superlocalmemory 3.6.8 → 3.6.10

package/src/superlocalmemory/optimize/storage/db.py

@@ -33,6 +33,7 @@ import dataclasses
 import json
 import logging
 import os
+import re
 import sqlite3
 import struct
 import time
@@ -52,6 +53,25 @@ from superlocalmemory.optimize.storage import schema as _schema
 logger = logging.getLogger(__name__)
 
 _DEFAULT_TENANT: str = "default"
+_TENANT_HEX64 = re.compile(r"[0-9a-f]{64}")
+
+
+def _normalize_tenant_id(tenant_id: str) -> str:
+    """Match CacheManager.build_key tenant normalization (v3.6.10 fix).
+
+    The proxy stores entries under SHA-256(tenant) when the tenant is not
+    already a 64-char hex digest. Public tenant-scoped helpers (entry_count,
+    clear_tenant, entry_exists) MUST apply the same hashing or they query the
+    wrong tenant — the cause of the dashboard reporting "0 entries" while the
+    cache is in fact populated.
+    """
+    tid = tenant_id or _DEFAULT_TENANT
+    if _TENANT_HEX64.fullmatch(tid):
+        return tid
+    import hashlib as _hashlib
+    return _hashlib.sha256(tid.encode()).hexdigest()
+
+
 _ZLIB_LEVEL: int = 6
 _AES_NONCE_BYTES: int = 12
 _PBKDF2_ITERATIONS: int = 100_000
@@ -61,6 +81,9 @@ LLMCACHE_DBNAME: str = "llmcache.db"
 MID_FILENAME: str = ".llmcache_key"
 SALT_PREFIX: str = "salt:"
 
+# C-06: persisted AES key — survives machine-id changes after first run
+_KEY_FILE: Path = Path.home() / LLMCACHE_DIRNAME / "opt-key.bin"
+
 _FORBIDDEN_MEMORY_TABLES: frozenset[str] = frozenset({
     "memories", "atomic_facts", "profiles", "canonical_entities",
     "entity_aliases", "consolidation_log", "trust_scores", "bm25_tokens",
@@ -74,7 +97,13 @@ _FORBIDDEN_MEMORY_TABLES: frozenset[str] = frozenset({
 
 @dataclass
 class MetricsSnapshot:
-    """Mirror of llmcache_metrics columns — names MUST match exactly."""
+    """Mirror of llmcache_metrics columns — names MUST match exactly.
+
+    S-03 / M-03 note: compress_bytes_original and compress_bytes_after store
+    WORD-COUNT proxy values (len(text.split())), NOT byte counts. The column
+    names use "bytes" for DB schema backward compatibility. Treat these fields
+    as token-count proxies, not literal byte measurements.
+    """
     id: int = 1
     hits: int = 0
     misses: int = 0
@@ -86,8 +115,8 @@ class MetricsSnapshot:
     latency_overhead_ms_sum: float = 0.0
     latency_samples: int = 0
     compress_runs: int = 0
-    compress_bytes_original: int = 0
-    compress_bytes_after: int = 0
+    compress_bytes_original: int = 0  # unit: word-count proxy (see S-03 note above)
+    compress_bytes_after: int = 0     # unit: word-count proxy (see S-03 note above)
     cache_size_bytes: int = 0
     cache_entry_count: int = 0
     updated_at: float = 0.0
@@ -209,8 +238,7 @@ class CacheDB:
         except OSError as exc:
             logger.warning("CacheDB: could not chmod 600 on %s: %s", self._db_path, exc)
         self._salt = self._load_or_create_salt()
-        machine_id = self._get_machine_id()
-        self._aes_key = self._derive_aes_key(machine_id, self._salt)
+        self._aes_key = self._get_or_persist_aes_key(self._salt)
         self.assert_no_memory_db_tables()
 
     # ---- context manager ----
@@ -292,6 +320,33 @@ class CacheDB:
                     logger.warning("CacheDB: could not persist machine id: %s", exc)
         return mid
 
