superlocalmemory 4.2.11__tar.gz → 4.2.12__tar.gz

{superlocalmemory-4.2.11/src/superlocalmemory.egg-info → superlocalmemory-4.2.12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: superlocalmemory
-Version: 4.2.11
+Version: 4.2.12
 Summary: The governed memory brain for AI agents: one local memory for Claude Code, Codex, Cursor, Copilot, ChatGPT and any MCP client
 Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
 License-Expression: AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/agents/slm-governance-advisor.md

@@ -81,4 +81,4 @@ slm-scope · slm-governance · slm-profile · slm-remember · slm-recall · slm-
 # What NOT to do
 Never session_init twice; never forget without dry-run preview; never run `slm gdpr erase --yes` unasked; never store secrets; never bypass role checks; never claim an erasure succeeded without verifying via recall.
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/agents/slm-loop-runner.md

@@ -79,4 +79,4 @@ assessment. The gate is the authority.
 
 ---
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/agents/slm-memory-advisor.md

@@ -51,4 +51,4 @@ slm-recall · slm-remember · slm-session · slm-scope · slm-profile · slm-gov
 # What NOT to do
 Never session_init twice; never forget dry_run=False without reporting preview; never switch the active profile unasked; never dump a whole file into remember; never invent a memory; never claim "saved" without success:true / clean CLI exit; never bypass scope or governance restrictions.
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/agents/slm-optimize-advisor.md

@@ -41,4 +41,4 @@ slm-compress · slm-cache · slm-status · slm-profile
 # What NOT to do
 Never compress code-for-edit/JSON-to-parse/<500 chars; never store secrets/ccr_ids; never let optimize failure block/alter the task; never claim a specific savings %; never hand a ccr_id to another agent.
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/rules/AGENTS.md

@@ -187,4 +187,4 @@ Using this memory from a web assistant or another computer: see the slm-web-acce
 - **slm-governance-advisor** — scope/role compliance, retention and lifecycle, GDPR
 - **slm-loop-runner** — bounded, gate-verified loops (the `/slm-loop` command delegates to it)
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/rules/CLAUDE.md.fragment

@@ -1,4 +1,4 @@
-<!-- BEGIN SuperLocalMemory v4.2.11 -->
+<!-- BEGIN SuperLocalMemory v4.2.12 -->
 
 ## SuperLocalMemory (SLM) — Agent Rules
 
@@ -51,6 +51,6 @@ slm-recall · slm-remember · slm-session · slm-status · slm-cache · slm-comp
 ### Subagents
 slm-memory-advisor (memory decisions, session hygiene, scope/profile guidance) · slm-optimize-advisor (context compression + KV cache) · slm-governance-advisor (scope/roles/compliance/GDPR) · slm-loop-runner (bounded, gate-verified loops)
 
-<!-- END SuperLocalMemory v4.2.11 -->
+<!-- END SuperLocalMemory v4.2.12 -->
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-bot-memory/SKILL.md

@@ -178,4 +178,4 @@ store every local bot on that computer can reach.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-cache/SKILL.md

@@ -144,4 +144,4 @@ These subcommands control daemon-level cache settings and bulk removal. They do
 
 ---
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-compress/SKILL.md

@@ -147,4 +147,4 @@ Content over 1 MB (1 000 000 bytes UTF-8) is processed but `reversible` is force
 
 ---
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-getting-started-bot/SKILL.md

@@ -194,4 +194,4 @@ quality on a shared box.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-governance/SKILL.md

@@ -276,4 +276,4 @@ Before running any destructive or state-changing operation (`forget`,
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-graph/SKILL.md

@@ -318,4 +318,4 @@ All are host-only: a remote caller cannot use them.
 
 ---
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-loop/SKILL.md

@@ -171,4 +171,4 @@ Guides: `docs/skill-evolution.md`, `docs/bounded-loops-bridge.md`.
 
 ---
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-media/SKILL.md

@@ -337,4 +337,4 @@ minutes (previous vectors are kept) and let them decide.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-mesh/SKILL.md

@@ -386,4 +386,4 @@ mesh availability.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-profile/SKILL.md

@@ -164,4 +164,4 @@ anything but `core` or `mesh`.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-recall/SKILL.md

@@ -515,4 +515,4 @@ See `slm-profile`.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-remember/SKILL.md

@@ -372,4 +372,4 @@ sessions on this machine, so use it only when the user asks to move. See
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-scope/SKILL.md

@@ -180,4 +180,4 @@ to review the impact. See `slm-remember` for the full deletion discipline.
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-session/SKILL.md

@@ -288,4 +288,4 @@ answered. To also surface shared or global facts, call `recall` with
 
 ---
 
-*SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later*

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/plugin-src/skills/slm-status/SKILL.md

@@ -228,4 +228,4 @@ the memory-engine upgrade is pending.
 
