v0.5.0 local transcript memory
reza

reza

Local-first memory for every AI coding agent on this PC.

Capture project context, chat turns, handoffs, and cross-tool threads into searchable SQLite so Codex, Claude, Cursor, VS Code, Aider, and future tools can retrieve the same working context.

$pip install reza
$reza install-hooks
$reza sync-all
$reza context current --budget 8000

Local status model

One memory layer, many agents

Project DB

.reza/context.db stores files, sessions, turns, FTS, sources, checkpoints, and threads.

Global registry

~/.reza/registry.db routes cross-project status, handoff, and search.

Always-on mode

reza watch polls adapters, ingests .reza/handoffs/, and reports adapter failures.

Agent access

Any editor can call reza context current, reza session search, or reza global search.

Capture

Adapters import local transcripts, handoffs, and history files without network sniffing.

Retrieve

Search turns, files, sessions, and threads with source-backed snippets.

Govern

Local SQLite, redaction before indexing, exclusions, and audit-friendly source metadata.

Fast retrieval commands

These are the commands other tools should call at session start or when they need prior context.

$reza session search "auth middleware" --json
$reza session handoff --thread thread-abc --budget 8000
$reza context pack --files src/auth.py tests/test_auth.py --budget 6000
$reza global search "stripe webhook"