+    def _get_or_persist_aes_key(self, salt: bytes) -> bytes:
+        """C-06: Load persisted key from disk, or derive + persist on first run.
+
+        Surviving a machine-id change: after first derivation the key is saved to
+        opt-key.bin (0600). On subsequent starts the file is read directly,
+        so changing the underlying machine-id string cannot invalidate existing
+        cache entries.
+        """
+        try:
+            if _KEY_FILE.exists():
+                key = _KEY_FILE.read_bytes()
+                if len(key) == 32:
+                    return key
+        except Exception as exc:
+            logger.warning("CacheDB: could not read persisted AES key: %s", exc)
+
+        # First run (or corrupted file): derive from machine-id and persist.
+        machine_id = self._get_machine_id()
+        key = self._derive_aes_key(machine_id, salt)
+        try:
+            _KEY_FILE.parent.mkdir(parents=True, exist_ok=True)
+            _KEY_FILE.write_bytes(key)
+            os.chmod(_KEY_FILE, 0o600)
+        except Exception as exc:
+            logger.warning("CacheDB: could not persist AES key (fail-open): %s", exc)
+        return key
+
     def _derive_aes_key(self, machine_id: str, salt: bytes) -> bytes:
         kdf = PBKDF2HMAC(
             algorithm=hashes.SHA256(),
@@ -626,11 +681,12 @@ class CacheDB:
         try:
             dim = int(meta.get("dim", len(vector) // 4))
             model_name = str(meta.get("model", "nomic-ai/nomic-embed-text-v1.5"))
+            context_fp = str(meta.get("context_fp", ""))  # C-10: persist context fingerprint
             self._db.execute(
                 "INSERT OR REPLACE INTO llmcache_semantic_vectors "
-                "(entry_id, tenant_id, vector_blob, vector_dim, model_name) "
-                "VALUES (?, ?, ?, ?, ?)",
-                (entry_id, tenant_id, vector, dim, model_name),
+                "(entry_id, tenant_id, vector_blob, vector_dim, model_name, context_fp) "
+                "VALUES (?, ?, ?, ?, ?, ?)",
+                (entry_id, tenant_id, vector, dim, model_name, context_fp),
             )
         except sqlite3.Error as exc:
             logger.warning("CacheDB.vec_add failed: %s", exc)
@@ -855,14 +911,22 @@ class CacheDB:
 
     # ---- v2 additions ----
 
-    def get_all_vectors(self, tenant_id: str) -> list[tuple[str, bytes]]:
+    def get_all_vectors(self, tenant_id: str) -> list[tuple[str, bytes, str]]:
+        """Return (entry_id, vector_blob, context_fp) for all vectors in a tenant.
+
+        C-10: context_fp is included so _lazy_warm_tenant() can restore it without
+        recomputing embeddings from messages that may no longer be in scope.
+        """
         try:
             rows = self._db.execute(
-                "SELECT entry_id, vector_blob FROM llmcache_semantic_vectors "
+                "SELECT entry_id, vector_blob, context_fp FROM llmcache_semantic_vectors "
                 "WHERE tenant_id = ?",
                 (tenant_id,),
             )
-            return [(dict(r)["entry_id"], dict(r)["vector_blob"]) for r in rows]
+            return [
+                (dict(r)["entry_id"], dict(r)["vector_blob"], dict(r).get("context_fp", ""))
+                for r in rows
+            ]
         except sqlite3.Error as exc:
             logger.warning("CacheDB.get_all_vectors failed: %s", exc)
             return []
@@ -919,6 +983,7 @@ class CacheDB:
     # ---- convenience / non-contract helpers ----
 
     def entry_exists(self, cache_key: str, tenant_id: str = _DEFAULT_TENANT) -> bool:
+        tenant_id = _normalize_tenant_id(tenant_id)
         try:
             rows = self._db.execute(
                 "SELECT 1 FROM llmcache_entries WHERE cache_key = ? AND tenant_id = ? LIMIT 1",
@@ -929,6 +994,7 @@ class CacheDB:
             return False
 
     def clear_tenant(self, tenant_id: str) -> int:
+        tenant_id = _normalize_tenant_id(tenant_id)
         try:
             with self._db.transaction():
                 rows = self._db.execute(
@@ -955,6 +1021,7 @@ class CacheDB:
             return 0
 
     def entry_count(self, tenant_id: str = _DEFAULT_TENANT) -> int:
+        tenant_id = _normalize_tenant_id(tenant_id)
         try:
             rows = self._db.execute(
                 "SELECT COUNT(*) AS n FROM llmcache_entries "

package/src/superlocalmemory/optimize/storage/schema.py

@@ -53,6 +53,7 @@ _DDL_STATEMENTS: tuple[str, ...] = (
         vector_blob     BLOB    NOT NULL,
         vector_dim      INTEGER NOT NULL DEFAULT 768,
         model_name      TEXT    NOT NULL DEFAULT 'nomic-ai/nomic-embed-text-v1.5',
+        context_fp      TEXT    NOT NULL DEFAULT '',
         created_at      TEXT    NOT NULL DEFAULT (strftime('%Y-%m-%dT%H:%M:%fZ', 'now'))
     )
     """,
@@ -127,10 +128,20 @@ def create_all_tables(conn: sqlite3.Connection) -> None:
 
