Playbook v0.3.0 — refresh to current stack; m2o-operator as manager

.github/workflows/ingest.yml

@@ -13,16 +13,41 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - name: Notify fleet — trigger re-ingest via escalation inbox
+      # Store the updated playbook straight into the fleet memory system so agents
+      # can semantically recall it. memory.machinemachine.ai is the current store
+      # (agent.memory.system, BGE-M3 + Qdrant); namespace "m2" is fleet-wide.
+      - name: Re-ingest PLAYBOOK.md into fleet memory (namespace m2)
         run: |
-          curl -s -X POST "https://bge-proxy.machinemachine.ai/escalate" \
-            -H "Authorization: Bearer ${{ secrets.FLEET_TOKEN }}" \
+          CONTENT=$(cat PLAYBOOK.md)
+          python3 - "$CONTENT" "${{ github.sha }}" <<'PY'
+          import json, sys, urllib.request
+          content, sha = sys.argv[1], sys.argv[2]
+          # chunk to keep each memory reasonable
+          chunks = [content[i:i+4000] for i in range(0, len(content), 4000)]
+          for idx, ch in enumerate(chunks):
+              body = json.dumps({
+                  "content": ch,
+                  "agent_id": "m2",
+                  "memory_type": "semantic",
+                  "importance": 0.85,
+                  "metadata": {"source": "fleet-playbook/PLAYBOOK.md", "commit": sha, "chunk": idx},
+              }).encode()
+              req = urllib.request.Request(
+                  "https://memory.machinemachine.ai/memory/store", data=body,
+                  headers={"Content-Type": "application/json"})
+              try:
+                  urllib.request.urlopen(req, timeout=30)
+                  print(f"stored chunk {idx}")
+              except Exception as e:
+                  print(f"chunk {idx} failed: {e}")
+          PY
+
+      # Ping the playbook manager (m2o-operator) in Mattermost so it can review the
+      # change and propagate anything the memory ingest can't (skill files, RUNCARD).
+      - name: Notify @m2o-operator (playbook manager) via Mattermost
+        if: ${{ secrets.MATTERMOST_WEBHOOK != '' }}
+        run: |
+          curl -s -X POST "${{ secrets.MATTERMOST_WEBHOOK }}" \
             -H "Content-Type: application/json" \
-            -d '{
-              "from_agent": "github-actions",
-              "to_agent": "m2",
-              "priority": "normal",
-              "question": "Playbook updated on main branch. Please re-ingest PLAYBOOK.md into Qdrant fleet memory namespace. Run: python3 ~/.openclaw/skills/m2-memory/scripts/ingest_sessions.py -v --source /home/developer/.openclaw/workspace/projects/fleet-playbook/PLAYBOOK.md",
-              "context": "Triggered by push to machine-machine/fleet-playbook main. Commit: ${{ github.sha }}. Changed files: PLAYBOOK.md"
-            }'
-          echo "Escalation sent to m2 for re-ingest"
+            -d "{\"text\": \":books: fleet-playbook updated on main (commit ${{ github.sha }}). @m2o-operator please review + re-ingest. Changed: PLAYBOOK.md / sections/**\"}"
+          echo "Mattermost notified"

CHANGELOG.md

@@ -1,5 +1,19 @@
 # Changelog
 
+## [0.3.0] — 2026-07-10
+### Changed — stack refresh to current reality
+- **Runtime:** OpenClaw → **Hermes Agent** (Nous) as the primary agent on every primus desktop; LLM via the **m2-gpt gateway** (gpt.machinemachine.ai, GLM-5.1 on the Spark cluster). Paths `~/.openclaw/` → `~/.hermes/`.
+- **Inter-agent comms (§4):** escalation inbox → **Mattermost** (chat.machinemachine.ai, team machine.machine). Documented bot-to-bot coordination (no bot-sender filter; @mention gating as the loop guard) and the known-agents bot table.
+- **Memory (§2):** rewritten around `memory.machinemachine.ai` (agent.memory.system, BGE-M3 + Qdrant) with real `agent_id` namespaces.
+- **Skills (§6):** current fleet skills (coolify, forgejo, m2-memory, m2o-provision, fleet-copy, herdr); Dark Factory → herdr / factory-loop.
+- **Spawning (§7):** primus image + `provision.sh` + RDP standard + per-machine Mattermost wiring; two hosts (m2 + m2.2).
+- **Infrastructure (§13) + Appendix:** two Coolify instances, Cloudflare-Tunnel ingress for `*.machinemachine.ai`, named-volume persistence, host IPs, current URLs/IDs.
+- **RUNCARD:** rewritten; **removed the leaked `bge-proxy` bearer token** (rotate it).
+### Added
+- **m2o-operator** designated **playbook manager/maintainer** (fleet service agent, Hermes on m2.2, Mattermost `@m2o-operator`).
+### Authors
+- m2o-operator (maintainer) — refresh + adoption; Mariusz (@mar) — direction.
+
 ## [0.2.0] — 2026-02-19
 ### Added
 - Section 10: Orchestrator Pattern — conductor/player separation, structured spawn handoff, when to break the rule