superlocalmemory 3.6.15 → 3.6.17

package/CHANGELOG.md

@@ -5,6 +5,31 @@ All notable changes to SuperLocalMemory V3 will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.6.17] - 2026-06-21 — Community PR round + dashboard-feedback fix + SQLite tuning
+
+Eight community pull requests merged after line-by-line review, plus fixes for the open issues. Every change was validated against the full test suite under the real 3.12 runtime; default single-machine behavior is unchanged.
+
+### Added
+
+- **HTTP write-path observability** (PR #52, @barrygfox). The HTTP fast paths (`/observe`, `/remember`, the AutoCapture pipeline, the materializer) now emit `EventBus` events tagged `source_protocol="http"`, so the dashboard event stream is no longer structurally empty. New event types: `memory.observed`, `memory.captured`, `memory.dropped`, `memory.queued`. Emission is best-effort and never affects the caller's response.
+- **Marker-bounded adapter writes** (PR #54, @barrygfox). `CopilotAdapter` now wraps its content in `<!-- SLM-START -->` / `<!-- SLM-END -->` markers and merges into `.github/copilot-instructions.md` instead of overwriting it, preserving user- and agent-curated content. `disable()` strips the SLM block instead of deleting the file. New `hooks/memory_protocol.py` is the shared single source of truth for the marker contract.
+
+### Fixed
+
+- **Dashboard feedback was completely broken** (issues #53/#59). The dashboard thumbs-up/down/pin and dwell handlers called `FeedbackCollector.record_dashboard_feedback()` — a method that did not exist, so every write raised `AttributeError` (caught by the route, so no lock leak, but the feature was dead). Implemented the method, mapping the dashboard vocabulary onto stored `(signal_type, value)` pairs; the raw query is hashed, never stored.
+- **NULL columns reloaded as `[]` instead of `None`** (PR #50, @barrygfox). `_jl()` collapsed "no default" and explicit `default=None`, defeating downstream `is None` guards and causing `Mean of empty slice` warnings in the Fisher–Langevin coupling. Fixed with a `_MISSING` sentinel + an empty-array guard.
+- **Lifecycle hooks hard-coded daemon port `:8765`** (PR #51, @barrygfox). Hooks now resolve the port from the per-user `~/.superlocalmemory/daemon.port` file; the non-loopback SSRF guard on `SLM_HOOK_DAEMON_URL` is preserved.
+- **`atomic_write` honored a stale sync-log skip** (PR #55, @barrygfox). The on-disk file is re-hashed before a durable skip, so an out-of-band edit (`git restore`, manual edit) is no longer silently ignored.
+- **Embedding/reranker workers ran single-threaded** (PR #56, @barrygfox). `OMP_NUM_THREADS` is restored in those subprocess workers (which load torch but never lightgbm, so the libomp SIGSEGV cannot occur).
+- **Anthropic provider ignored `api_base`** (PR #57, @barrygfox). The Anthropic backbone now honors a configured base URL (Anthropic-compatible proxy), mirroring the OpenAI provider.
+- **Dashboard screenshot committed as a raw binary** (PR #58, @MelleKoning). Converted to a Git LFS pointer per the existing `.gitattributes` rule.
+- **"Test Connection" blocked for remote dashboards** (issue #40 residue). In `SLM_REMOTE` mode, an allowlisted LAN client may probe its own LAN LLM endpoint, exactly like the loopback dashboard. The SSRF guard is not relaxed for any non-allowlisted caller.
+
+### Changed
+
+- **SQLite endurance knobs are env-tunable** (issue #53). `SLM_DB_BUSY_TIMEOUT_MS`, `SLM_DB_MAX_RETRIES`, and `SLM_DB_RETRY_BASE_DELAY` override the defaults for operators on slow/contended I/O. Unset env is byte-identical to the prior hard-coded constants.
+- **`outcome_queue` polling backs off when idle** (issue #53). The drain worker now relaxes its 0.25s poll (doubling, capped at 2s) when the queue drains empty and snaps back to 0.25s the instant there is work, reducing idle contention on the shared SQLite file.
+
 ## [3.6.14] - 2026-06-18 — Audit-hardened: memory bounds, cross-tenant cache isolation, atomic credentials
 
 Shipped through two adversarial audit passes (Qualixar Iron Pattern Stages 8–9), validated against a green 5933-test suite under the real 3.12 runtime. Default single-machine behavior is unchanged.

package/README.md

@@ -2,10 +2,10 @@
   <img src="https://superlocalmemory.com/assets/logo-mark.png" alt="SuperLocalMemory" width="200"/>
 </p>
 
-<h1 align="center">SuperLocalMemory V3.6.15</h1>
+<h1 align="center">SuperLocalMemory V3.6.17</h1>
 <p align="center"><strong>Cache. Compress. Remember. Three surfaces — proxy, MCP tools, or skill. Every setup covered.</strong><br/>
 <em>To the best of our knowledge, the only zero-cloud agent memory that beats Mem0's zero-LLM score on LoCoMo. Mode A: 74.8% vs Mem0 64.2% — no GPU, no API key, on CPU.</em></p>
-<p align="center"><code>v3.6.15</code> — <strong>Plugin-native. Profile-aware. Distributed-ready.</strong><br/>
+<p align="center"><code>v3.6.17</code> — <strong>Plugin-native. Profile-aware. Distributed-ready.</strong><br/>
 Proxy: <code>slm wrap claude</code> &nbsp;·&nbsp; MCP: add <code>slm_compress</code> to your config &nbsp;·&nbsp; Skill: zero-config</p>
 <p align="center"><strong>3 published research papers</strong> (arXiv preprints + Zenodo-archived) · <a href="https://arxiv.org/abs/2603.02240">arXiv:2603.02240</a> · <a href="https://arxiv.org/abs/2603.14588">arXiv:2603.14588</a> · <a href="https://arxiv.org/abs/2604.04514">arXiv:2604.04514</a></p>
 
@@ -162,7 +162,7 @@ Full docs: [docs/multi-machine.md](docs/multi-machine.md) · [docs/distributed-d
 |:-----|:--------|:-----|
 | **npm** (recommended) | `npm install -g superlocalmemory` | Node 14+, installs Python deps automatically |
 | **pip** | `pip install superlocalmemory` | Python 3.11+, direct install |
-| **Claude Code Plugin** (WP-06) | `/plugin install superlocalmemory@qualixar` | Self-bootstraps venv, isolated SLM_DATA_DIR, additive — 14-tool core |
+| **Claude Code Plugin** (WP-06) | `/plugin marketplace add qualixar/superlocalmemory` then `/plugin install superlocalmemory@qualixar` | Self-bootstraps venv, isolated SLM_DATA_DIR, additive — 14-tool core. Ships the skills/agents/hooks/commands |
 | **Portable / IDE connect** (WP-08) | `slm connect <ide> [--here]` | Wire any IDE without reinstalling; `slm connect claude-code` → plugin pointer |
 