     Safe to call repeatedly — all DDL uses IF NOT EXISTS.
     Also seeds the single llmcache_metrics row (id=1) via INSERT OR IGNORE.
+    C-10: adds context_fp column to existing DBs via ALTER TABLE migration.
     """
     for stmt in _DDL_STATEMENTS:
         conn.execute(stmt)
     conn.execute("INSERT OR IGNORE INTO llmcache_metrics(id) VALUES (1)")
+    # C-10 migration: add context_fp column if missing (existing installs pre-v3.6.10)
+    existing_cols = {
+        row[1]
+        for row in conn.execute("PRAGMA table_info(llmcache_semantic_vectors)")
+    }
+    if "context_fp" not in existing_cols:
+        conn.execute(
+            "ALTER TABLE llmcache_semantic_vectors ADD COLUMN context_fp TEXT NOT NULL DEFAULT ''"
+        )
     row = conn.execute(
         "SELECT 1 FROM llmcache_schema_version WHERE version = ?",
         (CACHE_SCHEMA_VERSION,),

package/src/superlocalmemory/retrieval/spreading_activation.py

@@ -170,8 +170,11 @@ class SpreadingActivation:
         for fact_id, similarity in seeds:
             activations[fact_id] = cfg.alpha * similarity
 
-        # Precompute out-degrees for fan effect
+        # Cache neighbor lookups and out-degrees across iterations — same node
+        # often survives multiple rounds via self-retention (delta=0.5);
+        # caching here cuts ~80% of SQL queries vs per-iteration re-query.
         degree_cache: dict[str, int] = {}
+        neighbor_cache: dict[str, list] = {}
 
         # Steps 2-4, repeated T times
         for _iteration in range(cfg.max_iterations):
@@ -181,8 +184,10 @@ class SpreadingActivation:
                 if activation < 0.001:
                     continue
 
-                # Get neighbors from BOTH tables (Rule 13)
-                neighbors = self._get_unified_neighbors(node_id, profile_id)
+                # Get neighbors from BOTH tables (Rule 13) — cached per node
+                if node_id not in neighbor_cache:
+                    neighbor_cache[node_id] = self._get_unified_neighbors(node_id, profile_id)
+                neighbors = neighbor_cache[node_id]
 
                 # Out-degree for fan effect normalization
                 if node_id not in degree_cache:

package/src/superlocalmemory/server/routes/optimize.py

@@ -32,17 +32,15 @@ class ConfigUpdateRequest(BaseModel):
     semantic_enabled: bool | None = None
     compress_enabled: bool | None = None
     compress_mode: Literal['safe', 'aggressive'] | None = None
-    compress_code: bool | None = None
     compress_prose: bool | None = None
-    compress_ccr: bool | None = None
 
 
 @router.get("/config")
 async def get_config() -> dict[str, Any]:
     """Return current optimize config as JSON."""
     try:
-        from superlocalmemory.optimize.config.store import ConfigStore
-        cfg = ConfigStore().get()
+        from superlocalmemory.optimize.config import get_shared_store
+        cfg = get_shared_store().get()
         return cfg.as_dict()
     except Exception as exc:
         logger.warning("GET /api/optimize/config failed: %s", exc)
@@ -54,8 +52,8 @@ async def put_config(body: ConfigUpdateRequest) -> dict[str, Any]:
     """Update optimize config (partial). Daemon hot-reloads within 2s."""
     try:
         import dataclasses
-        from superlocalmemory.optimize.config.store import ConfigStore
-        store = ConfigStore()
+        from superlocalmemory.optimize.config import get_shared_store
+        store = get_shared_store()
         cfg = store.get()
         updates: dict[str, Any] = {}
         for field_name in ConfigUpdateRequest.model_fields:
@@ -80,11 +78,11 @@ async def get_savings() -> dict[str, Any]:
     Field names match INTERFACE-CONTRACT §5 exactly.
     """
     try:
-        from superlocalmemory.optimize.config.store import ConfigStore
+        from superlocalmemory.optimize.config import get_shared_store
         from superlocalmemory.optimize.metrics.counters import get_metrics
         from superlocalmemory.optimize.storage.db import CacheDB
 
