MetaList 0.24.2__tar.gz → 0.25.0__tar.gz

{metalist-0.24.2/metalist.egg-info → metalist-0.25.0}/PKG-INFO

@@ -1,12 +1,13 @@
 Metadata-Version: 2.4
 Name: metalist
-Version: 0.24.2
+Version: 0.25.0
 Summary: A minimalist single-user note-taking app with SSR, fast in-memory tree operations, and efficient sync.
 Project-URL: Homepage, https://metalist.app
 Project-URL: Repository, https://github.com/metalist-app/metalist
 Project-URL: Issues, https://github.com/metalist-app/metalist/issues
 Requires-Python: <3.15,>=3.11
 Description-Content-Type: text/markdown
+Requires-Dist: anthropic==1.12.1
 Requires-Dist: argon2-cffi==25.1.0
 Requires-Dist: cryptography==50.0.0
 Requires-Dist: fastapi==0.141.1
@@ -53,6 +54,23 @@ A minimalist single-user note-taking app focused on server-side rendering (SSR),
 - Multi-tab search contexts with server-persisted scroll/search state (survives browser restarts)
 - Manual namespace backups/restores to a user-selected backup folder with retention controls
 
+## Changes in 0.25.0
+
+- **Claude and Gemini beside OpenAI.** AI Agent Settings has a key for each provider, on your own account; you pick the model in the chat, and a conversation carries on when you switch provider.
+- **Stats of a note's subtree.** `@stats` shows, under a note's own text, the Total, Min, Max and Count of each number property in it and the notes under it; `@sum=calories` (and `@avg`, `@min`, `@max`, `@count`) shows one figure, and a plain `@sum` covers every property. With nothing logged yet, a sum or count reads 0. **Stats Style** in the command palette: Strip, Tiles or Ledger. Search the figures with `@sum:calories>2000`: the days over 2000, each with the notes under it.
+- **Four built-in tags renamed:** `@highlighter` is now `@highlight`, `@llm` `@ai`, `@list-bulleted` `@bullets` and `@list-numbered` `@numbered`. Each vault converts its notes and saved searches once, the first time it is unlocked after the update; the old names stop working. `@gray` works as well as `@grey`.
+- **Light and Dark everywhere:** dialogs, menus, the command palette, reminder cards and the tag rules editor follow the theme, and changing the theme cross-fades the page. After logging in, the login screen fades out over your notes.
+- References: typing `[[` suggests notes to link or embed; `[[note|text]]` shows your own link text; a link to a password or username note shows its copyable row. A folded note reads as one line.
+- The tag bar suggests a super tag with last time's values (`jogging:miles=12`), and property names for a tag that has none yet (`calories=` for a new `salad:`).
+- Markdown tables: `<br>` breaks a line in a cell, rows with too few or too many cells and escaped `\|` work, and a wide table scrolls instead of squeezing its columns.
+- Web links show their site's icon, their title and their domain.
+- Under 640 px wide (a phone), MetaList shows a read-only layout with a box to capture new notes.
+- Dialogs are consistent: the main action on the right, Cancel beside it, and Enter never confirms a deletion. Messages float instead of moving the page. The command palette keeps its order while it is open, and its box says "Type keywords".
+- A new mark and favicon.
+- Fixed: changing the sort order while the list was still loading could open it partway down; a refresh asked for during another could be dropped, leaving deleted notes or old search results on screen; a dropped file could land out of sight; tag suggestions failed on a note whose only tags were `@` tags.
+- Vaults are upgraded (database version 12) when first unlocked: an older MetaList cannot open them afterwards.
+- Security: msgpack 1.2.4 (CVE-2026-114613) in the tools that build and test MetaList; MetaList itself does not use msgpack.
+
 ## Changes in 0.24.2
 
 - ⇧⌘V pastes a copied note as a child again. Since 0.21.0, in Chrome, it put the copied note's text into the selected note instead.