 After any install path: `slm setup` → `slm doctor` → `slm warmup` (optional, pre-downloads ~500MB embedding model).
@@ -218,18 +218,31 @@ Per-IDE configs available for Claude Code, Cursor, Windsurf, VS Code Copilot, Co
 
 ## Claude Code Plugin
 
-Install directly in Claude Code without a system-level npm/pip install:
+Install directly in Claude Code — no system-level npm/pip needed. This is how you
+get the **skills, agents, hooks, commands, and rules** (the MCP server is
+bootstrapped automatically). It is a two-step flow — add the marketplace once,
+then install:
 
 ```bash
+# 1. Add the Qualixar marketplace (one-time — the repo IS the marketplace)
+/plugin marketplace add qualixar/superlocalmemory
+
+# 2. Install the plugin
 /plugin install superlocalmemory@qualixar
 ```
 
 - Self-bootstraps a Python venv, installs all deps in an isolated `SLM_DATA_DIR`
-- Registers 14-tool core MCP surface (`core14` profile by default)
+- Registers the 14-tool core MCP surface (`core14` profile by default)
+- Ships the SLM skills / agents / hooks / commands / rules
 - Additive — does not replace an existing SLM install
 - `slm connect claude-code` detects an existing plugin install and links them
 
-See [docs/getting-started.md](docs/getting-started.md) for full plugin walkthrough.
+> **Plugin vs `pip`/`npm`:** `pip install superlocalmemory` / `npm i -g superlocalmemory`
+> give you the `slm` CLI + the MCP server (the *tools*). The **skills/agents/hooks/
+> commands** come only through the plugin above. Use the plugin for Claude Code; use
+> pip/npm for the CLI or other IDEs.
+
+To update later: `/plugin marketplace update qualixar` then `/plugin install superlocalmemory@qualixar`.
 
 ---
 
@@ -294,6 +307,8 @@ slm dashboard    # Opens at http://localhost:8765
 
 | Version | Codename | Key Features |
 |---|---|---|
+| **v3.6.17** | Community | 8 contributor PRs (observability events, marker-bounded adapter writes, daemon port discovery, anthropic `api_base`, OpenMP workers, atomic-write rehash, `_jl` sentinel, LFS pointer); dashboard-feedback fix (#53/#59); env-tunable SQLite knobs + idle backoff; remote LLM test-probe (#40) |
+| **v3.6.16** | Docs | Corrected Claude Code plugin install — adds the required `/plugin marketplace add` step; clarifies plugin vs pip/npm delivery |
 | **v3.6.15** | Multi-scope | **Opt-in [shared memory](docs/shared-memory.md)** (personal/shared/global, off by default), default-deny scope at every read path, recall scope-race fix, contributor PRs #42/#43/#44, fixes #46–#49 |
 | **v3.6.14** | Plugin-native | Claude Code Plugin (WP-06), MCP profiles (WP-01), IDE connect (WP-08), asset consolidation, UI polish (WP-12) |
 | **v3.6.x** | Optimize Everywhere / Distributed-ready | Three surfaces (proxy/MCP/skill), `SLM_REMOTE=1` LAN mode, remote dashboard, custom LLM endpoints |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "superlocalmemory",
-  "version": "3.6.15",
+  "version": "3.6.17",
   "description": "Information-geometric agent memory with mathematical guarantees. 4-channel retrieval, Fisher-Rao similarity, zero-LLM mode, EU AI Act compliant. Works with Claude, Cursor, Windsurf, and 17+ AI tools.",
   "keywords": [
     "ai-memory",

package/plugin/.claude-plugin/plugin.json

@@ -16,5 +16,5 @@
   "mcpServers": "./.mcp.json",
   "name": "superlocalmemory",
   "repository": "https://github.com/qualixar/superlocalmemory",
-  "version": "3.6.15"
+  "version": "3.6.17"
 }

package/plugin/CLAUDE.md

@@ -1,4 +1,4 @@
-<!-- BEGIN SuperLocalMemory v3.6.15 -->
+<!-- BEGIN SuperLocalMemory v3.6.17 -->
 
 ## SuperLocalMemory (SLM) — Agent Rules
 
@@ -39,6 +39,6 @@ slm-recall · slm-remember · slm-session · slm-status · slm-cache · slm-comp
 ### Subagents
 slm-memory-advisor (memory decisions, session hygiene) · slm-optimize-advisor (context compression + KV cache)
 
-<!-- END SuperLocalMemory v3.6.15 -->
+<!-- END SuperLocalMemory v3.6.17 -->
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin/agents/slm-memory-advisor.md

@@ -41,4 +41,4 @@ recall→`slm recall "<q>" --limit N` (add `--include-global`/`--include-shared`
 # What NOT to do
 Never session_init twice; never forget dry_run=False without reporting preview; never dump a whole file into remember; never invent a memory; never claim "saved" without success:true / clean CLI exit.
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin/agents/slm-optimize-advisor.md

@@ -35,4 +35,4 @@ stats→`slm optimize status`/`savings` · compress→`slm compress` · cache→
 # 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 %.
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin/requirements.txt

@@ -1 +1 @@
-superlocalmemory==3.6.15
+superlocalmemory==3.6.17

package/plugin/skills/slm-cache/SKILL.md

@@ -137,4 +137,4 @@ These subcommands control daemon-level cache settings. They do not read or write
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin/skills/slm-compress/SKILL.md

@@ -140,4 +140,4 @@ Content over 1 MB (1 000 000 bytes UTF-8) is processed but `reversible` is force
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin/skills/slm-graph/SKILL.md

@@ -297,4 +297,4 @@ Without this, the six graph tools are not registered and will appear as unknown
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin/skills/slm-recall/SKILL.md

@@ -201,4 +201,4 @@ trusts that what you surface came from the store.
 
 ---
 
-*SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later*

package/plugin/skills/slm-remember/SKILL.md

@@ -191,4 +191,4 @@ slm remember "<content>" --scope shared --shared-with alice,bob
 
 ---
 
-*SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later*

package/plugin/skills/slm-session/SKILL.md

@@ -204,4 +204,4 @@ slm doctor [--json]    # preflight check including daemon and embedding worker
 
 ---
 
-*SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later*

package/plugin/skills/slm-status/SKILL.md

@@ -146,4 +146,4 @@ Note: the `slm optimize` subcommands have known pre-existing parse-test failures
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/agents/slm-memory-advisor.md

@@ -41,4 +41,4 @@ recall→`slm recall "<q>" --limit N` (add `--include-global`/`--include-shared`
 # What NOT to do
 Never session_init twice; never forget dry_run=False without reporting preview; never dump a whole file into remember; never invent a memory; never claim "saved" without success:true / clean CLI exit.
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/agents/slm-optimize-advisor.md

@@ -35,4 +35,4 @@ stats→`slm optimize status`/`savings` · compress→`slm compress` · cache→
 # 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 %.
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/commands/slm-optimize.md

@@ -19,4 +19,4 @@ Rules (apply strictly):
 
 MCP unavailable → CLI fallback: `slm optimize status`.
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/commands/slm-recall.md

@@ -13,4 +13,4 @@ Recall from SuperLocalMemory using the query: $ARGUMENTS
    - `slm recall "$ARGUMENTS" --limit 10`
    - then `slm search "$ARGUMENTS"`
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/commands/slm-remember.md

@@ -13,4 +13,4 @@ Save to SuperLocalMemory: $ARGUMENTS
 5. Confirm only on success:true. If success is not true, report the error — never claim "saved."
 6. MCP unavailable → CLI fallback: `slm remember "$ARGUMENTS" --tags <tags>` (note: `--importance` is MCP-only, not a CLI flag).
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/commands/slm-status.md

@@ -12,4 +12,4 @@ Show SuperLocalMemory status and optimization counters.
 