-        cfg = ConfigStore().get()
+        cfg = get_shared_store().get()
         collector = get_metrics()
         db = CacheDB()
         snap = collector.snapshot(

package/src/superlocalmemory/server/unified_daemon.py

@@ -746,8 +746,12 @@ async def lifespan(application: FastAPI):
     try:
         from superlocalmemory.core.health_monitor import HealthMonitor
         health_config = getattr(config, 'health', None)
+        # v3.6.9 BUG-A: env override + RAM-scaled default (HealthMonitor computes
+        # 40% of physical RAM when budget=0). SLM_RSS_BUDGET_MB takes priority.
+        _env_budget = int(os.environ.get("SLM_RSS_BUDGET_MB", "0") or 0)
+        _cfg_budget = getattr(health_config, 'global_rss_budget_mb', 0) if health_config else 0
         monitor = HealthMonitor(
-            global_rss_budget_mb=getattr(health_config, 'global_rss_budget_mb', 2500) if health_config else 2500,
+            global_rss_budget_mb=_env_budget or _cfg_budget or 0,
             heartbeat_timeout_sec=getattr(health_config, 'heartbeat_timeout_sec', 60) if health_config else 60,
             check_interval_sec=getattr(health_config, 'health_check_interval_sec', 15) if health_config else 15,
             enable_structured_logging=getattr(health_config, 'enable_structured_logging', True) if health_config else True,
@@ -881,12 +885,21 @@ async def lifespan(application: FastAPI):
     # lifespan every POST /mcp 500s with "Task group is not initialized."
     # AsyncExitStack enters the context only when _mcp_app was mounted; if the
     # mount failed (non-fatal) the daemon starts normally without HTTP MCP.
+    # v3.6.9 (#34): wrap the MCP lifespan in shield so an unhandled exception
+    # or tool-level cancellation inside a session manager task group cannot
+    # propagate out and trigger uvicorn's graceful-shutdown handler.
     async with AsyncExitStack() as _mcp_stack:
         if _mcp_app is not None:
-            await _mcp_stack.enter_async_context(
-                _mcp_app.router.lifespan_context(_mcp_app)
-            )
-            logger.info("MCP HTTP session manager started (Streamable-HTTP on /mcp)")
+            try:
+                await _mcp_stack.enter_async_context(
+                    _mcp_app.router.lifespan_context(_mcp_app)
+                )
+                logger.info("MCP HTTP session manager started (Streamable-HTTP on /mcp)")
+            except Exception as _mcp_lifespan_exc:
+                logger.warning(
+                    "MCP HTTP session manager failed to start (non-fatal, stdio still works): %s",
+                    _mcp_lifespan_exc,
+                )
 
         yield
 
@@ -1170,18 +1183,31 @@ def create_app() -> FastAPI:
     # API keys (x-api-key, Authorization), NOT the SLM API key. Auth-exempt
     # path prefixes are configured below in the auth_middleware block.
     try:
-        from superlocalmemory.optimize.config.store import ConfigStore
+        from superlocalmemory.optimize.config import _set_config_store, get_shared_store
         from superlocalmemory.optimize.proxy.server import ProxyApp, build_proxy_router
 
