🐾 feat(library): add 'email' library type

Kairos is about to provision per-user app-managed (workspace) libraries for
synced mail, and none of the existing types fit correspondence. New entry in
LIBRARY_TYPE_DEFAULTS mirrors journal's entry-level chunking (one message ≈
one entry) with instructions tuned for email: correspondents, dates,
requests/commitments, and quoted-reply context. Registered in the Library
node choices and the API serializer choice list; forms and
load_library_types derive from the registry, and the compose init sidecar
re-runs the seeder on deploy.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.env.example

@@ -91,6 +91,19 @@ SPELUNKER_S3_REGION_NAME=us-east-1
 SPELUNKER_S3_USE_SSL=True
 SPELUNKER_S3_VERIFY=True
 
+# --- Kairos S3 (cross-bucket reads for ingest, source="kairos-mail") ---
+# Consumed by: worker only
+# Kairos renders synced mail to text documents in its own bucket and posts
+# ingest requests with source="kairos-mail". These creds should be scoped
+# read-only to the Kairos bucket in your secret manager.
+KAIROS_S3_ENDPOINT_URL=https://nyx.helu.ca:8555
+KAIROS_S3_ACCESS_KEY_ID=
+KAIROS_S3_SECRET_ACCESS_KEY=
+KAIROS_S3_BUCKET_NAME=kairos
+KAIROS_S3_REGION_NAME=us-east-1
+KAIROS_S3_USE_SSL=True
+KAIROS_S3_VERIFY=True
+
 # --- Celery / RabbitMQ (Oberon) ---------------------------------------------
 # Consumed by: app (producer), worker (consumer). NOT mcp.
 # Remember to percent-encode any password characters that have meaning in a

.gitea/workflows/cve-scan-docker-build.yml

@@ -73,8 +73,6 @@ jobs:
         with:
           images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
           tags: |
-            type=ref,event=branch
-            type=ref,event=pr
             type=sha,prefix=
             type=raw,value=latest,enable=${{ gitea.ref == 'refs/heads/main' }}
 

docker-compose.yaml

@@ -346,6 +346,13 @@ services:
       - SPELUNKER_S3_REGION_NAME=${SPELUNKER_S3_REGION_NAME}
       - SPELUNKER_S3_USE_SSL=${SPELUNKER_S3_USE_SSL}
       - SPELUNKER_S3_VERIFY=${SPELUNKER_S3_VERIFY}
+      - KAIROS_S3_ENDPOINT_URL=${KAIROS_S3_ENDPOINT_URL}
+      - KAIROS_S3_ACCESS_KEY_ID=${KAIROS_S3_ACCESS_KEY_ID}
+      - KAIROS_S3_SECRET_ACCESS_KEY=${KAIROS_S3_SECRET_ACCESS_KEY}
+      - KAIROS_S3_BUCKET_NAME=${KAIROS_S3_BUCKET_NAME}
+      - KAIROS_S3_REGION_NAME=${KAIROS_S3_REGION_NAME}
+      - KAIROS_S3_USE_SSL=${KAIROS_S3_USE_SSL}
+      - KAIROS_S3_VERIFY=${KAIROS_S3_VERIFY}
       # Celery / RabbitMQ
       - CELERY_BROKER_URL=${CELERY_BROKER_URL}
       - CELERY_RESULT_BACKEND=${CELERY_RESULT_BACKEND}
@@ -371,7 +378,12 @@ services:
     volumes:
       - media:/mnt/media
     healthcheck:
-      test: ["CMD", "celery", "-A", "mnemosyne", "inspect", "ping", "-d", "celery@$$HOSTNAME"]
+      # No -d destination: exec-form CMD has no shell, so $$HOSTNAME never
+      # expanded and the literal "celery@$HOSTNAME" matched no node → every
+      # check failed. There's one worker per container, so an unfiltered ping
+      # (any node replies = healthy) is correct. -t gives the reply room to
+      # round-trip through the broker on Oberon (~450ms observed) under jitter.
+      test: ["CMD", "celery", "-A", "mnemosyne", "inspect", "ping", "-t", "8"]
       interval: 60s
       timeout: 10s
       retries: 3