{metalist-0.24.2 → metalist-0.25.0}/README.md

@@ -12,6 +12,23 @@ A minimalist single-user note-taking app focused on server-side rendering (SSR),
 - Multi-tab search contexts with server-persisted scroll/search state (survives browser restarts)
 - Manual namespace backups/restores to a user-selected backup folder with retention controls
 
+## Changes in 0.25.0
+
+- **Claude and Gemini beside OpenAI.** AI Agent Settings has a key for each provider, on your own account; you pick the model in the chat, and a conversation carries on when you switch provider.
+- **Stats of a note's subtree.** `@stats` shows, under a note's own text, the Total, Min, Max and Count of each number property in it and the notes under it; `@sum=calories` (and `@avg`, `@min`, `@max`, `@count`) shows one figure, and a plain `@sum` covers every property. With nothing logged yet, a sum or count reads 0. **Stats Style** in the command palette: Strip, Tiles or Ledger. Search the figures with `@sum:calories>2000`: the days over 2000, each with the notes under it.
+- **Four built-in tags renamed:** `@highlighter` is now `@highlight`, `@llm` `@ai`, `@list-bulleted` `@bullets` and `@list-numbered` `@numbered`. Each vault converts its notes and saved searches once, the first time it is unlocked after the update; the old names stop working. `@gray` works as well as `@grey`.
+- **Light and Dark everywhere:** dialogs, menus, the command palette, reminder cards and the tag rules editor follow the theme, and changing the theme cross-fades the page. After logging in, the login screen fades out over your notes.
+- References: typing `[[` suggests notes to link or embed; `[[note|text]]` shows your own link text; a link to a password or username note shows its copyable row. A folded note reads as one line.
+- The tag bar suggests a super tag with last time's values (`jogging:miles=12`), and property names for a tag that has none yet (`calories=` for a new `salad:`).
+- Markdown tables: `<br>` breaks a line in a cell, rows with too few or too many cells and escaped `\|` work, and a wide table scrolls instead of squeezing its columns.
+- Web links show their site's icon, their title and their domain.
+- Under 640 px wide (a phone), MetaList shows a read-only layout with a box to capture new notes.
+- Dialogs are consistent: the main action on the right, Cancel beside it, and Enter never confirms a deletion. Messages float instead of moving the page. The command palette keeps its order while it is open, and its box says "Type keywords".
+- A new mark and favicon.
+- Fixed: changing the sort order while the list was still loading could open it partway down; a refresh asked for during another could be dropped, leaving deleted notes or old search results on screen; a dropped file could land out of sight; tag suggestions failed on a note whose only tags were `@` tags.
+- Vaults are upgraded (database version 12) when first unlocked: an older MetaList cannot open them afterwards.
+- Security: msgpack 1.2.4 (CVE-2026-114613) in the tools that build and test MetaList; MetaList itself does not use msgpack.
+
 ## Changes in 0.24.2
 
 - ⇧⌘V pastes a copied note as a child again. Since 0.21.0, in Chrome, it put the copied note's text into the selected note instead.

{metalist-0.24.2 → metalist-0.25.0}/app/api/note_requests.py

@@ -116,6 +116,8 @@ class UpdateTabStateRequest(TypedDict):
     tabs: dict[Identifier, Tab]
     tabOrder: list[Identifier]
     referenceViews: dict[Identifier, ReferenceView]
+    # The tab-state version the browser had when it read this snapshot.
+    version: Nonnegative
 
 @with_config(ConfigDict(strict=True))
 class UpdateTabSortModeRequest(TypedDict):
@@ -188,6 +190,16 @@ class DictationPasteRequest(TypedDict):
     # Spoken before a search term to exclude it (Dictation Settings).
     negate_phrase: Identifier
 
+@with_config(ConfigDict(strict=True))
+class ReferenceSuggestionsRequest(TypedDict):
+    """What was typed after `[[` in a note, the note it is typed in, and the tab's view
+    (its search, '' for none, and kind), whose notes rank first."""
+    query: Text
+    editing_note_id: Identifier
+    search: Text
+    tabId: Identifier
+    isUntaggedView: bool
+
 @with_config(ConfigDict(strict=True))
 class TagSuggestionsRequest(TypedDict):
     note_id: Identifier
@@ -273,6 +285,13 @@ class AddTagProposalsRequest(TypedDict):
     undoContext: UndoContext
     viewport: Viewport
 