 Note: MCP get_status is intentionally NOT used here — it is outside the core profile and would error. Use `slm status` (CLI) + `slm_optimize_stats` (MCP) only.
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "3.6.15",
+  "version": "3.6.17",
   "pluginName": "superlocalmemory",
   "displayName": "SuperLocalMemory",
   "repository": "https://github.com/qualixar/superlocalmemory",

package/plugin-src/requirements.txt

@@ -1 +1 @@
-superlocalmemory==3.6.15
+superlocalmemory==3.6.17

package/plugin-src/rules/AGENTS.md

@@ -88,4 +88,4 @@ When the SLM MCP server is unavailable, use these CLI equivalents:
 | `slm_cache_get`    | `key`                                                                          | KV cache get; returns hit, value   |
 | `slm_optimize_stats` | `()`                                                                         | Returns compress_runs, tokens_saved_compress, cache_kv_hits |
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/rules/CLAUDE.md.fragment

@@ -1,4 +1,4 @@
-<!-- BEGIN SuperLocalMemory v3.6.15 -->
+<!-- BEGIN SuperLocalMemory v3.6.17 -->
 
 ## SuperLocalMemory (SLM) — Agent Rules
 
@@ -39,6 +39,6 @@ slm-recall · slm-remember · slm-session · slm-status · slm-cache · slm-comp
 ### Subagents
 slm-memory-advisor (memory decisions, session hygiene) · slm-optimize-advisor (context compression + KV cache)
 
-<!-- END SuperLocalMemory v3.6.15 -->
+<!-- END SuperLocalMemory v3.6.17 -->
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/skills/slm-cache/SKILL.md

@@ -137,4 +137,4 @@ These subcommands control daemon-level cache settings. They do not read or write
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/skills/slm-compress/SKILL.md

@@ -140,4 +140,4 @@ Content over 1 MB (1 000 000 bytes UTF-8) is processed but `reversible` is force
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/skills/slm-graph/SKILL.md

@@ -297,4 +297,4 @@ Without this, the six graph tools are not registered and will appear as unknown
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/plugin-src/skills/slm-recall/SKILL.md

@@ -201,4 +201,4 @@ trusts that what you surface came from the store.
 
 ---
 
-*SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later*

package/plugin-src/skills/slm-remember/SKILL.md

@@ -191,4 +191,4 @@ slm remember "<content>" --scope shared --shared-with alice,bob
 
 ---
 
-*SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later*

package/plugin-src/skills/slm-session/SKILL.md

@@ -204,4 +204,4 @@ slm doctor [--json]    # preflight check including daemon and embedding worker
 
 ---
 
-*SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later*
+*SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later*

package/plugin-src/skills/slm-status/SKILL.md

@@ -146,4 +146,4 @@ Note: the `slm optimize` subcommands have known pre-existing parse-test failures
 
 ---
 
-SuperLocalMemory v3.6.15 · Qualixar · AGPL-3.0-or-later
+SuperLocalMemory v3.6.17 · Qualixar · AGPL-3.0-or-later

package/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "superlocalmemory"
-version = "3.6.15"
+version = "3.6.17"
 description = "Information-geometric agent memory with mathematical guarantees"
 readme = "README.md"
 license = {text = "AGPL-3.0-or-later"}

package/scripts/build-plugin.js

@@ -30,7 +30,7 @@ import process from 'node:process';
 // ---------------------------------------------------------------------------
 // Constants
 // ---------------------------------------------------------------------------
-const VERSION = '3.6.15';
+const VERSION = '3.6.17';
 const MANIFEST_REL = 'plugin-src/manifest.json';
 const GENERATED_BANNER = `# _GENERATED — DO NOT HAND-EDIT
 

package/src/superlocalmemory/__init__.py

@@ -32,7 +32,7 @@ if "OMP_NUM_THREADS" not in os.environ:
     os.environ["OMP_NUM_THREADS"] = "2"
 # ---------------------------------------------------------------------------
 
-__version__ = "3.6.15"
+__version__ = "3.6.17"
 
 _REQUIRED_VERSIONS = {
     "sentence_transformers": "5.3.0",

package/src/superlocalmemory/core/embeddings.py

@@ -473,6 +473,11 @@ class EmbeddingService:
                 "TOKENIZERS_PARALLELISM": "false",
                 "TORCH_DEVICE": "cpu",
                 "ORT_DISABLE_COREML": "1",
+                # Restore parallel OpenMP. The package caps OMP_NUM_THREADS
+                # globally to avoid a torch+lightgbm libomp SIGSEGV in the
+                # main process. This worker loads torch but never lightgbm,
+                # so there is no collision risk and full parallelism is safe.
+                "OMP_NUM_THREADS": str(os.cpu_count() or 4),
             }
             from superlocalmemory.core.platform_utils import popen_platform_kwargs
             self._worker_proc = subprocess.Popen(

package/src/superlocalmemory/dynamics/fisher_langevin_coupling.py

@@ -111,6 +111,8 @@ class FisherLangevinCoupling:
             CouplingState with derived temperature, direction, and weight.
         """
         var_arr = np.asarray(fisher_variance, dtype=np.float64)
+        if var_arr.size == 0:
+            return CouplingState()
 