mnemosyne/library/api/serializers.py

@@ -15,6 +15,7 @@ LIBRARY_TYPE_CHOICES = [
     "film",
     "art",
     "journal",
+    "email",
     "business",
     "finance",
 ]

mnemosyne/library/content_types.py

@@ -241,6 +241,38 @@ LIBRARY_TYPE_DEFAULTS = {
             "4) The commercial purpose — positioning, pricing, capability demonstration."
         ),
     },
+    "email": {
+        "chunking_config": {
+            "strategy": "entry_level",
+            "chunk_size": 512,
+            "chunk_overlap": 32,
+            "respect_boundaries": ["message", "quote", "paragraph"],
+        },
+        "embedding_instruction": (
+            "Represent this email message for retrieval. "
+            "Focus on the sender, recipients, subject, dates, requests and "
+            "commitments made, and the people, organizations, and events discussed."
+        ),
+        "reranker_instruction": (
+            "Re-rank email messages based on relevance to the query. "
+            "Prioritize messages matching the correspondents, subject matter, "
+            "time period, and any specific commitments or requests mentioned."
+        ),
+        "llm_context_prompt": (
+            "The following excerpts are from personal email correspondence. "
+            "This is private content — answer with discretion. Attribute "
+            "statements to their senders, note dates, and distinguish what was "
+            "asked from what was agreed. Quoted text below a reply is earlier "
+            "context, not the sender's own words."
+        ),
+        "vision_prompt": (
+            "Analyze this image from an email message. Identify:\n"
+            "1) Image type (photograph, screenshot, scanned document, chart, signature graphic).\n"
+            "2) What it depicts — people, places, documents, data.\n"
+            "3) Any visible text, dates, or figures.\n"
+            "4) Its role in the message — attachment content, inline illustration, or boilerplate."
+        ),
+    },
     "finance": {
         "chunking_config": {
             "strategy": "section_aware",
@@ -282,7 +314,7 @@ def get_library_type_config(library_type):
 
     Args:
         library_type: One of 'fiction', 'nonfiction', 'technical', 'music',
-                      'film', 'art', 'journal', 'business', 'finance'
+                      'film', 'art', 'journal', 'email', 'business', 'finance'
 
     Returns:
         dict with keys: chunking_config, embedding_instruction,

mnemosyne/library/models.py

@@ -77,6 +77,7 @@ class Library(StructuredNode):
             "film": "Film",
             "art": "Art",
             "journal": "Journal",
+            "email": "Email",
             "business": "Business",
             "finance": "Finance",
         },

mnemosyne/library/templates/library/library_list.html

@@ -15,6 +15,18 @@
     </div>
 </div>
 
+<form method="get" class="mb-4 flex flex-wrap gap-3 items-end">
+    <div class="form-control">
+        <label class="label"><span class="label-text">Scope</span></label>
+        <select name="scope" class="select select-bordered select-sm">
+            <option value="all"      {% if scope == "all" %}selected{% endif %}>All libraries</option>
+            <option value="global"   {% if scope == "global" %}selected{% endif %}>Global only</option>
+            <option value="daedalus" {% if scope == "daedalus" %}selected{% endif %}>Daedalus workspaces only</option>
+        </select>
+    </div>
+    <button type="submit" class="btn btn-sm btn-outline">Filter</button>
+</form>
+
 {% if error %}
 <div class="alert alert-warning mb-4">
     <span>{{ error }}</span>
@@ -53,8 +65,12 @@
 {% else %}
     {% if not error %}
     <div class="text-center py-12 opacity-60">
+        {% if scope == "all" %}
         <p class="text-lg">No libraries yet.</p>
         <p class="mt-2">Create your first library to get started.</p>
+        {% else %}
+        <p class="text-lg">No libraries match this filter.</p>
+        {% endif %}
     </div>
     {% endif %}
 {% endif %}