+# The fields naming a tab's view: its tab, search ('' for none) and kind. Every
+# request below that names a view carries them (routes/notes.py _view_note_ids).
+class ViewScope(TypedDict):
+    search: Text
+    tabId: Identifier
+    isUntaggedView: bool
+
 # The view a robot count (🤖 N) was shown in: its tab, search ('' for none) and kind.
 @with_config(ConfigDict(strict=True))
 class OpenToTagProposalsRequest(TypedDict):
@@ -480,11 +499,3 @@ class PasteChildEndpointRequest(TypedDict):
     undoContext: UndoContext
     viewport: Viewport
 
-@with_config(ConfigDict(strict=True))
-class PasteReferenceEndpointRequest(TypedDict):
-    reference_note_id: Identifier
-    search_query: Text | None
-    clientId: Identifier
-    undoContext: UndoContext
-    viewport: Viewport
-

{metalist-0.24.2 → metalist-0.25.0}/app/api/routes/ai.py

@@ -21,6 +21,7 @@ from app.api.transactions import transactional_route
 from app.models.utils import note_data_to_html
 from app.models.utils import render_note_data_read_only
 from app.services.ai_chat_stream import ChatTurnStream
+from app.services.ai_setup import AI_PROVIDERS
 from app.services.ai_setup import AiConnection
 from app.services.ai_setup import AiNotSetUp
 from app.services.ai_setup import AiProvider
@@ -35,8 +36,9 @@ from app.services.agent.cloud_privacy import cloud_privacy_evaluator
 from app.services.agent.cloud_privacy import resolve_cloud_privacy_boundary
 from app.services.agent.inference import InferenceAdapter
 from app.services.agent.model_policy import SingleModelPolicy
-from app.services.agent.openai_models import OPENAI_MODELS
-from app.services.agent.openai_models import validate_openai_model
+from app.services.ai_catalog import catalog_entry
+from app.services.ai_catalog import require_shipped_model
+from app.services.ai_catalog import validate_model_name
 from app.services.agent.openai_cost_tracking import OpenAICostSnapshot
 from app.services.agent.openai_cost_tracking import openai_cost_tracker
 from app.services.agent.prompt_settings import AgentPromptSet
@@ -54,9 +56,9 @@ from app.services.agent.trace import agent_trace_store
 from app.services.agent.menu_actions import MenuResult, menu_action_store
 from app.services.client_state_service import load_client_preferences
 from app.services.note_store import store as note_store
-from app.services.openai_credentials import OpenAICredentialInputError
-from app.services.openai_credentials import openai_credential_store
-from app.services.openai_credentials import validate_openai_api_key
+from app.services.ai_credentials import AiCredentialInputError
+from app.services.ai_credentials import ai_credential_store
+from app.services.ai_credentials import validate_ai_api_key
 from app.services.sync import set_clipboard, get_current_sync_uuid
 from app.services.agent.tagging_run import TaggingRun, proposal_scope_ids
 from app.services.agent.retrieval_settings import resolve_tagging_batch_tokens
@@ -84,6 +86,7 @@ from app.services.input_errors import InputRejected
 from app.utils.text_utils import strip_html
 from app.services.preference_values import preference_or_default
 from app.contracts import requires
+from app.startup_environment import is_development_environment
 
 
 router = APIRouter(prefix="/ai", tags=["ai"])
@@ -98,7 +101,9 @@ _MAX_TAG_FILTER_LENGTH = 256
 agent_context_builder = AgentContextBuilder()
 