 ---
 
-SuperLocalMemory v4.2.11 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v4.2.12 · Qualixar · AGPL-3.0-or-later

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "superlocalmemory"
-version = "4.2.11"
+version = "4.2.12"
 description = "The governed memory brain for AI agents: one local memory for Claude Code, Codex, Cursor, Copilot, ChatGPT and any MCP client"
 readme = "README.md"
 license = "AGPL-3.0-or-later"

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/__init__.py

@@ -32,7 +32,7 @@ if "OMP_NUM_THREADS" not in os.environ:
     os.environ["OMP_NUM_THREADS"] = "2"
 # ---------------------------------------------------------------------------
 
-__version__ = "4.2.11"
+__version__ = "4.2.12"
 
 _REQUIRED_VERSIONS = {
     "sentence_transformers": "5.6.1",

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/compliance/eu_ai_act.py

@@ -17,6 +17,10 @@ import logging
 from dataclasses import dataclass
 from datetime import UTC, datetime
 
+from superlocalmemory.core.data_locality import (
+    effective_data_locality,
+    remote_parts_sentence,
+)
 from superlocalmemory.core.modes import get_capabilities
 from superlocalmemory.storage.models import Mode
 
@@ -42,24 +46,37 @@ class ComplianceReport:
 class EUAIActChecker:
     """Report mode-level technical posture without giving legal certification."""
 
-    def check_compliance(self, mode: Mode) -> ComplianceReport:
-        """Generate compliance report for a mode."""
+    def check_compliance(
+        self, mode: Mode, *, embedding: object = None, llm: object = None,
+    ) -> ComplianceReport:
+        """Generate compliance report for a mode.
+
+        With the configured ``embedding`` and ``llm`` blocks (#303) the data
+        locality is the effective one, so a remote embedding endpoint under
+        Mode A or B is reported as data leaving the device. Without them it is
+        the mode's default.
+        """
         caps = get_capabilities(mode)
+        locality = effective_data_locality(mode, embedding, llm)
         findings: list[str] = []
 
         # Data locality
-        data_local = caps.data_stays_local
+        data_local = locality.stays_local
         if not data_local:
-            findings.append(
-                "Data leaves device for cloud LLM processing. "
-                "Requires Data Processing Agreement (DPA) with provider."
-            )
+            if caps.data_stays_local:
+                findings.append(remote_parts_sentence(locality))
+            else:
+                findings.append(
+                    "Data leaves device for cloud LLM processing. "
+                    "Requires Data Processing Agreement (DPA) with provider."
+                )
 
         # Generative AI usage
         uses_gen_ai = caps.llm_fact_extraction or caps.llm_answer_generation
-        local_gen_ai = uses_gen_ai and data_local
+        cloud_gen_ai = uses_gen_ai and (not caps.data_stays_local or locality.llm_remote)
+        local_gen_ai = uses_gen_ai and not cloud_gen_ai
 
-        if uses_gen_ai and not data_local:
+        if cloud_gen_ai:
             findings.append(
                 "Uses cloud generative AI. EU AI Act Art. 52 requires "
                 "transparency: users must be informed AI generates content."

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/core/config.py

@@ -2031,9 +2031,13 @@ class SLMConfig:
             # endpoint/model/dimension. Otherwise use local defaults.
             _a_provider = embedding_provider or "sentence-transformers"
             if _a_provider == "openai" and embedding_endpoint:
+                # #302: a provider gets its own model id or none at all, never
+                # another provider's. The bundle's nomic id and 768 are not
+                # something an endpoint is known to serve, so blank stays blank
+                # and the save doors ask for both (embedding_endpoint_check).
                 _a_emb = EmbeddingConfig(
-                    model_name=embedding_model_name or "nomic-ai/nomic-embed-text-v1.5",
-                    dimension=embedding_dimension or 768,
+                    model_name=embedding_model_name,
+                    dimension=embedding_dimension,
                     provider="openai",
                     api_endpoint=embedding_endpoint,
                     api_key=embedding_key,
@@ -2083,9 +2087,10 @@ class SLMConfig:
             # (e.g. local vLLM, LiteLLM, Ollama /v1), honour it.
             _b_provider = embedding_provider or "ollama"
             if _b_provider == "openai" and embedding_endpoint:
+                # #302: same rule as Mode A above.
                 _b_emb = EmbeddingConfig(
-                    model_name=embedding_model_name or "nomic-ai/nomic-embed-text-v1.5",
-                    dimension=embedding_dimension or 768,
+                    model_name=embedding_model_name,
+                    dimension=embedding_dimension,
                     provider="openai",
                     api_endpoint=embedding_endpoint,
                     api_key=embedding_key,