mnemosyne/library/tests/test_content_types.py

@@ -21,6 +21,7 @@ class LibraryTypeDefaultsTests(TestCase):
         "film",
         "art",
         "journal",
+        "email",
         "business",
         "finance",
     }

mnemosyne/library/tests/test_views.py

@@ -0,0 +1,92 @@
+"""Tests for the library CRUD HTML views.
+
+Currently covers ``library_list``'s Daedalus-workspace scope filter. The
+view loads every ``Library`` node from Neo4j and narrows it by a ``scope``
+GET param (``all`` / ``global`` / ``daedalus``). These tests stub out
+Neo4j entirely — patching ``neo4j_available`` and injecting a fake
+``Library`` class via ``sys.modules`` — so they assert on the queryset
+``.filter(...)`` call the view makes and the context it renders, not on
+real graph behaviour. Mirrors the mocking style in
+``test_search_views_admin_scope.py``.
+"""
+
+from __future__ import annotations
+
+from types import SimpleNamespace
+from unittest.mock import MagicMock, patch
+
+from django.contrib.auth import get_user_model
+from django.test import TestCase
+from django.urls import reverse
+
+User = get_user_model()
+
+
+class LibraryListScopeFilterTests(TestCase):
+    """Cover the ``scope`` filter branches of ``library_list``."""
+
+    def setUp(self):
+        self.user = User.objects.create_user(
+            username="op", email="op@example.com", password="pw"
+        )
+        self.client.force_login(self.user)
+        self.url = reverse("library:library-list")
+
+    def _fake_library_cls(self):
+        """Return (Library stub, nodes mock) where ``nodes`` chains fluently.
+
+        ``Library.nodes`` → ``.filter(...)`` → ``.order_by(...)`` all return
+        the same MagicMock so the view's queryset building works regardless
+        of which branch it takes, and ``.filter`` records its kwargs.
+        """
+        fake_nodes = MagicMock()
+        fake_nodes.filter.return_value = fake_nodes
+        fake_nodes.order_by.return_value = []
+        return SimpleNamespace(nodes=fake_nodes), fake_nodes
+
+    def _get(self, fake_library_cls, **params):
+        with patch("library.views.neo4j_available", return_value=True), \
+             patch.dict(
+                 "sys.modules",
+                 {"library.models": SimpleNamespace(Library=fake_library_cls)},
+             ):
+            return self.client.get(self.url, params)
+
+    def test_default_scope_is_all_and_does_not_filter(self):
+        fake_cls, fake_nodes = self._fake_library_cls()
+        response = self._get(fake_cls)
+
+        self.assertEqual(response.status_code, 200)
+        self.assertEqual(response.context["scope"], "all")
+        fake_nodes.filter.assert_not_called()
+        fake_nodes.order_by.assert_called_once_with("name")
+
+    def test_global_scope_filters_workspace_isnull_true(self):
+        fake_cls, fake_nodes = self._fake_library_cls()
+        response = self._get(fake_cls, scope="global")
+
+        self.assertEqual(response.context["scope"], "global")
+        fake_nodes.filter.assert_called_once_with(workspace_id__isnull=True)
+
+    def test_daedalus_scope_filters_workspace_isnull_false(self):
+        fake_cls, fake_nodes = self._fake_library_cls()
+        response = self._get(fake_cls, scope="daedalus")
+
+        self.assertEqual(response.context["scope"], "daedalus")
+        fake_nodes.filter.assert_called_once_with(workspace_id__isnull=False)
+
+    def test_unknown_scope_does_not_filter(self):
+        """An unexpected scope value degrades to the unfiltered list."""
+        fake_cls, fake_nodes = self._fake_library_cls()
+        response = self._get(fake_cls, scope="bogus")
+
+        self.assertEqual(response.context["scope"], "bogus")
+        fake_nodes.filter.assert_not_called()
+
+    def test_neo4j_unavailable_sets_error_and_empty_list(self):
+        with patch("library.views.neo4j_available", return_value=False):
+            response = self.client.get(self.url)
+
+        self.assertEqual(response.status_code, 200)
+        self.assertEqual(list(response.context["libraries"]), [])
+        self.assertEqual(response.context["error"], "Neo4j is not available.")

