superlocalmemory 3.6.9 → 3.6.10

package/CHANGELOG.md

@@ -5,20 +5,88 @@ 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).
 
-## [3.6.9] - 2026-06-11
+## [3.6.10] - 2026-06-14 — Optimize correctness (cache + lossless compression) · MCP per-agent identity · runtime toggles · benchmark + shadow-capture · Issue #38
 
-### Fixed
-- **BUG-A: Health monitor no longer kills the embedding worker under memory pressure.** The watchdog now spares the embedding worker (load-bearing for recall quality) and prefers the reranker instead. The RSS budget defaults to 40% of physical RAM (floor 2500 MB) rather than a hardcoded 2500 MB, preventing thrash on machines with ≥8 GB. `SLM_RSS_BUDGET_MB` env and a new `"health"` config section give full operator control. `HealthConfig` dataclass + `SLMConfig.load()` parsing added so config.json `"health"` keys now actually take effect.
-- **#34: Mesh tools no longer block the daemon event loop.** All 8 async mesh tools wrapped `_ensure_registered` and `_mesh_request` in `asyncio.to_thread` so the blocking loopback HTTP calls leave the event loop (v3.6.7 in-process transport made this a self-deadlock). MCP lifespan hardened so a tool-level exception cannot propagate to uvicorn's shutdown handler. `heartbeat_active` and `registered` now return live state instead of hardcoded literals.
-- **#35: `session_init` now returns a `session_id`.** Clients pass it to `remember()` and `close_session()`. `store_fast` lifts `session_id` from metadata onto the `MemoryRecord` row so facts are correctly attributed. `close_session` queries the DB for the most recent session instead of chasing a phantom `_last_session_id` that was never assigned — `summary_events_created` now returns real counts.
-- **#36-1: HTTP MCP reachable from LAN.** New `SLM_MCP_ALLOWED_HOSTS` (opt-in, default localhost-only) overrides MCP's DNS-rebinding protection. Accepts comma-separated `host:port*` patterns or `*` to disable protection entirely on a trusted LAN.
-- **#36-2: `slm mcp` and the HTTP daemon no longer race for port 8765.** `ensure_daemon` now checks TCP connectivity in addition to PID file + HTTP health, detecting systemd-started daemons mid-startup before attempting a second bind. `SLM_DAEMON_PORT` is now fully wired end-to-end (previously only read in `commands.py` URL building, not in the actual uvicorn bind or `_start_daemon_subprocess`).
-- **#32: Docs correctly state Python 3.11+ requirement** (was "3.10 or later"). Added Ubuntu 22.04 deadsnakes install instructions and a new `docs/install-linux.md` guide.
+This release makes the **Optimize** subsystem (the HTTP proxy that caches and compresses LLM API calls) correct, observable, and **independently controllable at runtime**, adds **per-agent identity** to the HTTP MCP transport, ships a **benchmark + shadow-capture** harness that proves the cache and compression behaviour, and fixes GitHub issue #38. Cache and compression remain **default-OFF** and are now separately toggleable from the dashboard.
+
+### Added — Independent runtime control of cache vs compression
+
+- **Cache and compression are separate switches**, both live at runtime from the dashboard — enable caching only, compression only, both, or neither, without restarting the daemon. The config watchdog rebuilds the proxy hook chain in place on change.
+
+### Added — MCP per-agent identity over HTTP (`/mcp/{agent_id}`)
+
+- **Per-agent attribution without per-agent processes.** The HTTP MCP endpoint accepts an agent-id path segment — `http://127.0.0.1:8765/mcp/claude`, `/mcp/hermes`, `/mcp/gemini`, etc. The daemon extracts it (root_path-aware ASGI wrapper) into a per-request `ContextVar`, so `remember`, `recall`, `observe`, `delete_memory`, `update_memory`, `session_init`, and event emission all tag the correct agent. Many agents share one daemon instead of one `slm mcp` stdio process each. Bare `/mcp/` is unchanged → default `mcp_client` (backward compatible). Precedence: URL path → `SLM_AGENT_ID` env (stdio) → `mcp_client`. URL agent-ids are sanitized (charset-restricted, 64-char cap). See `docs/distributed-deployment.md`.
+
+### Added — Benchmark + shadow-capture (`benchmarks/optimize/`)
+
+- **Correctness benchmark** drives the shipped cache + compression code: exact-cache replay is 100% hit + byte-identical, exact false-hit rate is **0** (the wrong-answer guard), semantic-tier wiring honours its threshold, and safe-mode compression is **lossless** for JSON/code/prose (code forwarded unchanged). Enforced in CI by `tests/optimize/test_benchmark.py`.
+- **Shadow-capture mode** (`SLM_OPTIMIZE_CAPTURE=1`): pure-passthrough proxy records real `{request, response, model, tokens, content_type}` exchanges to `~/.superlocalmemory/optimize_capture.jsonl` (0600, `O_NOFOLLOW`, git-ignored, 1 MB/side cap, secrets never recorded) for replay into the benchmark. No cache/compress while capturing.
+
+### Changed — Compression rebuilt on a research-backed layered design
+
