Skip to content

CHANGELOG.md

Latest commit

 

History

History
200 lines (133 loc) · 7.15 KB

CHANGELOG.md

File metadata and controls

200 lines (133 loc) · 7.15 KB
Raw

Changelog

v0.7.1 - 2026-07-20

Changed

  • Refreshed the redacted real-local-run GIF, TUI screenshots, and static HTML overview evidence for the v0.7.1 release surface.
  • Updated release-facing metadata across the CLI, plugin, Homebrew Formula, README, Pages, and sample report.

v0.7.0 - 2026-07-19

Added

  • Added local governance reports for cost audits, prioritized recommendations, observed MCP usage, cross-session context trends, and read-only Git delivery evidence.
  • Added optional local model aliases and per-million-token pricing overrides through AGENTTRACE_PRICING_FILE.
  • Added report scope, pricing confidence, project-root resolution, cache-aware doctor output, and governance appendices to overview exports.
  • Added Action Center, Efficiency, and Delivery workspaces to the TUI, plus bilingual UI copy, live search, paste handling, and report scrollbars.

Changed

  • Updated the CLI, TUI, reports, plugin, Homebrew Formula, Pages assets, and release metadata to v0.7.0.
  • Made CLI report actions mutually exclusive and hardened validation for gate thresholds and output formats.
  • Made delivery and MCP output explicit about evidence limits: commit correlation is not authorship or merge proof, and invocation logs do not imply complete MCP inventory coverage.

Validation

  • Added governance, TUI interaction, discovery, report, and release-surface coverage to the Rust test and CI paths.

v0.6.0 - 2026-07-19

Added

  • Replaced the Go implementation with a Rust workspace while preserving one agenttrace binary for both CLI reports and the terminal TUI.
  • Added cached background session loading, Hermes/OpenCode SQLite sources, Detailed/Aggregate/Limited data capability labels, coverage reporting, privacy-safe tool-step metadata, actionable issue filters, and shared Core findings/comparison rules across CLI and TUI.
  • Added deterministic generated parser fixtures and CI checks for provider coverage, data degradation, step redaction, and single-binary entrypoints.

Changed

  • Split the Rust TUI into state, presentation, filtering, and test modules and moved shared data-health/comparison logic into agenttrace-core.
  • Streamed JSONL object parsing instead of retaining whole-file JSON trees, reducing peak memory by about 44% on a 2.53 GiB local session corpus while keeping report totals unchanged.
  • Added Ctrl+d/Ctrl+u half-page movement and G end navigation to the TUI.
  • Updated public docs, Pages, plugin, and Skill surfaces to describe the Rust implementation and honest per-source evidence limits.

Fixed

  • Invalidated SQLite session snapshots when the database, WAL, or SHM file changes.
  • Hardened preserved-history loading against malformed short identifiers and stabilized the real-data CLI/TUI release smoke checks.

Validation

  • The Rust workspace, release binary, parser fixtures, CLI/TUI PTY entrypoints, report contracts, Homebrew syntax, and Pages artifact are covered by the local release gate.

v0.5.4 - 2026-05-24

Changed

  • Published cross-platform command-line release assets and checksums.

v0.5.3 - 2026-05-24

Changed

  • Refreshed release-facing artifacts for the v0.5 line.

v0.5.2 - 2026-05-24

Fixed

  • Fixed Claude Code JSONL metrics for assistant messages that include thinking, text, and a parallel tool_use batch so the report keeps one assistant turn, multiple tool calls, cache token attribution, and failed tool_result counting aligned. (#243)

v0.5.1 - 2026-05-19

Fixed

  • Clarified agenttrace --doctor cache-state wording so users can distinguish parsed session cache entries, entries reusable for the current scan, and cached directory listings. (#239)

v0.5.0 - 2026-05-18

Added

  • Added local baseline comparison for overview reports so a later run can be checked against a saved local JSON baseline. (#203)
  • Added incident timeline evidence to the TUI and report surfaces. (#204)
  • Added tool authority summaries to HTML, Markdown, and text overview reports. (#210, #212, #214, #219, #221)

Changed

  • Improved overview report readability for Unicode text, incident rows, and terminal-readable authority summaries. (#216, #217, #219, #221)
  • Aligned public README, docs, site metadata, and discovery surfaces with the current local coding-agent session coverage. (#197, #202, #228, #229, #230, #231, #232)
  • Removed stale package-channel and launch-kit surfaces so release-facing install guidance stays limited to available channels. (#225, #226)

Validation

  • Added and refreshed release-surface, report-semantics, Pages artifact, and parser-coverage checks for the v0.5.0 release train. (#178, #182, #184, #205, #208)

v0.4.6 - 2026-05-10

Fixed

  • Show sessions from ~/.pi/agent/sessions as Pi while keeping legacy ~/.omp/agent/sessions sessions labeled Oh My Pi.

v0.4.5 - 2026-05-10

Fixed

  • Added PI auto-discovery for ~/.pi/agent/sessions while keeping the legacy Oh My Pi ~/.omp/agent/sessions path for compatibility.

v0.4.4 - 2026-05-10

Added

  • Added a real local-data marketing refresh script for README and site assets.

Changed

  • Capped overview JSON anomaly details and added anomaly total/truncation metadata so large real histories stay readable for automation and promotional reports.
  • Refreshed README and site screenshots from a real local run.

v0.4.3 - 2026-05-10

Changed

  • Updated release surfaces for the v0.4.3 distribution.

v0.4.2 - 2026-05-05

Changed

  • Refreshed README GIF and screenshots from a real local run with color enabled.
  • Updated release surfaces for the v0.4.2 install paths.

Fixed

  • Kept Session List table values readable when terminal colors are enabled.

v0.4.1 - 2026-05-04

Changed

  • Refreshed the README's real local-run screenshots and summary metrics from the latest TUI against local session logs.
  • Updated release surfaces for the v0.4.1 distribution.

v0.4.0 - 2026-05-04

Changed

  • Polished the first-run TUI demo path with clearer selected-session context, scan-friendly status text, refreshed demo assets, and an updated recording script. (#91)
  • Improved TUI feedback around loading, empty diff states, and command-mode results so users get immediate guidance while navigating. (#122)
  • Made --waste use the same latest-session selection behavior as --latest, reuse loaded diagnostics, and show clearer waste-report copy. (#95)

Fixed

  • Stabilized overview report ordering for recent sessions and anomaly tie breakers across JSON, Markdown, and HTML outputs. (#104)
  • Aligned overview aggregate metrics with TUI discovery, including cache read/write tokens in the exported totals. (#114)
  • Clamped loop waste so reported waste cannot exceed total session cost. (#116)
  • Aligned TUI cache status wording with agenttrace --doctor. (#117)
  • Isolated auto-discovery tests from runner-specific environment configuration. (#120)

Validation

  • Added repeatable CI gates for output contracts, deterministic demo output, report semantics, release surfaces, and Pages artifacts. (#118)
  • Documented the launch-kit validation gates and release consistency checklist for public demo and install surfaces. (#115, #121)