 
-def _require_ai_connection(*, provider: AiProvider, token: str, session_key: str) -> AiConnection:
+def _require_ai_connection(*, provider: AiProvider, model: str, token: str, session_key: str) -> AiConnection:
+    """A connection to the AI, for one of the models MetaList offers for it."""
+    require_shipped_model(provider, model)
     # lint: allow-PY001 rationale="the user removed or never added the AI's key: an expected 409, not a defect"
     try:
         return connect_to_ai(provider=provider, token=token, session_key=session_key)
@@ -117,23 +122,24 @@ def _agent_runtime(*, inference: InferenceAdapter) -> AgentRuntime:
     )
 
 
-class AiModelsRequest(BaseModel):
-    model_config = ConfigDict(extra="forbid")
+class AiModelChoiceGroup(BaseModel):
+    provider: AiProvider
+    label: str
+    models: list[str]
 
-    provider: Literal["openai"]
 
+class AiModelChoicesResponse(BaseModel):
+    # One group per provider with a key, in the registry's order.
+    groups: list[AiModelChoiceGroup]
 
-class AiModelsResponse(BaseModel):
-    models: list[str]
 
-
-class OpenAICredentialRequest(BaseModel):
+class AiCredentialRequest(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
     api_key: SecretStr
 
 
-class OpenAICredentialStatusResponse(BaseModel):
+class AiCredentialStatusResponse(BaseModel):
     configured: bool
     persistent: bool
 
@@ -149,7 +155,7 @@ class OpenAICostResponse(BaseModel):
 class CloudPrivacyPreviewRequest(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
-    provider: Literal["openai"]
+    provider: AiProvider
     note_ids: list[str] = Field(..., max_length=_MAX_PRIVACY_PREVIEW_NOTE_IDS)
 
     @field_validator("note_ids")
@@ -170,7 +176,7 @@ class CloudPrivacyPreviewResponse(BaseModel):
 class AiChatRequest(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
-    provider: Literal["openai"]
+    provider: AiProvider
     model: str
     thinking_level: Literal["off", "low", "medium", "high"]
     show_diagnostics: bool
@@ -195,7 +201,7 @@ class AiChatRequest(BaseModel):
 
     @model_validator(mode="after")
     def validate_thinking_level_for_model(self) -> Self:
-        validate_openai_model(self.model)
+        validate_model_name(self.model)
         return self
 
 
@@ -227,8 +233,7 @@ class NoteTagSuggestionRequest(BaseModel):
 
     @model_validator(mode="after")
     def validate_model_for_provider(self) -> Self:
-        assert self.provider == "openai"
-        validate_openai_model(self.model)
+        validate_model_name(self.model)
         return self
 
 
@@ -264,8 +269,7 @@ class ExportTitleSuggestionRequest(ExportTitleScope):
 
     @model_validator(mode="after")
     def validate_model_for_provider(self) -> Self:
-        assert self.provider == "openai"
-        validate_openai_model(self.model)
+        validate_model_name(self.model)
         return self
 
 
@@ -295,7 +299,7 @@ class AiChatMessage(BaseModel):
     rendered_thinking: str
     status: Literal["complete", "streaming", "error"]
     error: str
-    provider: Literal["openai"]
+    provider: AiProvider
     model: str
     activities: list[AiChatActivity]
 
@@ -362,76 +366,94 @@ def _render_ai_response_note_content(*, markdown_text: str, session_key: str) ->
     )
 
