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-cursorandreza sync-codexread Cursor agent transcripts (~/.cursor/projects/…) and Codex rollout JSONL (~/.codex/sessions/…);reza sync-allruns registered sync paths. Imports are idempotent (safe to re-run; new turns append only). reza.ingestpackage — parsers for Cursor and Codex formats; sharedupsert_imported_sessionfor 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;--copyto clipboard with optionalpyperclip),reza session graph(files + node scope;--jsonfor 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 ingestmodule layout: legacy file-based ingest lives underreza.ingest.files; packagereza.ingestre-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.jsonlconversation file and sync all turns to reza; idempotent (only appends new turns)reza sync-claude --from-hook— Stop hook mode: readstranscript_path+cwdfrom stdin JSON; zero tokens needed from Claudereza install-claude-hook— writes a Claude Code Stop hook into~/.claude/settings.jsonso every response is auto-synced after it finishes;--uninstallflag to removereza session startnow writes session ID to.reza/current_sessionso the Stop hook auto-picks the right session without any argumentsreza/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 startwrites.reza/current_sessionside-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 sessionreza session search <query>— full-text search across all conversation history using FTS5 + BM25 ranking; supports--id(session filter),--limit,--jsonreza ingest <file>— ingest.mdor.jsontranscript files as conversation turns; auto-creates session from filename prefix; prevents double-importreza session handoffextended — new--id,--format markdown|json,--budget,--searchflags; renders structured markdown brief ready to paste into any AI toolreza watch— auto-ingests files dropped into.reza/handoffs/in real time- FTS5
conversation_turns_ftsvirtual table with Porter + Unicode61 stemming for accurate multilingual search - DELETE and UPDATE sync triggers keep FTS index consistent with base table
reza upgradebackfills FTS index for existing turn data
Changed
reza session handoff --jsonis now deprecated in favour of--format json
[0.2.0] — 2026-04-11
Added
conversation_turnstable — structured per-turn conversation history linked to sessions (role, content, token_est, turn_index)handoff_dropstable — tracks ingested file drops to prevent double-importreza/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.mdand.jsontranscript files as conversation turns;reza ingestfile-drop workflow.reza/handoffs/directory auto-created onreza initandreza upgradereza upgradenow runs schema migration so existing installations get new tables without re-initUNIQUE(session_id, turn_index)constraint ensures turn ordering integrity
Changed
reza upgradeappliesinit_schemabefore re-scanning (idempotent migration)
[0.1.0] — 2024-04-10
Added
reza init— initialize context database, scan project files, install git hooksreza status— quick project overviewreza watch— real-time file watcher usingwatchdogreza 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 formatreza hooks— install / uninstall pre-commit git hookreza 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.dbbeyond 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