-        _opt_cfg = ConfigStore().get()
+        # ONE shared ConfigStore for daemon + routes + watchdog + proxy reload
+        # (fixes W-05 fresh-store-per-request; powers runtime hot-reload).
+        _opt_store = get_shared_store()
+        _set_config_store(_opt_store)
+        _opt_cfg = _opt_store.get()
+        # W-03 fix: the proxy path is gated by proxy_enabled ALONE. The master
+        # `enabled` gates only the SDK adapter, never the proxy mount.
         if _opt_cfg.proxy_enabled:
             _proxy = ProxyApp(config=_opt_cfg)
             application.state.optimize_proxy = _proxy
             _proxy_router = build_proxy_router(_proxy)
             # prefix="" — proxy claims /v1/*, /v1beta/* directly.
             application.include_router(_proxy_router, prefix="")
+            # v3.6.10: runtime hot-reload — rebuild the proxy HookChain whenever
+            # optimize.json changes so cache_enabled / compress_enabled can be
+            # toggled INDEPENDENTLY from the UI with no restart. UI save fires the
+            # callback immediately; external edits are caught by the 2s watchdog.
+            _opt_store.register_change_callback(_proxy.reload_from_config)
+            _opt_store.start_watchdog()
             logger.info(
-                "optimize.proxy mounted on /v1/*, /v1beta/*  port=8765"
+                "optimize.proxy mounted on /v1/*, /v1beta/*  port=8765 "
+                "(runtime cache/compress hot-reload enabled)"
             )
         else:
             application.state.optimize_proxy = None
@@ -1210,10 +1236,39 @@ def create_app() -> FastAPI:
         from superlocalmemory.mcp.server import server as _mcp_fastmcp
         _mcp_fastmcp.settings.streamable_http_path = "/"
         _mcp_fastmcp._session_manager = None  # Defensive reset for idempotency
+        # v3.6.9 (#36): configure DNS-rebinding protection from env.
+        # Default: localhost-only (safe). Set SLM_MCP_ALLOWED_HOSTS=192.168.x.y:*
+        # (comma-separated, e.g. "192.168.50.144:*,slm.lan:*") to open to a LAN.
+        # Use "*" to disable protection entirely (trusted private network only).
+        # TransportSecuritySettings imported lazily here so that MCP mount
+        # works on older SDK versions when SLM_MCP_ALLOWED_HOSTS is not set.
+        _mcp_allowed = os.environ.get("SLM_MCP_ALLOWED_HOSTS", "").strip()
+        if _mcp_allowed:
+            from mcp.server.transport_security import TransportSecuritySettings
+            if _mcp_allowed == "*":
+                _mcp_fastmcp.settings.transport_security = TransportSecuritySettings(
+                    enable_dns_rebinding_protection=False,
+                )
+            else:
+                _hosts = [h.strip() for h in _mcp_allowed.split(",") if h.strip()]
+                _mcp_fastmcp.settings.transport_security = TransportSecuritySettings(
+                    enable_dns_rebinding_protection=True,
+                    allowed_hosts=_hosts,
+                    allowed_origins=[f"http://{h}" for h in _hosts],
+                )
+            logger.info("MCP transport security: allowed_hosts=%r", _mcp_allowed)
         global _mcp_app
         _mcp_app = _mcp_fastmcp.streamable_http_app()
-        application.mount("/mcp", _mcp_app)
-        logger.info("MCP HTTP transport mounted at /mcp (Streamable HTTP, port 8765)")
+
+        # v3.6.10: per-agent-ID routing — /mcp/{agent_id} extracts the agent
+        # identity from the URL path and places it in a ContextVar so all MCP
+        # tools (remember, recall, etc.) automatically use the correct namespace.
+        # AgentIDExtractorASGI lives in mcp/agent_context so it is unit-testable
+        # (tests/test_mcp/test_agent_context.py) rather than buried inline here.
+        from superlocalmemory.mcp.agent_context import AgentIDExtractorASGI
+
+        application.mount("/mcp", AgentIDExtractorASGI(_mcp_app))
+        logger.info("MCP HTTP transport mounted at /mcp (Streamable HTTP, port 8765; per-agent routing enabled)")
     except Exception as _mcp_exc:  # pragma: no cover — defensive
         logger.warning("MCP HTTP mount failed (non-fatal, stdio still works): %s", _mcp_exc)
 
@@ -2222,7 +2277,9 @@ if __name__ == "__main__":
     # freshly-sized file.
     rotate_oversized_logs()
     logging.basicConfig(level=logging.INFO, format="%(asctime)s %(message)s")
-    port = _DEFAULT_PORT
+    # v3.6.9 (#33): honour SLM_DAEMON_PORT env so operators can configure the
+    # port without changing the launch command. --port= arg takes precedence.
+    port = int(os.environ.get("SLM_DAEMON_PORT", "") or _DEFAULT_PORT)
     for arg in sys.argv:
         if arg.startswith("--port="):
             port = int(arg.split("=")[1])