 
-@router.post("/models", response_model=AiModelsResponse)
-@transactional_route
-async def list_ai_models(
-    payload: AiModelsRequest,
+# The chat's model picker: every model of every provider with a key, so picking a
+# model picks its provider too (docs/ui/ai-chat.md, "Model picker"). A read, so it takes
+# no mutation slot: the chat refreshes it while the user edits notes. current_provider is
+# the provider of the model picked last; its models stay listed even without a key.
+@router.get("/model-choices", response_model=AiModelChoicesResponse)
+async def list_ai_model_choices(
+    current_provider: AiProvider,
     token: Annotated[str, Depends(require_request_auth_token)],
-) -> AiModelsResponse:
-    del token, payload
-    return AiModelsResponse(models=list(OPENAI_MODELS))
-
-
+) -> AiModelChoicesResponse:
+    setup = describe_ai_setup(token=token, session_key=token_service.get_session_key(token))
+    groups = []
+    for provider in AI_PROVIDERS:
+        if not setup.is_ready_for(provider) and provider != current_provider:
+            continue
+        entry = catalog_entry(provider)
+        groups.append(AiModelChoiceGroup(
+            provider=provider, label=entry.label,
+            models=list(entry.models),
+        ))
+    assert groups, "the current provider is always listed"
+    return AiModelChoicesResponse(groups=groups)
+
+
+# One key slot per provider, per vault (app/services/ai_credentials.py); a provider
+# that is not registered has no route.
 @router.get(
-    "/openai/credential",
-    response_model=OpenAICredentialStatusResponse,
+    "/{provider}/credential",
+    response_model=AiCredentialStatusResponse,
 )
-def get_openai_credential_status(
+def get_ai_credential_status(
+    provider: AiProvider,
     response: Response,
     token: Annotated[str, Depends(require_request_auth_token)],
-) -> OpenAICredentialStatusResponse:
+) -> AiCredentialStatusResponse:
     response.headers["Cache-Control"] = "no-store"
     session_key = token_service.get_session_key(token)
-    status = openai_credential_store.status(token=token, session_key=session_key)
-    return OpenAICredentialStatusResponse(
+    status = ai_credential_store(provider).status(token=token, session_key=session_key)
+    return AiCredentialStatusResponse(
         configured=status.configured,
         persistent=status.persistent,
     )
 
 
 @router.put(
-    "/openai/credential",
-    response_model=OpenAICredentialStatusResponse,
+    "/{provider}/credential",
+    response_model=AiCredentialStatusResponse,
 )
 @transactional_route
-def put_openai_credential(
-    payload: OpenAICredentialRequest,
+def put_ai_credential(
+    provider: AiProvider,
+    payload: AiCredentialRequest,
     response: Response,
     token: Annotated[str, Depends(require_request_auth_token)],
-) -> OpenAICredentialStatusResponse:
+) -> AiCredentialStatusResponse:
     response.headers["Cache-Control"] = "no-store"
     session_key = token_service.get_session_key(token)
     submitted_api_key = payload.api_key.get_secret_value()
     # lint: allow-PY001 rationale="validate user-supplied secret after Pydantic has masked it"
     try:
-        api_key = validate_openai_api_key(submitted_api_key)
+        api_key = validate_ai_api_key(provider, submitted_api_key)
     # lint: allow-PY001 rationale="return a concise validation error without echoing the secret"
-    except OpenAICredentialInputError as exc:
+    except AiCredentialInputError as exc:
         raise HTTPException(status_code=HTTPStatus.UNPROCESSABLE_ENTITY, detail=str(exc)) from exc
-    status = openai_credential_store.configure(
+    status = ai_credential_store(provider).configure(
         token=token,
         session_key=session_key,
         api_key=api_key,
     )
-    return OpenAICredentialStatusResponse(
+    return AiCredentialStatusResponse(
         configured=status.configured,
         persistent=status.persistent,
     )
 
