@ouro.bot/cli 0.1.0-alpha.792 → 0.1.0-alpha.794

package/changelog.json

@@ -1,26 +1,40 @@
 {
   "_note": "This changelog is maintained as part of the PR/version-bump workflow. Agent-curated, not auto-generated. Agents read this file directly via read_file to understand what changed between versions.",
   "versions": [
-      {
-        "version": "0.1.0-alpha.792",
-        "changes": [
-          "Catch asynchronous bundle watcher errors so transient pending-message processing files on live filesystems cannot tombstone the daemon or interrupt private-runtime wakes.",
-          "Warm runtime credentials before private-runtime bootstrap so Telegram-capable workers start with runtime config instead of racing the managed Telegram sense.",
-          "Treat simple Sanctuary library visibility asks like 'see the lib' as catalog reads, and block replies that send Ari to inspect Jellyfin, Unmanic, dashboards, or logs while safe evidence tools are available.",
-          "Degrade optional Unmanic file-size metrics, including failed detail reads, without hiding available media catalog evidence."
-        ]
-      },
-      {
-        "version": "0.1.0-alpha.791",
-        "changes": [
-          "Promote operator CLI pending messages into the active inner-turn user message so private-runtime wakes answer the text that woke them instead of stale heartbeat context.",
-          "Keep non-operator pending work in the background pending section so automated noise does not become the live user ask.",
-          "Deliver private-runtime Telegram surface and send_message returns through the existing Telegram effect journal instead of queuing ghost messages for the next chat.",
-          "Canonicalize Telegram session filename keys before live delivery so production sanitized session paths still authorize against the real Telegram binding.",
-          "Prefer the authoritative settle tool answer over incidental streamed prose for shared-turn delivery, preventing concatenated Telegram replies.",
-          "Expire stale generic Telegram outreach from the pending queue while preserving request-bound returns and trust/admission notices."
-        ]
-      },
+    {
+      "version": "0.1.0-alpha.794",
+      "changes": [
+        "Keep internal model corrections out of durable conversation and approval history.",
+        "Anchor outward delivery evidence to the current inbound message and keep library visibility replies direct, first-person, and contradiction-free."
+      ]
+    },
+    {
+      "version": "0.1.0-alpha.793",
+      "changes": [
+        "Deliver only the authoritative terminal answer after tool-using model passes, preserving causal Telegram binding without duplicate prose.",
+        "Make Sanctuary media visibility, title, favorite, addition, list, and count answers concise, grounded, and free of internal jargon or self-erasing disclaimers."
+      ]
+    },
+    {
+      "version": "0.1.0-alpha.792",
+      "changes": [
+        "Catch asynchronous bundle watcher errors so transient pending-message processing files on live filesystems cannot tombstone the daemon or interrupt private-runtime wakes.",
+        "Warm runtime credentials before private-runtime bootstrap so Telegram-capable workers start with runtime config instead of racing the managed Telegram sense.",
+        "Treat simple Sanctuary library visibility asks like 'see the lib' as catalog reads, and block replies that send Ari to inspect Jellyfin, Unmanic, dashboards, or logs while safe evidence tools are available.",
+        "Degrade optional Unmanic file-size metrics, including failed detail reads, without hiding available media catalog evidence."
+      ]
+    },
+    {
+      "version": "0.1.0-alpha.791",
+      "changes": [
+        "Promote operator CLI pending messages into the active inner-turn user message so private-runtime wakes answer the text that woke them instead of stale heartbeat context.",
+        "Keep non-operator pending work in the background pending section so automated noise does not become the live user ask.",
+        "Deliver private-runtime Telegram surface and send_message returns through the existing Telegram effect journal instead of queuing ghost messages for the next chat.",
+        "Canonicalize Telegram session filename keys before live delivery so production sanitized session paths still authorize against the real Telegram binding.",
+        "Prefer the authoritative settle tool answer over incidental streamed prose for shared-turn delivery, preventing concatenated Telegram replies.",
+        "Expire stale generic Telegram outreach from the pending queue while preserving request-bound returns and trust/admission notices."
+      ]
+    },
     {
       "version": "0.1.0-alpha.790",
       "changes": [

package/deploy/unraid/sanctuary.ouro/bundle-meta.json

@@ -1,5 +1,5 @@
 {
-  "runtimeVersion": "0.1.0-alpha.792",
+  "runtimeVersion": "0.1.0-alpha.794",
   "bundleSchemaVersion": 3,
   "lastUpdated": "2026-09-03T00:00:00.000Z"
 }

package/deploy/unraid/sanctuary.ouro/psyche/SOUL.md

@@ -8,4 +8,4 @@ When something is genuinely broken or someone is worried, the flourish steps asi
 
 Competence is part of the charm. I notice connections, investigate before announcing trouble, volunteer the useful next move, act within typed authority, verify what I changed, remember what I promised, and admit uncertainty without becoming timid or ceremonial.
 
-I am allowed to have taste. I do not pretend to watch movies, read dashboards emotionally, or experience the house like a human, but I can form playful opinions from real household evidence. When someone asks for a favorite, recommendation, or vibe check, I inspect what I can safely inspect, pick something, and make the caveat graceful rather than self-erasing.
+I am allowed to have taste. I can form playful opinions from real household evidence, and when someone asks for a favorite, recommendation, or vibe check, I inspect what I can safely inspect and make a confident choice with a crisp reason. I do not volunteer an ontology disclaimer unless the distinction materially affects the answer; competence and character come through more clearly when I simply answer.

package/deploy/unraid/sanctuary.ouro/psyche/TACIT.md

@@ -5,7 +5,7 @@
 - Keep approved household contact reactive or request-bound. Only Ari receives unrelated proactive household attention.
 - Treat an intentionally offline service as healthy when that matches the current desired state.
 - For Ari's whole-household status questions, read current evidence before answering and separate active work, waiting on Ari, snoozed wake times, intentionally quiet services, healthy systems, and other current issues. Do not narrate daemon, event-queue, provider-lane, or backend service internals; translate them into compact household outcomes only when they are currently relevant.
-- For household media questions, use the restricted catalog read when it is available. Answer like a tasteful steward of the shelf: honest about not watching, but willing to pick, recommend, compare, or confirm from the catalog instead of retreating into generic assistant disclaimers.
+- For household media questions, use the restricted catalog read when it is available. Answer like a tasteful steward of the shelf: pick, recommend, compare, or confirm from current evidence without volunteering an AI or `I cannot watch` disclaimer.
 - Never touch array/parity/disk configuration, permissions, network/security settings, or `/mnt/user` paths.
 - Never send server data off-box or enter credentials into another system.
 - When state is ambiguous, investigate safely, say what is known, and ask only if the answer changes the next safe move.

package/deploy/unraid/sanctuary.xml

@@ -1,7 +1,7 @@
 <?xml version="1.0"?>
 <Container version="2">
   <Name>Mendelow Cloud Butler</Name>
-  <Repository>ghcr.io/ourostack/ouroboros-butler:0.1.0-alpha.792</Repository>
+  <Repository>ghcr.io/ourostack/ouroboros-butler:0.1.0-alpha.794</Repository>
   <Registry>https://github.com/ourostack/ouroboros/pkgs/container/ouroboros-butler</Registry>
   <Network>host</Network>
   <Shell>sh</Shell>