package/src/superlocalmemory/ui/index.html

@@ -1248,6 +1248,16 @@
                                     <input class="form-check-input" type="checkbox" id="opt-enabled">
                                     <label class="form-check-label" for="opt-enabled">Optimize Enabled</label>
                                 </div>
+                                <div class="form-check form-switch mb-2">
+                                    <input class="form-check-input" type="checkbox" id="opt-proxy-enabled">
+                                    <label class="form-check-label" for="opt-proxy-enabled">
+                                        Proxy Enabled
+                                        <span class="text-muted small ms-1">(restart required)</span>
+                                    </label>
+                                </div>
+                                <div id="opt-restart-notice" class="alert alert-warning py-1 px-2 small mb-2 d-none">
+                                    Proxy setting changed. Run <code>slm restart</code> to apply.
+                                </div>
                                 <div class="form-check form-switch mb-2">
                                     <input class="form-check-input" type="checkbox" id="opt-cache-enabled">
                                     <label class="form-check-label" for="opt-cache-enabled">Cache Enabled</label>
@@ -1263,26 +1273,14 @@
                                 <div class="mb-2">
                                     <label for="opt-compress-mode" class="form-label">Compression Mode</label>
                                     <select class="form-select form-select-sm" id="opt-compress-mode">
-                                        <option value="safe">Safe</option>
-                                        <option value="aggressive">Aggressive</option>
+                                        <option value="safe">Safe (lossless)</option>
+                                        <option value="aggressive">Aggressive (LLMLingua-2, prose only)</option>
                                     </select>
                                 </div>
-                                <div class="form-check form-switch mb-2">
-                                    <input class="form-check-input" type="checkbox" id="opt-compress-code">
-                                    <label class="form-check-label" for="opt-compress-code">Code Compression</label>
-                                </div>
                                 <div class="form-check form-switch mb-2">
                                     <input class="form-check-input" type="checkbox" id="opt-compress-prose">
                                     <label class="form-check-label" for="opt-compress-prose">Prose Compression</label>
                                 </div>
-                                <div class="form-check form-switch mb-2">
-                                    <input class="form-check-input" type="checkbox" id="opt-compress-ccr">
-                                    <label class="form-check-label" for="opt-compress-ccr">CCR</label>
-                                </div>
-                                <div class="form-check form-switch mb-2">
-                                    <input class="form-check-input" type="checkbox" id="opt-compress-align">
-                                    <label class="form-check-label" for="opt-compress-align">Alignment Compression</label>
-                                </div>
                             </div>
                             <div class="col-md-6">
                                 <h6>Savings</h6>
@@ -1299,6 +1297,12 @@
                                 <div class="small text-muted">Config version: <span id="opt-config-version">-</span></div>
                                 <div id="opt-stale-warning" class="text-warning small mt-1"></div>
                                 <button class="btn btn-sm btn-outline-primary mt-2" id="opt-copy-url">Copy URL</button>
+                                <div class="mt-2 small text-muted">
+                                    <strong>Point agents at the proxy:</strong><br>
+                                    <code>slm wrap claude</code> — Claude Code<br>
+                                    <code>--openai-api-base http://localhost:8765/v1</code> — Aider<br>
+                                    <code>OPENAI_BASE_URL=http://localhost:8765/v1</code> — OpenAI SDK
+                                </div>
                             </div>
                         </div>
                     </div>

package/src/superlocalmemory/ui/js/auto-settings.js

@@ -324,10 +324,12 @@ async function testConnection() {
     if (resultEl) { resultEl.textContent = 'Testing...'; resultEl.className = 'ms-2 small text-muted'; }
 
     try {
+        var testBody = {provider: provider, model: model};
+        if (apiKey) testBody.api_key = apiKey;
         var resp = await fetch('/api/v3/provider/test', {
             method: 'POST',
             headers: {'Content-Type': 'application/json'},
-            body: JSON.stringify({provider: provider, model: model, api_key: apiKey})
+            body: JSON.stringify(testBody)
         });
         var data = await resp.json();
         if (data.success) {

package/src/superlocalmemory/ui/js/ng-shell.js

@@ -338,6 +338,9 @@
   // ── Lazy Load Tab Data ─────────────────────────────────────
   function triggerTabLoad(tabId) {
     switch(tabId) {
+      case 'brain-pane':
+        if (typeof loadBrain === 'function') loadBrain();
+        break;
       case 'graph-pane':
         if (typeof loadGraph === 'function') loadGraph();
         // v3.4.4: Initialize chat panel if not already present

