagebassoon Python API for querying token usage history into pandas and polars (optional) dataframes and Arrow tables.
• Added support for tokscale v4.18.0.
• Added tokscale collection of daily per-session and per-model usage, session metadata, observed pricing, and collector-host metadata.
• Added ephemeral append-only schema-drift observations and resolution events with replay-safe identities and deduplicated observation counts.
• Added an Arrow-based StorageBackend protocol with dialect-specific schemas and views and backend-specific idempotent publication that preserves token usage history.
• Added support for local DuckDB databases using the StorageBackend protocol.
• Added support for BigQuery datasets with concurrent append publication, immediately queryable canonical views, and nightly transactional compaction.
• Added support for MotherDuck databases for remote persistence using the StorageBackend protocol.
• Added terminal summary, model, daily, session, and graph reports; bounded read-only SQL queries; CSV, JSON, and Parquet exports; and Python results as pandas, Polars, or Arrow.
• Added workspace, client, model, tag, source, and inclusive date filters to daily and model reports, session last-active date filters, and a session creation-date display and filter mode; date bounds combine with all other report filters.
• Added the Daily Activity calendar report with total-token, component, cost, cost-per-million, and agent session-time metrics; shared report filters; linear/log scaling; 3–10 intensity colors.
• Added solid, terminal-themed Daily Activity colors blended with the terminal background, bounded theme discovery, and --use-ascii to force character shading.
• Added JSON and CSV output to daily, model, activity, and session reports with existing filters, grouping, limits, optional obfuscation, full identifiers, and numeric costs and derived rates.
• Added persisted daily model timing components, derived milliseconds-per-thousand-token rates, and reconciliation against tokscale's reported rate.
• Added global workspace, client, and session tags with mutation provenance and source-scoped session notes for user-curated reports.
• Added permanent collection audits, models payload reconciliation, and diagnostics using append-only outcomes and issue events.
• Added portable Parquet snapshots and restore for local and remote object store archives, with consistent warehouse capture, catalog-based publication, integrity checks, retention, independent scheduled cadence, and restoration into initialized empty destinations across supported backends.
• Added support for Google Cloud Storage for snapshot archives.
• Added self-contained disk-backed snapshots, catalog reconstruction, cross-location copying, pinned manual recovery points, independent four-week retention, global latest selection, and explicit archive paths/URIs with warned automatic fallback.
• Added snapshot listing, inspection, integrity audits and aliases, protected pin/delete operations, source/run audits directly from archives, recovery initialization, scheduled maintenance shutdown, and atomic restore completion receipts.
• Added scheduled token usage history collection for Linux (systemd), macOS (launchd), and container environments (worker script) with status, log controls, per-operation timeouts, and retries.
• Added sharing controls: reports are raw by default with --sanitize, exports pseudonymize session, workspace, tag, and host fields and redact notes, doctor diagnostics output is sanitized by default, and raw queries warn on stderr.
• Added comprehensive unit test suite that ensures consistent behavior across backends and object stores.
• Added a live persistence test suite against all implemented remote backends and object stores including BigQuery, MotherDuck, and Google Cloud Storage. Tests are automated in the CI Live GitHub workflow.
• Added SQL parity test suite that ensures structural and synthetic replay behavior against all SQL dialects using SQLGlot.
• Added explicit, idempotent warehouse initialization and schema-compatibility preflight checks.
• Added local collector exclusion, replay-safe remote publication, and bounded persistence retries for concurrent ephemeral environments.
• Added operational logger to support debugging, with file rotation and configuration settings.
• Added project configuration through config.toml with primary namespaces tokscale, backend, snapshots, collection, and logging. Loading spinner is configurable (yaspin). Default behavior specified in the README.