         # Step 1: Fisher confidence from variance
         # Low variance = high confidence (memory is well-characterized)
@@ -203,6 +205,8 @@ class FisherLangevinCoupling:
             return self._base_temp
 
         var_arr = np.asarray(fisher_variance, dtype=np.float64)
+        if var_arr.size == 0:
+            return self._base_temp
         avg_var = float(np.mean(np.clip(var_arr, 1e-8, None)))
         fisher_conf = min(1.0, 1.0 / (1.0 + avg_var) + min(access_count * 0.02, 0.2))
         return self._base_temp / (fisher_conf + self._epsilon)

package/src/superlocalmemory/hooks/adapter_base.py

@@ -217,9 +217,16 @@ def atomic_write(
     prev = sync_log_last_content_sha256(sync_log_db, adapter_name, target_sha)
 
     if prev == new_hash and resolved_path.exists():
-        # Durable skip — no write, no new sync-log row (the prior row still
-        # reflects on-disk truth).
-        return WriteResult(wrote=False, bytes_written=0, content_sha256=new_hash)
+        # Durable skip only if the on-disk content also matches the new hash.
+        # The sync-log row alone is not authoritative: the file may have been
+        # mutated out-of-band (e.g. ``git restore``, manual edit) since the
+        # last sync. Re-hash the file and re-write if it diverges.
+        try:
+            disk_hash = hashlib.sha256(resolved_path.read_bytes()).hexdigest()
+        except OSError:
+            disk_hash = None
+        if disk_hash == new_hash:
+            return WriteResult(wrote=False, bytes_written=0, content_sha256=new_hash)
 
     resolved_path.parent.mkdir(parents=True, exist_ok=True)
     tmp = resolved_path.with_suffix(resolved_path.suffix + ".slm-tmp")

package/src/superlocalmemory/hooks/copilot_adapter.py

@@ -8,9 +8,15 @@ LLD-05 §6. Verified (verification-2026-04-17.md claim 5): plain markdown,
 no frontmatter, soft 2 KB / hard 4 KB cap. Adapter is INACTIVE when the
 project has no ``.github/`` directory — we do not create it ourselves.
 
+v3.4.23 fix: the SLM-managed content is wrapped in
+``<!-- SLM-START -->`` / ``<!-- SLM-END -->`` markers and merged into the
+host file rather than overwriting it. ``.github/copilot-instructions.md``
+is typically a curated, project-specific document; destructive rewrites
+deleted the user's prose.
+
 Hard rules covered here:
   - A1 / A2 / A3 / A7: via ``adapter_base.atomic_write``.
-  - A4: soft 2 KB + hard 4 KB cap enforcement.
+  - A4: soft 2 KB + hard 4 KB cap enforcement on the SLM section.
 """
 
 from __future__ import annotations
@@ -39,6 +45,12 @@ from superlocalmemory.hooks.context_payload import (
     format_topics,
     truncate_payload_for_cap,
 )
+from superlocalmemory.hooks.memory_protocol import (
+    SLM_MARKER_END,
+    SLM_MARKER_START,
+    memory_protocol_markdown,
+    strip_slm_block as _strip_existing_block,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -52,20 +64,37 @@ _BODY_TEMPLATE = (
     "## Entities\n{entities}\n\n"
     "## Never do\n"
     "- Do not modify files under `.slm/`\n"
-    "- Do not commit `*.slm-cache.db`\n"
+    "- Do not commit `*.slm-cache.db`\n\n"
 )
 
 
 def render_copilot(payload: ContextPayload) -> bytes:
-    return _BODY_TEMPLATE.format(
+    # Two-stage assembly: format the dynamic header (which contains {}
+    # placeholders), then concatenate the static memory-protocol block
+    # verbatim. The memory-protocol block legitimately contains literal
+    # braces (JSON-shaped argument examples for the agent) which must not
+    # be interpreted as format fields.
+    header = _BODY_TEMPLATE.format(
         version=payload.version,
         topics=format_topics(payload),
         entities=format_entities(payload),
-    ).encode("utf-8")
+    )
+    return (header + memory_protocol_markdown()).encode("utf-8")
+
+
+def _wrap_managed(rendered: bytes) -> str:
+    """Wrap rendered SLM content in ``<!-- SLM-START -->`` markers."""
+    return (
+        f"{SLM_MARKER_START}\n"
+        "<!-- Managed by SuperLocalMemory. Edits between SLM-START and "
+        "SLM-END will be overwritten. -->\n\n"
+        f"{rendered.decode('utf-8')}\n"
+        f"{SLM_MARKER_END}\n"
+    )
 
 
 class CopilotAdapter:
-    """Project-scope Copilot adapter."""
+    """Project-scope Copilot adapter (marker-bounded merge)."""
 
     def __init__(
         self,
@@ -124,8 +153,39 @@ class CopilotAdapter:
             )
             rendered = truncate_to_cap(rendered, cap=self._hard_cap)
 
+        # Marker-bounded merge — preserve any user-curated content in the
+        # host file. Strip any prior SLM block(s) and re-append a fresh one.
+        existing = ""
+        if resolved.exists():
+            try:
+                existing = resolved.read_text(encoding="utf-8")
+            except OSError as exc:
+                logger.warning(
+                    "copilot: cannot read %s: %s", resolved, exc,
+                )
+                return False
+
+        # Orphaned start marker — refuse to write rather than corrupt.
+        if (SLM_MARKER_START in existing
+                and SLM_MARKER_END not in existing):
+            logger.warning(
+                "copilot: %s present but %s missing in %s; refusing to write",
+                SLM_MARKER_START, SLM_MARKER_END, resolved,
+            )
+            return False
+
+        stripped = _strip_existing_block(existing)
+        section = _wrap_managed(rendered)
+        if stripped:
+            if not stripped.endswith("\n"):
+                stripped += "\n"
+            # One blank line between user content and the managed section.
+            new_content = stripped + "\n" + section
+        else:
+            new_content = section
+
         result: WriteResult = atomic_write(
-            resolved, rendered,
+            resolved, new_content.encode("utf-8"),
             adapter_name=self.name,
             profile_id=self._profile_id,
             sync_log_db=self._sync_log_db,
@@ -137,11 +197,20 @@ class CopilotAdapter:
             resolved = self.target_path
         except PathTraversalError:
             return
+        # Marker-bounded strip — never delete the host file (user-owned).
         if resolved.exists():
             try:
-                resolved.unlink()
-            except OSError:  # pragma: no cover
-                pass
+                existing = resolved.read_text(encoding="utf-8")
+            except OSError:
+                existing = ""
+            stripped = _strip_existing_block(existing)
+            if stripped != existing:
+                try:
+                    resolved.write_text(stripped, encoding="utf-8")
+                except OSError as exc:  # pragma: no cover
+                    logger.warning(
+                        "copilot: failed to strip on disable: %s", exc,
+                    )
         record_disable(
             resolved,
             adapter_name=self.name,

package/src/superlocalmemory/hooks/hook_handlers.py

@@ -36,7 +36,32 @@ _LAST_CONSOLIDATION = os.path.join(
 )
 