package/src/superlocalmemory/ui/js/optimize.js

@@ -23,14 +23,12 @@
       if (!resp.ok) return;
       var cfg = await resp.json();
       _setToggle('opt-enabled', cfg.enabled);
+      _setToggle('opt-proxy-enabled', cfg.proxy_enabled);
       _setToggle('opt-cache-enabled', cfg.cache_enabled);
       _setToggle('opt-semantic-enabled', cfg.semantic_enabled);
       _setToggle('opt-compress-enabled', cfg.compress_enabled);
       _setSelect('opt-compress-mode', cfg.compress_mode);
-      _setToggle('opt-compress-code', cfg.compress_code);
       _setToggle('opt-compress-prose', cfg.compress_prose);
-      _setToggle('opt-compress-ccr', cfg.compress_ccr);
-      _setToggle('opt-compress-align', cfg.compress_align);
       var verEl = document.getElementById('opt-config-version');
       if (verEl) verEl.textContent = cfg.config_version || '-';
     } catch (e) {
@@ -70,14 +68,12 @@
     var val = e.target.checked;
 
     var fieldMap = {
-      'opt-enabled': 'enabled',
-      'opt-cache-enabled': 'cache_enabled',
+      'opt-enabled':          'enabled',
+      'opt-proxy-enabled':    'proxy_enabled',
+      'opt-cache-enabled':    'cache_enabled',
       'opt-semantic-enabled': 'semantic_enabled',
       'opt-compress-enabled': 'compress_enabled',
-      'opt-compress-code': 'compress_code',
-      'opt-compress-prose': 'compress_prose',
-      'opt-compress-ccr': 'compress_ccr',
-      'opt-compress-align': 'compress_align'
+      'opt-compress-prose':   'compress_prose'
     };
 
     if (id === 'opt-compress-mode') {
@@ -109,6 +105,10 @@
       var body = {};
       body[field] = val;
       _putConfig(body);
+      if (id === 'opt-proxy-enabled' || id === 'opt-enabled') {
+        var notice = document.getElementById('opt-restart-notice');
+        if (notice) notice.classList.remove('d-none');
+      }
     }
   });
 

package/src/superlocalmemory.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: superlocalmemory
-Version: 3.6.8
+Version: 3.6.10
 Summary: Information-geometric agent memory with mathematical guarantees
 Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
 License: AGPL-3.0-or-later
@@ -58,6 +58,7 @@ Requires-Dist: huggingface_hub==0.36.2
 Requires-Dist: torch==2.11.0
 Requires-Dist: scikit-learn==1.8.0
 Requires-Dist: sqlite-vec==0.1.9
+Requires-Dist: llmlingua==0.2.2
 Provides-Extra: search
 Requires-Dist: sentence-transformers==5.3.0; extra == "search"
 Requires-Dist: optimum==2.1.0; extra == "search"
@@ -125,18 +126,22 @@ Dynamic: license-file
 <details>
 <summary><strong>What's New in V3.6 — Optimize: SKIP, SHRINK, DISCOUNT, REMEMBER</strong> (click to expand)</summary>
 
-> V3.6 is the only local-first layer that SKIPS repeat LLM calls (cache: 100% saved), SHRINKS prompts 60-95% (compress: extractive + LLMLingua-2), and DISCOUNTS prefix costs (align: native KV-cache) — and remembers everything — in one install. **Your first cache hit pays for the install time. Hours of coding on repeat, minimal API cost.**
+> V3.6 is the only local-first layer that SKIPS repeat LLM calls (cache: 100% saved on a hit), SHRINKS prose prompts (compress: lossless-by-default, opt-in LLMLingua-2), and DISCOUNTS prefix costs (align: native KV-cache) — and remembers everything — in one install. **Your first cache hit pays for the install time. Hours of coding on repeat, minimal API cost.**
+>
+> **v3.6.10:** cache and compression are now **independent runtime switches** (cache-only, compress-only, both, or neither — toggle live from the dashboard, no restart). Compression was rebuilt to be **lossless by default** (the old string/array/code truncation is gone); aggressive mode adds LLMLingua-2 for **prose only** — never code, numbers, structured data, or the current turn.
 
