Changelog

All notable changes to reza are documented here. Follows Keep a Changelog and Semantic Versioning.

Changelog

All notable changes to reza are documented here.

Format follows Keep a Changelog. Versioning follows Semantic Versioning.


[0.5.0] — 2026-04-17

Added

  • Cross-tool session import — reza sync-cursor and reza sync-codex read Cursor agent transcripts (~/.cursor/projects/…) and Codex rollout JSONL (~/.codex/sessions/…); reza sync-all runs registered sync paths. Imports are idempotent (safe to re-run; new turns append only).
  • reza.ingest package — parsers for Cursor and Codex formats; shared upsert_imported_session for deduplicated sessions and turns; file paths harvested from tool calls for graph scope.
  • Session metadata — sessions.source_tool, sessions.source_path, sessions.source_id (migration on upgrade) for provenance and deduplication.
  • CLI: reza session show, reza session load (handoff pack; --copy to clipboard with optional pyperclip), reza session graph (files + node scope; --json for automation).
  • Richer reza session list — tool column, turns, tokens, files touched, relative age; --source, --limit, --json.
  • VS Code graph webview: Sessions panel — list imported sessions, filter by tool, Highlight vs Subgraph only on the code graph, sync buttons, Pack copies handoff via reza session load … --copy.

Changed

  • reza ingest module layout: legacy file-based ingest lives under reza.ingest.files; package reza.ingest re-exports public APIs.

Documentation

  • README, npm README, and docs site updated for 0.5.0 (import commands, session scope, VS Code sessions UI).

[0.4.0] — 2026-04-12

Added

  • reza sync-claude <jsonl_path> — parse Claude Code's .jsonl conversation file and sync all turns to reza; idempotent (only appends new turns)
  • reza sync-claude --from-hook — Stop hook mode: reads transcript_path + cwd from stdin JSON; zero tokens needed from Claude
  • reza install-claude-hook — writes a Claude Code Stop hook into ~/.claude/settings.json so every response is auto-synced after it finishes; --uninstall flag to remove
  • reza session start now writes session ID to .reza/current_session so the Stop hook auto-picks the right session without any arguments
  • reza/claude_sync.py — new module: parse_jsonl (handles both string and [{type:"text"}] content), sync_claude_session (idempotent, auto-creates session if needed)
  • 17 new unit tests covering parse edge cases and full sync lifecycle

Changed

  • reza session start writes .reza/current_session side-effect (backwards-compatible)

[0.3.0] — 2026-04-12

Added

  • reza session turns add — append conversation turns to a session (single turn or bulk from JSON file)
  • reza session turns list — list all turns for a session
  • reza session search <query> — full-text search across all conversation history using FTS5 + BM25 ranking; supports --id (session filter), --limit, --json
  • reza ingest <file> — ingest .md or .json transcript files as conversation turns; auto-creates session from filename prefix; prevents double-import
  • reza session handoff extended — new --id, --format markdown|json, --budget, --search flags; renders structured markdown brief ready to paste into any AI tool
  • reza watch — auto-ingests files dropped into .reza/handoffs/ in real time
  • FTS5 conversation_turns_fts virtual table with Porter + Unicode61 stemming for accurate multilingual search
  • DELETE and UPDATE sync triggers keep FTS index consistent with base table
  • reza upgrade backfills FTS index for existing turn data

Changed

  • reza session handoff --json is now deprecated in favour of --format json

[0.2.0] — 2026-04-11

Added

  • conversation_turns table — structured per-turn conversation history linked to sessions (role, content, token_est, turn_index)
  • handoff_drops table — tracks ingested file drops to prevent double-import
  • reza/turns.py — new module: add_turn, add_turns_bulk, list_turns, turns_within_budget (budget-aware retrieval drops oldest turns first)
  • reza/ingest.py — new module: parse and ingest .md and .json transcript files as conversation turns; reza ingest file-drop workflow
  • .reza/handoffs/ directory auto-created on reza init and reza upgrade
  • reza upgrade now runs schema migration so existing installations get new tables without re-init
  • UNIQUE(session_id, turn_index) constraint ensures turn ordering integrity

Changed

  • reza upgrade applies init_schema before re-scanning (idempotent migration)

[0.1.0] — 2024-04-10

Added

  • reza init — initialize context database, scan project files, install git hooks
  • reza status — quick project overview
  • reza watch — real-time file watcher using watchdog
  • reza query — search context database (overview, find, recent, sessions, file info)
  • reza session — session management (start, save, end, list, handoff)
  • reza export — export context to markdown, JSON, or compact LLM prompt format
  • reza hooks — install / uninstall pre-commit git hook
  • reza upgrade — re-scan all files
  • 6-table SQLite schema: project_meta, files, dependencies, sessions, changes, conflicts
  • Purpose extraction from Python docstrings, JSDoc, Markdown headings, and first comments
  • Framework detection: Django, FastAPI, Flask, React, Vue, Next.js, Nuxt, Svelte, Astro, Express, Go, Rust, Maven, Gradle, Rails
  • Cross-LLM handoff via reza session handoff
  • Integrations for: Claude Code, Cursor, Kilocode, Aider, Continue.dev, GitHub Copilot, Codeium/Windsurf, OpenAI Codex
  • JSON output mode for all query commands (--json)
  • Full test suite with pytest

Future

  • VS Code extension published to Marketplace (local .vsix / repo build today)
  • Web UI for browsing .reza/context.db beyond the docs site
  • AST-based purpose extraction (more accurate than regex)
  • reza diff — show what changed between sessions
  • Team sync via shared DB (opt-in)
  • reza annotate — let LLMs write notes directly to file records