superlocalmemory-4.2.12/src/superlocalmemory/core/data_locality.py

@@ -0,0 +1,147 @@
+# Copyright (c) 2026 Varun Pratap Bhardwaj / Qualixar
+# Licensed under AGPL-3.0-or-later - see LICENSE file
+# Part of SuperLocalMemory V3 | https://qualixar.com | https://varunpratap.com
+
+"""Where memory text goes, worked out from the configuration (#303).
+
+Modes A and B accept an OpenAI-compatible embedding endpoint, and Mode B a
+hosted language model, so "data stays local" cannot be a fact about the mode
+name. ``ModeCapabilities.data_stays_local`` stays as the mode's default; this
+module answers the question for the configuration actually in force, and every
+surface that reports locality asks it.
+
+An endpoint is local when its literal host is loopback (``localhost``,
+``*.localhost``, 127.0.0.0/8, ::1) or a numeric private-LAN address
+(10/8, 172.16/12, 192.168/16, fc00::/7, link-local). Anything else, including
+a public hostname that cannot be resolved, is remote. There is no DNS lookup:
+a name is not proof of where it points, the same posture the remote reranker
+and the LLM endpoint checks already take, so this reuses their host helpers.
+
+Pure functions, no I/O.
+
+Part of Qualixar | Author: Varun Pratap Bhardwaj
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Any
+from urllib.parse import urlparse
+
+from superlocalmemory.core.modes import get_capabilities
+from superlocalmemory.retrieval.remote_reranker_config import (
+    _is_loopback_host,
+    _is_private_lan_host,
+)
+from superlocalmemory.storage.models import Mode
+
+# What Mode B's template writes into api_base for any LLM provider when none was
+# given. It is Ollama's address, so it says nothing about where a hosted
+# provider's requests go.
+_OLLAMA_TEMPLATE_BASE = "http://localhost:11434"
+
+EMBEDDING_PART = "embedding endpoint"
+LLM_PART = "language model"
+
+
+def _host_of(url: str) -> str:
+    text = (url or "").strip()
+    if not text:
+        return ""
+    if "://" not in text:
+        text = f"//{text}"  # "localhost:8045/v1" parses as a scheme otherwise
+    try:
+        return urlparse(text).hostname or ""
+    except ValueError:
+        return ""
+
+
+def is_local_endpoint(url: str) -> bool:
+    """True when ``url``'s literal host is loopback or a private-LAN address."""
+    host = _host_of(url)
+    return bool(host) and (_is_loopback_host(host) or _is_private_lan_host(host))
+
+
+def _embedding_is_remote(embedding: Any) -> bool:
+    provider = getattr(embedding, "provider", "") or ""
+    if provider == "openai":
+        endpoint = getattr(embedding, "api_endpoint", "") or ""
+        return bool(endpoint) and not is_local_endpoint(endpoint)
+    if provider == "ollama":
+        base = getattr(embedding, "ollama_base_url", "") or ""
+        return bool(base) and not is_local_endpoint(base)
+    if provider == "cloud":
+        return True
+    return False
+
+
+def _llm_is_remote(llm: Any) -> bool:
+    provider = (getattr(llm, "provider", "") or "").strip().lower()
+    if provider in ("", "none"):
+        return False
+    base = (getattr(llm, "api_base", "") or "").strip()
+    if provider == "ollama":
+        return bool(base) and not is_local_endpoint(base)
+    if base.rstrip("/") == _OLLAMA_TEMPLATE_BASE:
+        return True
+    return not (base and is_local_endpoint(base))
+
+
+@dataclass(frozen=True)
+class DataLocality:
+    """Whether memory text stays on this computer, and what carries it away."""
+
+    stays_local: bool
+    remote_parts: tuple[str, ...] = ()
+
+    @property
+    def label(self) -> str:
+        return "local-only" if self.stays_local else "provider-assisted"
+
+    @property
+    def embedding_remote(self) -> bool:
+        return EMBEDDING_PART in self.remote_parts
+
+    @property
+    def llm_remote(self) -> bool:
+        return LLM_PART in self.remote_parts
+
+
+def effective_data_locality(
+    mode: Mode, embedding: Any = None, llm: Any = None,
+) -> DataLocality:
+    """The mode's default, made true to the configuration.
+
+    A mode whose default is not local stays not local. A mode whose default is
+    local stays local only while the embedding endpoint is local and, for a
+    mode that calls a language model, that model is too. With no configuration
+    given the answer is the mode's default.
+    """
+    caps = get_capabilities(mode)
+    parts: list[str] = []
+    if embedding is not None and _embedding_is_remote(embedding):
+        parts.append(EMBEDDING_PART)
+    if llm is not None and caps.llm_fact_extraction and _llm_is_remote(llm):
+        parts.append(LLM_PART)
+    return DataLocality(
+        stays_local=caps.data_stays_local and not parts,
+        remote_parts=tuple(parts),
+    )
+
+
+def remote_parts_sentence(locality: DataLocality) -> str:
+    """"" while data stays local; else one plain sentence saying what carries it away."""
+    if locality.stays_local or not locality.remote_parts:
+        return ""
+    what = " and ".join(locality.remote_parts)
+    return f"The {what} you configured is not on this computer, so memory text leaves this computer."
+
+
+__all__ = [
+    "DataLocality",
+    "EMBEDDING_PART",
+    "LLM_PART",
+    "effective_data_locality",
+    "is_local_endpoint",
+    "remote_parts_sentence",
+]