 
-_DAEMON_URL = "http://127.0.0.1:8765"
+_DEFAULT_DAEMON_PORT = 8765
+
+
+def _daemon_url() -> str:
+    """Resolve the daemon base URL, preferring the per-user port file.
+
+    On a shared host each user runs their own daemon bound to a different
+    port; the active port is written to ``~/.superlocalmemory/daemon.port``
+    at startup. Reading it here keeps lifecycle hooks pointed at the
+    caller's own daemon instead of a hard-coded ``8765`` that may belong to
+    another user's instance. Falls back to the default port when the file is
+    absent or unreadable. Stdlib only — no SLM imports in the hot path.
+    """
+    port = _DEFAULT_DAEMON_PORT
+    try:
+        port_file = os.path.join(
+            os.path.expanduser("~"), ".superlocalmemory", "daemon.port",
+        )
+        with open(port_file) as fh:
+            port = int(fh.read().strip())
+    except Exception:
+        pass
+    return f"http://127.0.0.1:{port}"
+
+
+_DAEMON_URL = _daemon_url()
 
 
 def _daemon_post(path: str, body: dict, timeout: float = 3.0) -> bool:

package/src/superlocalmemory/hooks/memory_protocol.py

@@ -0,0 +1,102 @@
+# 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
+
+"""Shared utilities for marker-bounded writes into agent instruction files.
+
+Adapters that inject SLM content into IDE/agent instruction files (e.g.
+``.github/copilot-instructions.md``) use the constants and helpers here to
+demarcate the SLM-managed section so user-curated content outside the
+markers is preserved on every sync.
+
+Marker contract
+---------------
+SLM wraps its content in a pair of HTML comments::
+
+    <!-- SLM-START -->
+    ... managed content ...
+    <!-- SLM-END -->
+
+``strip_slm_block`` removes all such pairs idempotently; adapters call it
+before re-writing so a fresh block replaces the old one in place.
+"""
+
+from __future__ import annotations
+
+import logging
+
+logger = logging.getLogger(__name__)
+
+#: Opening marker for the SLM-managed section.
+SLM_MARKER_START = "<!-- SLM-START -->"
+#: Closing marker for the SLM-managed section.
+SLM_MARKER_END = "<!-- SLM-END -->"
+
+
+def strip_slm_block(text: str) -> str:
+    """Remove all SLM-managed sections from *text*.
+
+    Idempotent — returns *text* unchanged when no markers are present.
+    Strips every ``SLM-START``/``SLM-END`` pair to handle files that
+    accumulated duplicates from a previous bug or a competing writer.
+
+    If a ``SLM-START`` marker has no matching ``SLM-END``, the file is
+    returned unchanged to avoid eating user content; the caller should
+    treat this as an orphaned-marker error and skip the write.
+    """
+    out = text
+    while True:
+        start_idx = out.find(SLM_MARKER_START)
+        if start_idx == -1:
+            return out
+        end_idx = out.find(SLM_MARKER_END, start_idx)
+        if end_idx == -1:
+            logger.warning(
+                "memory_protocol: %s found but %s missing; leaving file unchanged",
+                SLM_MARKER_START,
+                SLM_MARKER_END,
+            )
+            return text
+        cut_end = end_idx + len(SLM_MARKER_END)
+        if cut_end < len(out) and out[cut_end] == "\n":
+            cut_end += 1
+        # Pull back up to two leading newlines added as a boundary separator.
+        cut_start = start_idx
+        while cut_start > 0 and out[cut_start - 1] == "\n":
+            cut_start -= 1
+            if start_idx - cut_start >= 2:
+                break
+        out = out[:cut_start] + out[cut_end:]
+
+
+def memory_protocol_markdown() -> str:
+    """Return the agent-facing Markdown memory protocol block.
+
+    Embedded verbatim into Markdown instruction files such as
+    ``.github/copilot-instructions.md``.  Trailing newline included so
+    callers can concatenate without worrying about boundary whitespace.
+    """
+    return (
+        "## Memory protocol\n"
+        "SLM tools are available via the `slm-hub` MCP gateway. Use them to "
+        "make this brain context grow across sessions.\n\n"
+        "- **At the start of work on an unfamiliar area**, call "
+        "`hub__call_tool` with `tool=\"slm__recall\"` and "
+        "`arguments={\"query\": \"<topic>\"}` to surface prior decisions "
+        "and patterns.\n"
+        "- **At the end of a substantial task** (a fix, a decision, a "
+        "non-trivial change, a session conclusion), call `hub__call_tool` "
+        "with `tool=\"slm__remember\"` and `arguments={\"content\": "
+        "\"<one-paragraph summary of what was decided / changed / "
+        "learned>\", \"tags\": \"<comma-separated kebab-case keywords>\"}`.\n"
+        "- A \"substantial task\" is anything you would write a commit "
+        "message or handoff note about — not every tool call.\n"
+    )
+
+
+__all__ = (
+    "SLM_MARKER_START",
+    "SLM_MARKER_END",
+    "strip_slm_block",
+    "memory_protocol_markdown",
+)

package/src/superlocalmemory/hooks/post_tool_async_hook.py

@@ -30,6 +30,24 @@ _ALLOWED_DAEMON_HOSTS: frozenset[str] = frozenset({
     "127.0.0.1", "localhost", "::1", "[::1]",
 })
 
+_DEFAULT_DAEMON_PORT = 8765
+
+
+def _port_file_url() -> str:
+    """Loopback daemon URL from the per-user port file (default 8765).
+
+    On a shared host each user runs their own daemon on a different port,
+    written to ``~/.superlocalmemory/daemon.port`` at startup. Falling back
+    to this instead of a hard-coded ``8765`` keeps the hook pointed at the
+    caller's own daemon. Stdlib only.
+    """
+    port = _DEFAULT_DAEMON_PORT
+    try:
+        port = int((Path.home() / ".superlocalmemory" / "daemon.port").read_text().strip())
+    except Exception:
+        pass
+    return f"http://127.0.0.1:{port}"
+
 
 def _sanitised_daemon_url() -> str:
     """Return the configured daemon URL only if it's loopback-scoped.
@@ -38,21 +56,21 @@ def _sanitised_daemon_url() -> str:
     shell profile) could set ``SLM_HOOK_DAEMON_URL`` to a remote host
     and exfiltrate the install token via the ``X-SLM-Hook-Token``
     header. We refuse any non-loopback URL and fall back to the local
-    daemon.
+    daemon (resolved via the per-user port file, not a hard-coded port).
     """
     raw = os.environ.get("SLM_HOOK_DAEMON_URL", "").strip()
     if not raw:
-        return "http://127.0.0.1:8765"
+        return _port_file_url()
     try:
         from urllib.parse import urlparse
         parsed = urlparse(raw)
     except Exception:  # pragma: no cover — urllib always importable
-        return "http://127.0.0.1:8765"
+        return _port_file_url()
     if parsed.scheme not in ("http", "https"):
-        return "http://127.0.0.1:8765"
+        return _port_file_url()
     host = (parsed.hostname or "").lower()
     if host not in _ALLOWED_DAEMON_HOSTS:
-        return "http://127.0.0.1:8765"
+        return _port_file_url()
     # Preserve the scheme + port (user may bind daemon on a non-default port).
     port = f":{parsed.port}" if parsed.port else ""
     return f"{parsed.scheme}://{host}{port}"

package/src/superlocalmemory/infra/event_bus.py

@@ -32,6 +32,10 @@ VALID_EVENT_TYPES = frozenset([
     "memory.updated",       # Existing memory modified
     "memory.deleted",       # Memory removed
     "memory.recalled",      # Memory retrieved by an agent
+    "memory.observed",      # /observe accepted content into the debounce buffer
+    "memory.captured",      # AutoCapture matched a buffered observation
+    "memory.dropped",       # AutoCapture rejected a buffered observation
+    "memory.queued",        # /remember accepted content into pending.db (async)
     "graph.updated",        # Knowledge graph rebuilt
     "pattern.learned",      # New pattern detected
     "agent.connected",      # New agent connects

package/src/superlocalmemory/learning/feedback.py

@@ -41,6 +41,18 @@ SIGNAL_VALUES: Dict[str, float] = {
     "access_pattern": 0.6,
 }
 