mnemosyne/library/views.py

@@ -31,14 +31,20 @@ logger = logging.getLogger(__name__)
 
 @login_required
 def library_list(request):
-    """List all libraries."""
+    """List libraries, optionally filtered by Daedalus-workspace scope."""
+    scope = request.GET.get("scope", "all")
     libraries = []
     error = None
     if neo4j_available():
         try:
             from .models import Library
 
-            libraries = Library.nodes.order_by("name")
+            qs = Library.nodes
+            if scope == "daedalus":
+                qs = qs.filter(workspace_id__isnull=False)
+            elif scope == "global":
+                qs = qs.filter(workspace_id__isnull=True)
+            libraries = qs.order_by("name")
         except Exception as e:
             error = f"Could not connect to Neo4j: {e}"
             logger.error(error)
@@ -47,7 +53,7 @@ def library_list(request):
     return render(
         request,
         "library/library_list.html",
-        {"libraries": libraries, "error": error},
+        {"libraries": libraries, "error": error, "scope": scope},
     )
 
 

mnemosyne/mnemosyne/settings.py

@@ -268,6 +268,15 @@ SPELUNKER_S3_REGION_NAME = env("SPELUNKER_S3_REGION_NAME", default="us-east-1")
 SPELUNKER_S3_USE_SSL = env.bool("SPELUNKER_S3_USE_SSL", default=False)
 SPELUNKER_S3_VERIFY = env.bool("SPELUNKER_S3_VERIFY", default=True)
 
+# Kairos renders synced mail to text documents in its own bucket.
+KAIROS_S3_ENDPOINT_URL = env("KAIROS_S3_ENDPOINT_URL", default="")
+KAIROS_S3_ACCESS_KEY_ID = env("KAIROS_S3_ACCESS_KEY_ID", default="")
+KAIROS_S3_SECRET_ACCESS_KEY = env("KAIROS_S3_SECRET_ACCESS_KEY", default="")
+KAIROS_S3_BUCKET_NAME = env("KAIROS_S3_BUCKET_NAME", default="kairos")
+KAIROS_S3_REGION_NAME = env("KAIROS_S3_REGION_NAME", default="us-east-1")
+KAIROS_S3_USE_SSL = env.bool("KAIROS_S3_USE_SSL", default=False)
+KAIROS_S3_VERIFY = env.bool("KAIROS_S3_VERIFY", default=True)
+
 # Registry keyed by the ingest `source` field. Unknown/blank sources fall
 # back to "daedalus" for backwards compatibility.
 SOURCE_S3_BUCKETS = {
@@ -289,6 +298,15 @@ SOURCE_S3_BUCKETS = {
         "use_ssl": SPELUNKER_S3_USE_SSL,
         "verify": SPELUNKER_S3_VERIFY,
     },
+    "kairos-mail": {
+        "endpoint_url": KAIROS_S3_ENDPOINT_URL,
+        "access_key_id": KAIROS_S3_ACCESS_KEY_ID,
+        "secret_access_key": KAIROS_S3_SECRET_ACCESS_KEY,
+        "bucket_name": KAIROS_S3_BUCKET_NAME,
+        "region_name": KAIROS_S3_REGION_NAME,
+        "use_ssl": KAIROS_S3_USE_SSL,
+        "verify": KAIROS_S3_VERIFY,
+    },
 }
 
 # --- Celery / RabbitMQ ---