 ### The Three Levers
 
 | Lever | Mechanism | Saving | Off by default? |
 |-------|-----------|:------:|:---------------:|
-| **Cache** | Skip repeat calls — exact-match SQLite lookup, vCache-gated semantic (opt-in) | **100% on a hit** (input + output) | Cache ON, Semantic OFF |
-| **Compress** | Shrink prompts — extractive JSON/code (lossless) + LLMLingua-2 prose (opt-in) | **60–95% on a miss** (input only) | Safe mode ON, Aggressive OFF |
+| **Cache** | Skip repeat calls — exact-match SQLite lookup (zero false hits), vCache-gated semantic (opt-in) | **100% on a hit** (input + output) | Cache ON, Semantic OFF |
+| **Compress** | Shrink prompts — **safe = lossless** normalization; **aggressive = LLMLingua-2 prose only** (opt-in) | Safe: small + lossless · Aggressive: large on prose | Safe mode, Aggressive OFF |
 | **Align** | Stabilize prefix — maximize provider prefix-cache discounts | **Lossless extra** | ON when compression is ON |
 
 **Memory** (v3.5's existing engine) runs in parallel — it shapes *what is in* the prompt (relevant facts); Optimize decides *whether and how* it is sent.
 
+> **Independent at runtime:** enable caching only, compression only, both, or neither — from the dashboard Optimize tab, applied live (no restart). Each AI client can also get its own memory identity over HTTP MCP via `http://127.0.0.1:8765/mcp/{agent_id}`.
+
 ### Quick Start
 
 ```bash
@@ -152,7 +157,7 @@ slm wrap claude
 |:--------|:-------------|
 | `slm optimize status\|on\|off\|savings` | Master Optimize control + savings report (USD/INR/tokens) |
 | `slm cache status\|clear\|invalidate\|ttl\|semantic` | Cache sub-control — exact + semantic tiers |
-| `slm compress status\|mode\|code\|prose\|ccr\|align` | Compression control — per-channel toggles |
+| `slm compress status\|mode\|prose` | Compression control — safe (lossless) / aggressive (LLMLingua-2 prose) |
 | `slm proxy [--port] [--provider]` | Start the interception proxy (port 8765) |
 | `slm wrap <agent>` | Proxy-activate an agent — one command to start saving |
 | `slm help-optimize [topic]` | Full developer reference + per-agent setup recipes |

package/src/superlocalmemory.egg-info/SOURCES.txt

@@ -271,6 +271,7 @@ src/superlocalmemory/mcp/__init__.py
 src/superlocalmemory/mcp/_daemon_proxy.py
 src/superlocalmemory/mcp/_pool_adapter.py
 src/superlocalmemory/mcp/_stdin_guard.py
+src/superlocalmemory/mcp/agent_context.py
 src/superlocalmemory/mcp/resources.py
 src/superlocalmemory/mcp/server.py
 src/superlocalmemory/mcp/shared.py
@@ -309,8 +310,6 @@ src/superlocalmemory/optimize/cache/stampede.py
 src/superlocalmemory/optimize/compress/__init__.py
 src/superlocalmemory/optimize/compress/align.py
 src/superlocalmemory/optimize/compress/ccr.py
-src/superlocalmemory/optimize/compress/extractive_code.py
-src/superlocalmemory/optimize/compress/extractive_json.py
 src/superlocalmemory/optimize/compress/prose_llmlingua.py
 src/superlocalmemory/optimize/compress/router.py
 src/superlocalmemory/optimize/config/__init__.py
@@ -325,6 +324,7 @@ src/superlocalmemory/optimize/metrics/persistence.py
 src/superlocalmemory/optimize/proxy/__init__.py
 src/superlocalmemory/optimize/proxy/_helpers.py
 src/superlocalmemory/optimize/proxy/anthropic_surface.py
+src/superlocalmemory/optimize/proxy/capture.py
 src/superlocalmemory/optimize/proxy/gemini_surface.py
 src/superlocalmemory/optimize/proxy/lifecycle.py
 src/superlocalmemory/optimize/proxy/openai_surface.py

package/src/superlocalmemory.egg-info/requires.txt

@@ -29,6 +29,7 @@ huggingface_hub==0.36.2
 torch==2.11.0
 scikit-learn==1.8.0
 sqlite-vec==0.1.9
+llmlingua==0.2.2
 
 [dev]
 pytest>=8.0