+# Dashboard UI vocabulary -> (signal_type, signal_value). The dashboard speaks
+# thumbs_up/thumbs_down/pin (explicit) and dwell_positive/dwell_negative
+# (derived from modal dwell time). Unknown types fall back to a neutral
+# user_correction signal rather than being dropped.
+_DASHBOARD_SIGNAL_MAP: Dict[str, tuple[str, float]] = {
+    "thumbs_up": ("user_positive", 1.0),
+    "thumbs_down": ("user_negative", 0.0),
+    "pin": ("user_pin", 1.0),
+    "dwell_positive": ("dwell_positive", 0.6),
+    "dwell_negative": ("dwell_negative", 0.2),
+}
+
 _CREATE_TABLE = """
 CREATE TABLE IF NOT EXISTS learning_feedback (
     id           INTEGER PRIMARY KEY AUTOINCREMENT,
@@ -219,6 +231,53 @@ class FeedbackCollector:
             finally:
                 conn.close()
 
+    # ------------------------------------------------------------------
+    # Public API: record dashboard feedback
+    # ------------------------------------------------------------------
+
+    def record_dashboard_feedback(
+        self,
+        memory_id: str,
+        query: str = "",
+        feedback_type: str = "",
+        profile_id: str = "default",
+    ) -> Optional[int]:
+        """Record an explicit feedback signal raised from the dashboard UI.
+
+        Maps the dashboard's vocabulary (``thumbs_up``/``thumbs_down``/``pin``
+        and the dwell-derived ``dwell_positive``/``dwell_negative``) onto a
+        stored ``(signal_type, signal_value)`` pair. ``memory_id`` is the fact
+        id; the raw ``query`` is hashed and never stored. Returns the inserted
+        row id, or ``None`` on missing ``memory_id``.
+
+        This method restores the dashboard feedback path: the HTTP routes in
+        ``server/routes/learning.py`` called it before it existed, so every
+        thumbs/pin/dwell write raised ``AttributeError`` (issues #53/#59).
+        """
+        if not memory_id:
+            return None
+        signal_type, value = _DASHBOARD_SIGNAL_MAP.get(
+            feedback_type, ("user_correction", 0.5),
+        )
+        qhash = _hash_query(query) if query else None
+        now = _utcnow_iso()
+
+        with self._lock:
+            conn = self._connect()
+            try:
+                cursor = conn.execute(
+                    "INSERT INTO learning_feedback "
+                    "(profile_id, fact_id, signal_type, signal_value, "
+                    "query_hash, created_at, metadata) "
+                    "VALUES (?, ?, ?, ?, ?, ?, ?)",
+                    (profile_id or "default", str(memory_id), signal_type,
+                     value, qhash, now, None),
+                )
+                conn.commit()
+                return cursor.lastrowid
+            finally:
+                conn.close()
+
     # ------------------------------------------------------------------
     # Public API: read feedback
     # ------------------------------------------------------------------

package/src/superlocalmemory/learning/outcome_queue.py

@@ -207,11 +207,19 @@ def _worker_loop(memory_db_path: Path, interval_s: float) -> None:
     )
     import time as _time
     next_reap = _time.monotonic() + _REAP_INTERVAL_S
-    while not _stop_event.wait(interval_s):
+    # Adaptive idle back-off: poll at interval_s under load, but relax the wait
+    # (doubling, capped) when the queue drains empty so an idle daemon stops
+    # contending on the shared SQLite file every 0.25s (issue #53). Snaps back
+    # to interval_s the instant there is work again.
+    _idle_cap = max(interval_s, 2.0)
+    cur_wait = interval_s
+    while not _stop_event.wait(cur_wait):
         try:
-            _drain_once(memory_db_path)
+            drained = _drain_once(memory_db_path)
         except Exception as exc:  # pragma: no cover — defensive
             logger.warning("outcome_queue drain crashed: %s", exc)
+            drained = 0
+        cur_wait = interval_s if drained else min(cur_wait * 2.0, _idle_cap)
         # Periodic reaper for CLI/dashboard outcomes that no Stop hook
         # will ever finalize. Runs OFF the drain path so a busy queue
         # doesn't starve the reaper.

package/src/superlocalmemory/llm/backbone.py

@@ -295,7 +295,14 @@ class LLMBackbone:
         }
         if system:
             payload["system"] = system
-        return _ANTHROPIC_URL, headers, payload
+        # Respect custom base_url (e.g. Anthropic-compatible proxy).
+        # Append /v1/messages to the root URL, mirroring how _build_openai
+        # handles api_base. Falls back to the official Anthropic endpoint.
+        url = (
+            self._base_url.rstrip("/") + "/v1/messages"
+            if self._base_url else _ANTHROPIC_URL
+        )
+        return url, headers, payload
 
     def _build_azure(
         self, prompt: str, system: str, max_tokens: int, temperature: float,

package/src/superlocalmemory/retrieval/reranker.py

@@ -197,6 +197,11 @@ class CrossEncoderReranker:
                 "TOKENIZERS_PARALLELISM": "false",
                 "TORCH_DEVICE": "cpu",
                 "ORT_DISABLE_COREML": "1",
+                # Restore parallel OpenMP. The package caps OMP_NUM_THREADS
+                # globally to avoid a torch+lightgbm libomp SIGSEGV in the
+                # main process. This worker loads torch but never lightgbm,
+                # so there is no collision risk and full parallelism is safe.
+                "OMP_NUM_THREADS": str(os.cpu_count() or 4),
             }
             from superlocalmemory.core.platform_utils import popen_platform_kwargs
             self._worker_proc = subprocess.Popen(

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

@@ -325,6 +325,7 @@ async def record_feedback(data: dict):
 
         row_id = feedback.record_dashboard_feedback(
             memory_id=str(memory_id), query=query, feedback_type=feedback_type,
+            profile_id=get_active_profile() or "default",
         )
 
         return {
@@ -369,6 +370,7 @@ async def record_dwell(data: dict):
 
         row_id = feedback.record_dashboard_feedback(
             memory_id=str(memory_id), query=query, feedback_type=feedback_type,
+            profile_id=get_active_profile() or "default",
         )
 
         return {

package/src/superlocalmemory/server/routes/v3_api.py

@@ -441,6 +441,17 @@ def _validate_provider_url(url: str, client_host: str) -> str | None:
         return "Cloud metadata endpoints are not allowed"
     if client_host in ("127.0.0.1", "::1", "localhost"):
         return None  # local dashboard may target its own local/LAN endpoints
+    # SLM_REMOTE residue (#40): an allowlisted LAN dashboard is trusted exactly
+    # like the loopback one and may probe its own LAN LLM endpoint. This does
+    # NOT relax the SSRF guard for arbitrary remote callers —
+    # is_lan_client_allowed is False unless remote mode is ON *and* the client
+    # IP is in SLM_MCP_ALLOWED_HOSTS.
+    try:
+        from superlocalmemory.core.remote_mode import is_lan_client_allowed
+        if is_lan_client_allowed(client_host):
+            return None
+    except Exception:  # pragma: no cover — defensive, never weaken on import error
+        pass
     try:
         ip = ipaddress.ip_address(host)
     except ValueError:

package/src/superlocalmemory/server/unified_daemon.py

@@ -171,6 +171,33 @@ from superlocalmemory.core.recall_gate import (
 # daemon startup via engine._process_pending_memories().
 _engine = None
 