superlocalmemory-4.2.12/src/superlocalmemory/core/embedding_endpoint_check.py

@@ -0,0 +1,53 @@
+# Copyright (c) 2026 Varun Pratap Bhardwaj / Qualixar
+# Licensed under AGPL-3.0-or-later - see LICENSE file
+# Part of SuperLocalMemory V3 | https://qualixar.com | https://varunpratap.com
+
+"""What an OpenAI-compatible embedding endpoint needs before it is saved (#302).
+
+The model name is sent verbatim to the endpoint and the width has to match the
+vectors it returns. Neither can be guessed: the local bundle's model id is not
+one the endpoint serves, and a guessed width either fails at the first request
+or rebuilds a store that held another one. So every door that saves an
+embedding config asks for both and says which is missing.
+
+Pure function, no I/O.
+
+Part of Qualixar | Author: Varun Pratap Bhardwaj
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+#: The ``code`` a refused save carries, for a client that wants to branch on it.
+INCOMPLETE_ENDPOINT_CODE = "embedding_endpoint_incomplete"
+
+
+def incomplete_endpoint_error(embedding: Any) -> str | None:
+    """An actionable message when an endpoint config lacks a model or a width, else ``None``.
+
+    Only an OpenAI-compatible config (provider ``openai`` with an endpoint) is
+    checked; every other provider has its own defaults.
+    """
+    if not (getattr(embedding, "provider", "") == "openai" and getattr(embedding, "api_endpoint", "")):
+        return None
+    missing = []
+    if not (getattr(embedding, "model_name", "") or "").strip():
+        missing.append("model name")
+    try:
+        width = int(getattr(embedding, "dimension", 0) or 0)
+    except (TypeError, ValueError):
+        width = 0
+    if width <= 0:
+        missing.append("dimension")
+    if not missing:
+        return None
+    return (
+        f"A custom embedding endpoint needs its {' and '.join(missing)}. Fill in "
+        f"the model the endpoint serves (for example Qwen3-Embedding) and the "
+        f"vector size it returns (use Test Embedding to read it). SuperLocalMemory "
+        f"will not guess either one."
+    )
+
+
+__all__ = ["INCOMPLETE_ENDPOINT_CODE", "incomplete_endpoint_error"]

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/core/engine.py

@@ -397,6 +397,7 @@ class MemoryEngine:
         _llm_up = getattr(self, "_llm", None) is not None
         for _warning in validate_mode_config(
             self._config.mode, has_llm=_llm_up, has_cloud_llm=_llm_up,
+            embedding=self._config.embedding, llm=self._config.llm,
         ):
             logger.warning("Mode config: %s", _warning)
 

