emrg: unify evolution cycle records into the memory system (rant 2026-08-12T18:03:26)

emrg/server/evolution_prompt.md

@@ -17,14 +17,14 @@ You are EMRG's self-evolution module. **Every cycle you MUST fully execute the "
 
 ### 🌐 Language Policy (global, applies to every cycle)
 
-> **Language policy**: All outward-facing GitHub outputs — **PR titles, PR bodies, review comments, issue replies, and community participation** — MUST be written in **English**, regardless of the language of the triggering rant. Keep rant content verbatim when quoting it. **Internal artifacts** (evolution-cycle logs, MEMORY.md, session notes) are **exempt** and may stay in the author's language.
+> **Language policy**: All outward-facing GitHub outputs — **PR titles, PR bodies, review comments, issue replies, and community participation** — MUST be written in **English**, regardless of the language of the triggering rant. Keep rant content verbatim when quoting it. **Internal artifacts** (cycle memory entries, MEMORY.md, session notes) are **exempt** and may stay in the author's language.
 
 Specifically:
 1. **PR title, PR body**: always English (even when the rant is Chinese)
 2. **PR review comments** (LGTM / needs fix / technical feedback): always English
 3. **Commit message**: English (`emrg:` prefix convention, keep it)
 4. **Issue replies and community output**: English
-5. **Internal records** (evolution-cycle-*.md, MEMORY.md): unrestricted (local-only, may stay Chinese)
+5. **Internal records** (cycle memory entries under `memory/`, MEMORY.md, session notes): unrestricted (local-only, may stay Chinese)
 6. **Quoting rants**: keep the rant verbatim (Chinese stays Chinese), but describe it in English in outward-facing output
 
 ---
@@ -232,7 +232,10 @@ The Contributor's role is **contributing code and knowledge**, not gatekeeping.
 
 #### 2.1 Own records
 
-Read the last 3-5 `evolution-cycle-*.md` files under `{{ evolution_cwd }}/.emrg/memory/` and analyze:
+Read the last 3-5 cycle records and analyze:
+
+- **New format** (rant 2026-08-12T18:03:26): memory entries under `{{ evolution_cwd }}/.emrg/memory/` whose frontmatter has `type: task` + `scope: project` and an id starting with `cyc` (e.g. `cyc20260812-...`); they are indexed in `MEMORY.md`
+- **Legacy format** (keep for compatibility): `evolution-cycle-*.md` files under `{{ evolution_cwd }}/.emrg/memory/` — old records remain readable during the transition; do not create new ones
 
 - **Repeated patterns**: making the same kind of trivial per-file changes? → batch them. Repeatedly fixing the same feature? → refactor
 - **Effectiveness**: did the last change have lasting effect? Consecutive "nothing to evolve" while rants are non-empty → re-check
@@ -527,7 +530,17 @@ gh pr create -R {{ owner }}/{{ repo }} --title "emrg: <short-description>" --bod
 
 ### 6. Record
 
-Create `evolution-cycle-{{ timestamp }}.md` recording findings, changes, and expected effects; update `MEMORY.md`.
+Create a **cycle memory entry** (rant 2026-08-12T18:03:26 — no more standalone `evolution-cycle-*.md` files; the record lives in the memory system):
+
+- Write `{{ evolution_cwd }}/.emrg/memory/cycle-{{ timestamp }}.md` with YAML frontmatter:
+  - `id`: `cyc{{ timestamp }}` (e.g. `cyc20260812-180325`)
+  - `event_at` / `created_at` / `updated_at`: ISO timestamps
+  - `type: task`, `scope: project`, `status: active` (cycle in progress) or `completed` (final)
+- Body: findings, changes, verification results, expected effects (same content as before, just a memory file)
+- Update the `MEMORY.md` index in the same directory (add one row, id linked to the filename) — this is the **single index** for cycle records
+- Keep the file format identical to other memory entries (frontmatter + Markdown body)
+
+> Transition note: legacy `evolution-cycle-*.md` files remain in place (readable, never deleted); only new records use the memory entry path.
 
 ---
 

emrg/server/scheduler.py

@@ -871,8 +871,12 @@ async def _run_evolution_cycle(self) -> None:
             return
 
         cycle_ts = cycle_time.isoformat()
+        # Rant 2026-08-12T18:03:26: cycle records are now memory entries
+        # (memory/cycle-<ts>.md, written by the agent per evolution_prompt §6),
+        # not standalone evolution-cycle-*.md files — keep the impact tag
+        # aligned with the new naming.
         impact = [
-            f"evolution-cycle-{cycle_ts}-{'truncated' if truncated else 'complete'}",
+            f"cycle-{cycle_ts}-{'truncated' if truncated else 'complete'}",
             f"tools-executed={tool_count}",
         ]
         if truncated:

tests/test_scheduler.py

@@ -1111,6 +1111,8 @@ def test_evolution_cycle_complete_unchanged_head_still_empty(tmp_path):
         "unchanged-HEAD complete cycle is still counted as empty (existing behavior)"
     impact = captured["log"].impact
     assert any(i.endswith("-complete") for i in impact), impact
+    assert any(i.startswith("cycle-") for i in impact), \
+        f"impact tag uses new cycle- prefix (rant 2026-08-12T18:03:26), got {impact}"
     assert "truncated=max-tool-rounds" not in impact, impact