+
+def _emit_event(
+    event_type: str,
+    payload: dict | None = None,
+    *,
+    source_agent: str = "http_client",
+) -> None:
+    """Emit a best-effort EventBus event from an HTTP write path.
+
+    Mirrors mcp.shared.emit_event but tags source_protocol="http" so the
+    dashboard can distinguish HTTP traffic from MCP tool calls. Never raises
+    — a bus failure must not affect the caller's response.
+    """
+    try:
+        from superlocalmemory.infra.event_bus import EventBus
+        from superlocalmemory.server.routes.helpers import DB_PATH
+        bus = EventBus.get_instance(DB_PATH)
+        bus.emit(
+            event_type,
+            payload=payload,
+            source_agent=source_agent,
+            source_protocol="http",
+        )
+    except Exception as exc:
+        logger.debug("EventBus emit failed (%s): %s", event_type, exc)
+
+
 # v3.4.53: Limit concurrent full (non-fast) recalls. Without this, N parallel
 # /recall calls spawn N × 6-channel threads → Ollama serialises, reranker
 # lock queues, and total wall time is N × single-recall-time. 3 concurrent
@@ -240,6 +267,14 @@ class ObserveBuffer:
             self._timer = threading.Timer(self._debounce_sec, self._flush)
             self._timer.daemon = True
             self._timer.start()
+        _emit_event(
+            "memory.observed",
+            payload={
+                "content_hash": content_hash,
+                "content_preview": content[:120],
+                "buffer_size": buf_size,
+            },
+        )
         return {"captured": True, "queued": True, "buffer_size": buf_size}
 
     def _flush(self) -> None:
@@ -268,6 +303,22 @@ class ObserveBuffer:
                         # The prior 'processed N' counted skipped (capture=False)
                         # items as successes — a false-positive write count.
                         captured_count += 1
+                        _emit_event(
+                            "memory.captured",
+                            payload={
+                                "category": decision.category,
+                                "confidence": getattr(decision, "confidence", None),
+                                "content_preview": content[:120],
+                            },
+                        )
+                    else:
+                        _emit_event(
+                            "memory.dropped",
+                            payload={
+                                "reason": getattr(decision, "reason", "no patterns matched"),
+                                "content_preview": content[:120],
+                            },
+                        )
                 except Exception as exc:
                     failed_count += 1
                     logger.warning(
@@ -1784,6 +1835,15 @@ def _register_daemon_routes(application: FastAPI) -> None:
                     req.content, metadata=metadata,
                     scope=scope, shared_with=shared_with,
                 )
+                _emit_event(
+                    "memory.stored",
+                    payload={
+                        "fact_ids": list(fact_ids) if fact_ids else [],
+                        "count": len(fact_ids) if fact_ids else 0,
+                        "path": "remember_sync",
+                        "content_preview": req.content[:120],
+                    },
+                )
                 return {"ok": True, "fact_ids": fact_ids, "count": len(fact_ids)}
             except Exception as exc:
                 raise HTTPException(500, detail=str(exc))
@@ -1823,6 +1883,14 @@ def _register_daemon_routes(application: FastAPI) -> None:
             pending_id = store_pending(
                 req.content, tags=req.tags or "", metadata=meta,
             )
+            _emit_event(
+                "memory.queued",
+                payload={
+                    "pending_id": pending_id,
+                    "tags": req.tags or "",
+                    "content_preview": req.content[:120],
+                },
+            )
             return {
                 "ok": True,
                 "fact_ids": fact_ids,
@@ -2196,6 +2264,16 @@ def _start_pending_materializer() -> None:
                         )
                         engine.store_fact_direct(fact)
                         mark_done(item["id"])
+                        _emit_event(
+                            "memory.stored",
+                            payload={
+                                "pending_id": item["id"],
+                                "memory_id": mem_id,
+                                "path": "materializer_drain",
+                                "content_preview": content[:120],
+                            },
+                            source_agent="materializer",
+                        )
                     except Exception as exc:
                         logger.warning(
                             "Pending %d failed: %s", item["id"], exc,

package/src/superlocalmemory/storage/database.py

@@ -12,7 +12,7 @@ Part of Qualixar | Author: Varun Pratap Bhardwaj
 """
 from __future__ import annotations
 
-import json, logging, sqlite3, threading, time
+import json, logging, os, sqlite3, threading, time
 from contextlib import contextmanager
 from pathlib import Path
 from types import ModuleType
@@ -27,10 +27,16 @@ from superlocalmemory.storage.models import (
 
 logger = logging.getLogger(__name__)
 
-def _jl(raw: Any, default: Any = None) -> Any:
-    """JSON-load a value, returning *default* on None/empty."""
+_MISSING = object()
+
+def _jl(raw: Any, default: Any = _MISSING) -> Any:
+    """JSON-load a value, returning *default* on None/empty.
+
+    _jl(raw)       -> [] when raw is None/empty  (list fields)
+    _jl(raw, None) -> None when raw is None/empty (optional fields)
+    """
     if raw is None or raw == "":
-        return default if default is not None else []
+        return [] if default is _MISSING else default
     return json.loads(raw)
 
 def _jd(val: Any) -> str | None:
@@ -38,9 +44,30 @@ def _jd(val: Any) -> str | None:
     return json.dumps(val) if val is not None else None
 
 
-_BUSY_TIMEOUT_MS = 10_000   # 10 seconds — wait for other writers
-_MAX_RETRIES = 5            # retry on transient SQLITE_BUSY
-_RETRY_BASE_DELAY = 0.1    # seconds — exponential backoff base
+def _env_int(name: str, default: int) -> int:
+    """Read a positive int from the environment, falling back on bad/absent."""
+    try:
+        val = int(os.environ.get(name, "").strip())
+        return val if val > 0 else default
+    except (ValueError, AttributeError):
+        return default
+
+
+def _env_float(name: str, default: float) -> float:
+    """Read a positive float from the environment, falling back on bad/absent."""
+    try:
+        val = float(os.environ.get(name, "").strip())
+        return val if val > 0 else default
+    except (ValueError, AttributeError):
+        return default
+
+
+# SQLite endurance tuning. Defaults preserve prior hard-coded behaviour exactly;
+# operators on slow/contended I/O can raise them via env (issue #53) without a
+# code change. Unset env => byte-identical to the previous constants.
+_BUSY_TIMEOUT_MS = _env_int("SLM_DB_BUSY_TIMEOUT_MS", 10_000)   # wait for writers
+_MAX_RETRIES = _env_int("SLM_DB_MAX_RETRIES", 5)                # retry on SQLITE_BUSY
+_RETRY_BASE_DELAY = _env_float("SLM_DB_RETRY_BASE_DELAY", 0.1)  # backoff base (s)
 
 
 def _scope_where(

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

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: superlocalmemory
-Version: 3.6.15
+Version: 3.6.17
 Summary: Information-geometric agent memory with mathematical guarantees
 Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
 License: AGPL-3.0-or-later
@@ -96,10 +96,10 @@ Dynamic: license-file
   <img src="https://superlocalmemory.com/assets/logo-mark.png" alt="SuperLocalMemory" width="200"/>
 </p>
 