+- **Removed the homegrown extractive JSON/code compressors** (they truncated JSON strings, capped arrays, and stubbed code bodies — lossy and unsafe). Safe mode is now **lossless** (whitespace + compact-JSON normalization only; code untouched). Aggressive mode adds **LLMLingua-2** (`microsoft/llmlingua-2-xlm-roberta-large-meetingbank`) for **prose only** — never code, numbers, structured data, instructions, or the current turn. Dead `compress_json`/`compress_code` toggles removed.
+
+### Fixed — Cache correctness & observability
+
+- **Semantic cache tier is now wired** into the proxy `get/check/set` path (was dead code); default-off, conservative 0.98 return threshold, benchmarked false-hit guard. See `docs/optimize-config.md` for threshold-tuning guidance.
+- **Token accounting corrected** — input tokens (the biggest save) and real output tokens are now counted on cache hits instead of 0/estimates.
+- **Cache survives machine-id changes** — the AES-GCM key/salt is persisted (0600), fixing the total-miss after VM clone / container / OS reinstall.
+- **UI opt-in is real** — a `proxy_enabled` toggle now actually mounts the proxy.
+
+### Fixed — Issue #38 (frontend)
+
+- **Brain page no longer stuck on "Couldn't load Brain"** on a healthy backend — pane activation fires `fetchBrain` and reads the install token from the same source as the working endpoints.
+- **"Test Connection" no longer 401s on an empty API key** in no-auth (Mode B) — when the key is empty no `Authorization` header is sent.
+
+### Security (v3.6.10 audit, Stage 8/9)
+
+- Capture file: `O_NOFOLLOW` + unconditional `0600` (symlink-append + TOCTOU closed); in-memory stream accumulator bounded (CWE-400); synchronous capture writes offloaded off the event loop; `extract_usage` uses an explicit provider allowlist (no silent mis-parse). URL agent-ids sanitized at the single extraction chokepoint.
+
+## [3.6.9] - 2026-06-11 — Full 7-layer recall quality · event-loop safety · health watchdog
+
+This release fixes seven GitHub issues, one critical production incident, five post-implementation audit bugs, and two event-loop deadlocks introduced in v3.6.7. All 7 retrieval layers (semantic, BM25, temporal, spreading activation, Hopfield, entity graph, cross-encoder reranker) now operate at full designed quality with no capability tradeoffs.
+
+### Fixed — Production incident
+
+- **BUG-A (CRITICAL): Health monitor RSS watchdog no longer kills the embedding worker.** On machines with <16 GB RAM the hardcoded 2500 MB budget caused the watchdog to kill the ~1 GB ONNX embedding worker during recall bursts. Semantic channel silently scored 0.0, recall fell back to keyword-only FTS5/BM25 ("DEGRADED MODE"), and the v3.6.8 health monitor reported false-healthy because it checked worker liveness rather than semantic signal quality. Fixed by: (1) defaulting the budget to 40% of physical RAM (floor 2500 MB) with `SLM_RSS_BUDGET_MB` env override; (2) protecting the embedder from the kill list — the reranker is targeted first as it can be recreated without losing recall quality; (3) adding `HealthConfig` dataclass and `SLMConfig.load()` parsing so `config.json` `"health"` keys now actually take effect.
+
+### Fixed — GitHub issues
+
+- **#34: Mesh tools no longer deadlock the daemon.** All 8 async mesh MCP tools called blocking `urllib.urlopen` loopback HTTP against the daemon's own `/mcp` endpoint introduced in v3.6.7. From inside the event loop this self-deadlocked uvicorn's single-thread executor, causing a graceful-shutdown timeout that killed the daemon. All 8 call sites wrapped in `asyncio.to_thread`. MCP lifespan hardened so a tool-level exception cannot propagate to uvicorn's shutdown handler. `heartbeat_active` and `registered` return live state instead of hardcoded literals.
+- **#35: `session_init` now returns a `session_id`.** Three spots fixed: (1) `tools_active.py` — `session_init` now passes `session_id` through to callers; (2) `engine.py` `store_fast` — lifts `session_id` from metadata onto the `MemoryRecord` row so facts are correctly attributed to sessions; (3) `close_session` — queries the DB for the most recent session instead of chasing a phantom `_last_session_id` that was never assigned, and `summary_events_created` now returns real counts.
+- **#36-1: HTTP MCP now reachable from LAN hosts.** The new `SLM_MCP_ALLOWED_HOSTS` env (opt-in, default localhost-only) overrides FastMCP's DNS-rebinding protection that rejected non-`127.0.0.1` `Host` headers on LAN deployments. Accepts comma-separated `host:port*` patterns or `*`. Security default is unchanged — the endpoint stays localhost-only without explicit opt-in.
+- **#36-2: `slm mcp` and the HTTP daemon no longer race for port 8765.** `ensure_daemon` now checks TCP connectivity in addition to PID file + HTTP health, detecting systemd-started daemons mid-startup before attempting a second bind. `SLM_DAEMON_PORT` is now fully wired end-to-end (previously only read in `commands.py` URL construction, ignored in the actual uvicorn bind and `_start_daemon_subprocess`). Wrapped in `try/except ValueError` so a malformed env value fails with a clear message rather than a silent zero-port bind.
