chore(release): 0.8.0

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-mem",
-  "version": "0.7.0",
+  "version": "0.8.0",
   "description": "Bounded project memory tools and seven deterministic capture hooks, without a resident worker.",
   "mcpServers": {
     "agent-mem": {

CHANGELOG.md

@@ -1,7 +1,8 @@
 # Changelog
 
-## Unreleased
+## 0.8.0
 
+- Native-memory import and portable projects (agent-mem#40 F17, AC9, #58): `amem import claude-auto` and `amem import codex` read Claude Code auto memory and Codex memories into `memory` candidates for person review only (never curated notes), screened by the configured policy and the secret screen, at most 100 per run, idempotent on repeat, with sources never changed and `CLAUDE.md`, `AGENTS.md`, `*.local.md` and `rules/` excluded. `amem project export` gains `--format jsonl|markdown` (default `json` unchanged); `jsonl` keeps authored rows and aliases, `markdown` is read-only. The person-only `amem project import <key> --from <file.jsonl> [--confirm] [--allow-shared]` previews by default and re-imports curated notes with their original entry ids after validating every row first; capture rows, goals and work lists stay export-only. See `docs/contracts.md`.
 - Korean recall set and pinned baseline (agent-mem#62, task #472): `amem eval --set ko` scores a second built-in synthetic set (`src/memory/eval-set-ko.json`, 37 invented entries and 38 questions covering particle variants, spaced and unspaced compounds, mixed Korean-English terms, Latin abbreviations and short/long tier). Questions take an optional `tag` label that scoring ignores. Baseline on the current engine: recall@1/5/10 0.8684/0.9474/0.9474, MRR 0.8934, pinned in `test/eval.test.ts` and recorded in `docs/contracts.md`. Measurement only: no ranking or tokenizer change.
 - Consolidation follow-ups (task #439): an accepted consolidation note is never a source of the next run (no recurring-fix or chain summaries of summaries; an accepted recurring-fix summary no longer pairs with its own sources), the report counts only newly queued or still-pending candidates and says how many were already reviewed, `amem consolidate --project` refuses a project other than the current one (its repository policy cannot be resolved from here), and acceptance reads one placement proof per locked section instead of four.
 - Sync and doctor signal polish (task #437): after an abort, a transport child of git that ignores SIGTERM is SIGKILLed with its process group before the group is forgotten; `sync --if-due` ended by a person's signal exits like that signal (130, 143, 129, 131) instead of 1, as `amem doctor` does; `amem doctor` gives each git lookup a 5 second limit, so a signal sent to doctor alone is handled after at most one call (a failed lookup reads as unknown, as before).

README.md

@@ -2,7 +2,7 @@
 
 Curated memory with append-only JSONL logs, a derived SQLite index, short- and long-term tiers, a typed graph, goals with leases, and git sync between machines. Extracted from [agent-hub](https://github.com/STAIxBWLB/agent-hub), with compatible storage and entry IDs.
 
-Version: 0.7.0. Requires Bun 1.4.2 or later; no runtime dependencies.
+Version: 0.8.0. Requires Bun 1.4.2 or later; no runtime dependencies.
 
 ## Quick start
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@staix/agent-mem",
-  "version": "0.7.0",
+  "version": "0.8.0",
   "description": "Curated, shared memory for coding agents: append-only logs, a derived index, tiers and a typed graph.",
   "license": "MIT",
   "type": "module",