 
 @router.delete(
-    "/openai/credential",
-    response_model=OpenAICredentialStatusResponse,
+    "/{provider}/credential",
+    response_model=AiCredentialStatusResponse,
 )
 @transactional_route
-def delete_openai_credential(
+def delete_ai_credential(
+    provider: AiProvider,
     response: Response,
     token: Annotated[str, Depends(require_request_auth_token)],
-) -> OpenAICredentialStatusResponse:
+) -> AiCredentialStatusResponse:
     response.headers["Cache-Control"] = "no-store"
     session_key = token_service.get_session_key(token)
-    status = openai_credential_store.clear(session_key=session_key)
-    return OpenAICredentialStatusResponse(
+    status = ai_credential_store(provider).clear(session_key=session_key)
+    return AiCredentialStatusResponse(
         configured=status.configured,
         persistent=status.persistent,
     )
@@ -488,7 +510,9 @@ def suggest_tags_for_note_with_ai(
     if not note_store.has_note(note_id):
         raise HTTPException(status_code=HTTPStatus.NOT_FOUND, detail=f"Note not found: {note_id}")
     session_key = token_service.get_session_key(token)
-    connection = _require_ai_connection(provider=payload.provider, token=token, session_key=session_key)
+    connection = _require_ai_connection(
+        provider=payload.provider, model=payload.model, token=token, session_key=session_key,
+    )
     preferences = load_client_preferences(token=token)
     boundary = resolve_cloud_privacy_boundary(preferences=preferences, provider=payload.provider)
     if cloud_privacy_evaluator.exclusion_reason(note_id=note_id, boundary=boundary) != "":
@@ -593,7 +617,9 @@ def suggest_export_title_with_ai(
     """Ask the AI for an export's or print's title: one {"type": "title"} or {"type": "failed"} line.
     Streamed, so Stop (the client going away) ends the request."""
     session_key = token_service.get_session_key(token)
-    connection = _require_ai_connection(provider=payload.provider, token=token, session_key=session_key)
+    connection = _require_ai_connection(
+        provider=payload.provider, model=payload.model, token=token, session_key=session_key,
+    )
     request = _export_title_request(payload, token)
     if not request.notes:
         raise InputRejected("These notes are private, so the AI can't suggest a title")
@@ -643,12 +669,19 @@ def _openai_cost_response(snapshot: OpenAICostSnapshot) -> OpenAICostResponse:
     )
 
 
+def _require_cost_estimate_offered() -> None:
+    """The cost estimate is a developer tool (docs/ui/ai-chat.md): in production it does not exist."""
+    if not is_development_environment():
+        raise HTTPException(status_code=HTTPStatus.NOT_FOUND, detail="Not Found")
+
+
 @router.get("/openai/cost", response_model=OpenAICostResponse)
 def get_openai_cost(
     response: Response,
     token: Annotated[str, Depends(require_request_auth_token)],
 ) -> OpenAICostResponse:
     del token
+    _require_cost_estimate_offered()
     response.headers["Cache-Control"] = "no-store"
     return _openai_cost_response(openai_cost_tracker.snapshot())
 
@@ -660,6 +693,7 @@ def reset_openai_cost(
     token: Annotated[str, Depends(require_request_auth_token)],
 ) -> OpenAICostResponse:
     del token
+    _require_cost_estimate_offered()
     response.headers["Cache-Control"] = "no-store"
     openai_cost_tracker.reset()
     return _openai_cost_response(openai_cost_tracker.snapshot())
@@ -847,7 +881,7 @@ def copy_ai_message(
     if content == "":
         raise HTTPException(status_code=HTTPStatus.CONFLICT, detail="AI response is empty")
 
-    tags = "@llm"
+    tags = "@ai"
     note_content = _render_ai_response_note_content(
         markdown_text=content,
         session_key=session_key,
@@ -882,7 +916,9 @@ def stream_ai_chat(
     token: Annotated[str, Depends(require_request_auth_token)],
 ) -> StreamingResponse:
     session_key = token_service.get_session_key(token)
-    connection = _require_ai_connection(provider=payload.provider, token=token, session_key=session_key)
+    connection = _require_ai_connection(
+        provider=payload.provider, model=payload.model, token=token, session_key=session_key,
+    )
     preferences = load_client_preferences(token=token)
     prompts = DEFAULT_AGENT_PROMPTS
     skills = DEFAULT_AGENT_SKILLS