-<h1 align="center">SuperLocalMemory V3.6.15</h1>
+<h1 align="center">SuperLocalMemory V3.6.17</h1>
 <p align="center"><strong>Cache. Compress. Remember. Three surfaces — proxy, MCP tools, or skill. Every setup covered.</strong><br/>
 <em>To the best of our knowledge, the only zero-cloud agent memory that beats Mem0's zero-LLM score on LoCoMo. Mode A: 74.8% vs Mem0 64.2% — no GPU, no API key, on CPU.</em></p>
-<p align="center"><code>v3.6.15</code> — <strong>Plugin-native. Profile-aware. Distributed-ready.</strong><br/>
+<p align="center"><code>v3.6.17</code> — <strong>Plugin-native. Profile-aware. Distributed-ready.</strong><br/>
 Proxy: <code>slm wrap claude</code> &nbsp;·&nbsp; MCP: add <code>slm_compress</code> to your config &nbsp;·&nbsp; Skill: zero-config</p>
 <p align="center"><strong>3 published research papers</strong> (arXiv preprints + Zenodo-archived) · <a href="https://arxiv.org/abs/2603.02240">arXiv:2603.02240</a> · <a href="https://arxiv.org/abs/2603.14588">arXiv:2603.14588</a> · <a href="https://arxiv.org/abs/2604.04514">arXiv:2604.04514</a></p>
 
@@ -256,7 +256,7 @@ Full docs: [docs/multi-machine.md](docs/multi-machine.md) · [docs/distributed-d
 |:-----|:--------|:-----|
 | **npm** (recommended) | `npm install -g superlocalmemory` | Node 14+, installs Python deps automatically |
 | **pip** | `pip install superlocalmemory` | Python 3.11+, direct install |
-| **Claude Code Plugin** (WP-06) | `/plugin install superlocalmemory@qualixar` | Self-bootstraps venv, isolated SLM_DATA_DIR, additive — 14-tool core |
+| **Claude Code Plugin** (WP-06) | `/plugin marketplace add qualixar/superlocalmemory` then `/plugin install superlocalmemory@qualixar` | Self-bootstraps venv, isolated SLM_DATA_DIR, additive — 14-tool core. Ships the skills/agents/hooks/commands |
 | **Portable / IDE connect** (WP-08) | `slm connect <ide> [--here]` | Wire any IDE without reinstalling; `slm connect claude-code` → plugin pointer |
 
 After any install path: `slm setup` → `slm doctor` → `slm warmup` (optional, pre-downloads ~500MB embedding model).
@@ -312,18 +312,31 @@ Per-IDE configs available for Claude Code, Cursor, Windsurf, VS Code Copilot, Co
 
 ## Claude Code Plugin
 
-Install directly in Claude Code without a system-level npm/pip install:
+Install directly in Claude Code — no system-level npm/pip needed. This is how you
+get the **skills, agents, hooks, commands, and rules** (the MCP server is
+bootstrapped automatically). It is a two-step flow — add the marketplace once,
+then install:
 
 ```bash
+# 1. Add the Qualixar marketplace (one-time — the repo IS the marketplace)
+/plugin marketplace add qualixar/superlocalmemory
+
+# 2. Install the plugin
 /plugin install superlocalmemory@qualixar
 ```
 
 - Self-bootstraps a Python venv, installs all deps in an isolated `SLM_DATA_DIR`
-- Registers 14-tool core MCP surface (`core14` profile by default)
+- Registers the 14-tool core MCP surface (`core14` profile by default)
+- Ships the SLM skills / agents / hooks / commands / rules
 - Additive — does not replace an existing SLM install
 - `slm connect claude-code` detects an existing plugin install and links them
 
-See [docs/getting-started.md](docs/getting-started.md) for full plugin walkthrough.
+> **Plugin vs `pip`/`npm`:** `pip install superlocalmemory` / `npm i -g superlocalmemory`
+> give you the `slm` CLI + the MCP server (the *tools*). The **skills/agents/hooks/
+> commands** come only through the plugin above. Use the plugin for Claude Code; use
+> pip/npm for the CLI or other IDEs.
+
+To update later: `/plugin marketplace update qualixar` then `/plugin install superlocalmemory@qualixar`.
 
 ---
 
@@ -388,6 +401,8 @@ slm dashboard    # Opens at http://localhost:8765
 
 | Version | Codename | Key Features |
 |---|---|---|
+| **v3.6.17** | Community | 8 contributor PRs (observability events, marker-bounded adapter writes, daemon port discovery, anthropic `api_base`, OpenMP workers, atomic-write rehash, `_jl` sentinel, LFS pointer); dashboard-feedback fix (#53/#59); env-tunable SQLite knobs + idle backoff; remote LLM test-probe (#40) |
+| **v3.6.16** | Docs | Corrected Claude Code plugin install — adds the required `/plugin marketplace add` step; clarifies plugin vs pip/npm delivery |
 | **v3.6.15** | Multi-scope | **Opt-in [shared memory](docs/shared-memory.md)** (personal/shared/global, off by default), default-deny scope at every read path, recall scope-race fix, contributor PRs #42/#43/#44, fixes #46–#49 |
 | **v3.6.14** | Plugin-native | Claude Code Plugin (WP-06), MCP profiles (WP-01), IDE connect (WP-08), asset consolidation, UI polish (WP-12) |
 | **v3.6.x** | Optimize Everywhere / Distributed-ready | Three surfaces (proxy/MCP/skill), `SLM_REMOTE=1` LAN mode, remote dashboard, custom LLM endpoints |

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

@@ -177,6 +177,7 @@ src/superlocalmemory/hooks/cursor_adapter.py
 src/superlocalmemory/hooks/hook_daemon.py
 src/superlocalmemory/hooks/hook_handlers.py
 src/superlocalmemory/hooks/ide_connector.py
+src/superlocalmemory/hooks/memory_protocol.py
 src/superlocalmemory/hooks/portable_kit.py
 src/superlocalmemory/hooks/post_tool_async_hook.py
 src/superlocalmemory/hooks/post_tool_outcome_hook.py