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/core/engine_ingestion.py

@@ -13,7 +13,7 @@ from __future__ import annotations
 import hashlib
 import logging
 import uuid
-from typing import TYPE_CHECKING, Protocol
+from typing import TYPE_CHECKING, Any, Protocol
 
 from superlocalmemory.core.derivation_lineage import checkpoint_with_lineage
 from superlocalmemory.core.ingestion_command import (
@@ -37,6 +37,35 @@ _PREBUILT_FACT_KEY = "_slm_prebuilt_fact_v1"
 _DERIVATION_VERSION = "v3.7-ingestion-1"
 
 
+def _nothing_left_live(db: Any, operation: IngestionOperation) -> bool:
+    """True when every queryable fact of the save is archived, erased or being erased.
+
+    That is a save retired before it was enriched (a folder note revised or
+    deleted while it was still queryable, #297), not a broken one. A fact that
+    is still live in this profile, or that sits in another profile, keeps the
+    "mismatch or missing fact" error: those are real faults.
+    """
+    ids = list(operation.queryable_fact_ids)
+    if not ids:
+        return False
+    marks = ",".join("?" for _ in ids)
+    rows = db.execute(
+        f"SELECT fact_id, profile_id, COALESCE(archive_status, 'live') AS status "
+        f"FROM atomic_facts WHERE fact_id IN ({marks})", tuple(ids))
+    try:
+        erasing = {str(dict(r)["fact_id"]) for r in db.execute(
+            f"SELECT fact_id FROM projection_tombstones WHERE profile_id=? "
+            f"AND fact_id IN ({marks})", (operation.profile_id, *ids))}
+    except Exception:  # noqa: BLE001 -- a schema without tombstones erases nothing
+        erasing = set()
+    for row in map(dict, rows):
+        if row["profile_id"] != operation.profile_id:
+            return False
+        if row["status"] == "live" and str(row["fact_id"]) not in erasing:
+            return False
+    return True
+
+
 def _require_known_profile(engine: "MemoryEngine", profile_id: str) -> None:
     """Fail closed on a write routed to a profile that does not exist.
 
@@ -828,6 +857,11 @@ def build_engine_ingestion_command(
             operation.profile_id,
         )
         if len(facts) != len(operation.queryable_fact_ids):
+            if _nothing_left_live(engine._db, operation):
+                from superlocalmemory.core.materialization_control import (
+                    MaterializationSuperseded,
+                )
+                raise MaterializationSuperseded(operation.operation_id)
             raise ValueError("queryable fact profile mismatch or missing fact")
         memory_ids = {fact.memory_id for fact in facts}
         if len(memory_ids) != 1:

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/core/ingestion_command.py

@@ -23,7 +23,10 @@ from enum import Enum
 from typing import Any, Callable
 
 from superlocalmemory.core.ingestion_lease_owner import owner_token, release_dead_owner_leases
-from superlocalmemory.core.materialization_control import MaterializationDeferred
+from superlocalmemory.core.materialization_control import (
+    MaterializationDeferred,
+    MaterializationSuperseded,
+)
 from superlocalmemory.storage.database import DatabaseManager
 from superlocalmemory.storage.idempotency_identity import same_principal
 
@@ -1007,6 +1010,19 @@ class IngestionCommand:
                 operation_id,
                 owner=self._owner,
             )
+        except MaterializationSuperseded:
+            # #297: the save was retired before it was enriched. Done, with
+            # nothing enriched back; its queryable ids stay on the row for the
+            # erasure scrubs, and "superseded" says why it has no final facts.
+            logger.info("ingestion %s was retired before enrichment; settled", operation_id)
+            return self.repository.finish_enriching(
+                operation_id,
+                owner=self._owner,
+                target=IngestionState.COMPLETE,
+                final_fact_ids=(),
+                derivation_version=self._derivation_version,
+                derivation_state={"superseded": True},
+            )
         except Exception as exc:
             return self.repository.finish_enriching(
                 operation_id,

{superlocalmemory-4.2.11 → superlocalmemory-4.2.12}/src/superlocalmemory/core/materialization_control.py

@@ -18,3 +18,12 @@ class MaterializationDeferred(RuntimeError):
     The queryable projection is already durable.  This signal must be handled
     by :class:`IngestionCommand` as a requeue, never as a failed attempt.
     """
+
+
+class MaterializationSuperseded(RuntimeError):
+    """Nothing of a queryable save is live any more: it was retired before enrichment.
+
+    A folder note revised or deleted while its save was still queryable has its
+    facts archived or erased (#297). Enriching it would put retired content back,
+    so :class:`IngestionCommand` settles the operation instead of failing it.
+    """