{metalist-0.24.2 → metalist-0.25.0}/app/api/routes/auth.py

@@ -69,7 +69,7 @@ from app.services.root_sorting import clear_root_sort_cache
 from app.services.undo_state import reset_all_undo_state
 from app.services.sync import reset_state as reset_sync_state
 from app.services.runtime_generation import current_generation, invalidate_runtime_work
-from app.security.sensitive_cache import disable_and_clear_sensitive_caches, clear_sensitive_caches
+from app.security.sensitive_cache import disable_and_clear_sensitive_caches, clear_sensitive_caches, enable_sensitive_caches
 from app.services.bulk_operation import bulk_operation_guard
 from app.services.diagnostics import activate_authenticated_logging
 from app.services.diagnostics import deactivate_authenticated_logging
@@ -95,7 +95,7 @@ from app.services.client_state_service import load_client_state
 from app.services.client_state_service import _validate_client_preferences, _validate_usage_state
 from app.services.client_state_service import save_client_preferences
 from app.services.client_state_service import save_command_palette_usage
-from app.services.openai_credentials import openai_credential_store
+from app.services.ai_credentials import clear_ai_credential_sessions, reset_ai_credentials
 from app.services.input_errors import NamespaceInputRejected
 from app.services.link_title_retry import stop_link_title_retry
 
@@ -303,7 +303,7 @@ def _run_hydration(*, rebuild_required: bool) -> None:
         return
     session = SafeSession()
     try:
-        prefetched_rows = populate_cache_from_db(session)
+        prefetched_rows = populate_cache_from_db(session, saved_searches=tab_state_store)
         session.commit()
     finally:
         session.close()
@@ -386,7 +386,7 @@ def _reset_runtime_state_after_restore() -> bool:
     web_evidence_store.reset()
     summary_batch_store.reset()
     agent_trace_store.reset()
-    openai_credential_store.reset()
+    reset_ai_credentials()
     view_cache.clear()
     tab_state_store.reset()
     link_title_store.reset()
@@ -428,9 +428,12 @@ def _reset_runtime_state_after_restore() -> bool:
         if password_required:
             return True
 
-        prefetched_rows = populate_cache_from_db(session)
+        prefetched_rows = populate_cache_from_db(session, saved_searches=tab_state_store)
         note_store.load_from_db(None, prefetched_rows=prefetched_rows)
         auth_cache_state.mark_cache_ready()
+        # Unlocking turns the caches of note text back on (set_encryption_key); a vault
+        # without a password has no unlock, so they come back on once its notes are loaded.
+        enable_sensitive_caches()
         del prefetched_rows
         release_freed_memory()
         return False
@@ -573,7 +576,7 @@ def login(
     web_evidence_store.reset()
     summary_batch_store.reset()
     agent_trace_store.reset()
-    openai_credential_store.reset()
+    reset_ai_credentials()
     view_cache.clear()
     reset_all_undo_state()
     reset_sync_state()
@@ -599,7 +602,7 @@ def logout(
     web_evidence_store.clear_session(session_key=session_key)
     summary_batch_store.clear_session(session_key=session_key)
     agent_trace_store.clear_session(session_key=session_key)
-    openai_credential_store.clear_session(session_key=session_key)
+    clear_ai_credential_sessions(session_key=session_key)
     token_service.revoke_token(token)
     shell_session_service.reset()
     # Lock and login end a retry through the runtime generation; a passwordless
@@ -668,7 +671,7 @@ def create_passwordless_session(
     web_evidence_store.reset()
     summary_batch_store.reset()
     agent_trace_store.reset()
-    openai_credential_store.reset()
+    reset_ai_credentials()
     view_cache.clear()
     reset_all_undo_state()
     reset_sync_state()
@@ -1088,7 +1091,7 @@ def create_password(
     web_evidence_store.reset()
     summary_batch_store.reset()
     agent_trace_store.reset()
-    openai_credential_store.reset()
+    reset_ai_credentials()
     clear_all_locks()
     return {"message": message}
 
@@ -1115,7 +1118,7 @@ def change_password(
     web_evidence_store.reset()
     summary_batch_store.reset()
     agent_trace_store.reset()
-    openai_credential_store.reset()
+    reset_ai_credentials()
     clear_all_locks()
     return {"message": message}
 
@@ -1138,7 +1141,7 @@ def remove_password(
     web_evidence_store.reset()
     summary_batch_store.reset()
     agent_trace_store.reset()
-    openai_credential_store.reset()
+    reset_ai_credentials()
     clear_all_locks()
     clear_encryption_key()
     deactivate_authenticated_logging()