+- **#33 + #37: Environment variable reference table published.** All ~90 `SLM_*` variables documented in `docs/distributed-deployment.md` with type, default, scope, and example. Closes both issues.
+- **#32: Python version requirement corrected.** Docs previously stated "3.10 or later" — the codebase enforces 3.11+. Updated `getting-started.md`, added Ubuntu 22.04 deadsnakes install path and a new `docs/install-linux.md`.
+
+### Fixed — Event-loop blocking (post-v3.6.7 audit)
+
+The v3.6.7 in-process HTTP MCP transport changed the execution context of all MCP tools from stdio subprocess threads to async event-loop coroutines. Three core tools made blocking HTTP or I/O calls that safe in a subprocess become event-loop deadlocks in the new context.
+
+- **`session_init` no longer blocks the event loop.** `pool_recall` called `DaemonPoolProxy.recall` which uses blocking `urllib.urlopen` internally. Wrapped in `asyncio.to_thread`.
+- **`observe` no longer blocks the event loop.** `auto.capture()` is a synchronous function that makes an internal HTTP call. Wrapped in `asyncio.to_thread`.
+- **`remember` no longer blocks the event loop.** Both `is_daemon_running()` (file-system check + HTTP probe) and `daemon_request()` (blocking HTTP POST) wrapped in `asyncio.to_thread`.
+
+### Fixed — Post-implementation audit (5 bugs)
+
+- **`close_session` crash eliminated.** `tools_active.py` `close_session` called `engine._db._get_conn()`, a private method removed in a prior refactor. Changed to the public `engine._db.execute()` API.
+- **`TransportSecuritySettings` import no longer fails on older MCP SDK versions.** The import was at module level in `unified_daemon.py`, causing an `ImportError` on MCP SDK < 1.27. Moved inside the `if _mcp_allowed:` conditional so it only loads when the HTTP transport is actually being activated.
+- **`SLM_DAEMON_PORT` no longer silently becomes 0 on invalid input.** `int(os.environ.get("SLM_DAEMON_PORT", "") or 8765)` evaluated to `int("")` → `ValueError` → port 0. Wrapped in `try/except ValueError` with a fallback to 8765 and a log warning.
+- **Health monitor cmdline truncation increased.** `cmdline[:80]` silently dropped long Python command lines (common with virtualenv paths), producing false "not SLM" negatives. Increased to `cmdline[:200]`.
+- **Daemon `_DEFAULT_PORT` no longer crashes on malformed env.** Same `ValueError` guard added to the module-level `_DEFAULT_PORT` assignment in `cli/daemon.py`.
+
+### Changed — Recall performance (zero quality tradeoff)
+
+- **SpreadingActivation: 418ms → 36ms (12×).** Neighbor lookups for each graph node are now cached across propagation iterations. The neighbor list is static within a single recall — re-querying SQL per iteration (up to ~120 queries per propagation) was pure waste. The cache is local to the call, so correctness is unchanged.
+- **`fast=True` deprecated.** `fast=True` was added in v3.4.40 when SpreadingActivation took 418ms. With SA now completing in ~36ms, `fast=True` (which disables SA) is actually *slower* than `fast=False` and reduces recall quality by dropping a full retrieval channel. `MemoryEngine.recall()` now logs a `WARNING` and silently treats `fast=True` as `False`. The parameter is retained for API backward compatibility and will be removed in v3.7.x. All built-in callers already passed `fast=False`.
+- Hopfield `prefilter_candidates` and entity-graph scoring candidates retain their original designed values (1000 and 100 respectively) — no quality tradeoffs were made in pursuit of latency targets.
 
 ### Added
+
 - `HealthConfig` dataclass in `core/config.py` and `"health"` section parsing in `SLMConfig.load()`.
-- `docs/distributed-deployment.md` — complete guide for LXC/container/multi-machine setups, including a full ~90-entry `SLM_*` environment variable reference table (closes #33 + #37).
-- `docs/install-linux.md` — Ubuntu 22.04 / Debian install guide with venv, pipx, and pyenv options, plus a systemd unit template.
+- `SLM_RSS_BUDGET_MB` environment variable for operator control of the health watchdog kill threshold.
+- `docs/distributed-deployment.md` — complete guide for LXC/container/multi-machine setups with full `SLM_*` environment variable reference (closes #33 + #37).
+- `docs/install-linux.md` — Ubuntu 22.04 / Debian install guide with venv, pipx, and pyenv paths plus a systemd unit template.
 
 ## [3.6.8] - 2026-06-11 — Runtime recall-health monitor (self-healing recall)
 

package/README.md

@@ -33,18 +33,22 @@
 <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
@@ -60,7 +64,7 @@ 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/package.json

@@ -1,6 +1,6 @@
 {
   "name": "superlocalmemory",
-  "version": "3.6.9",
+  "version": "3.6.10",
   "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/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "superlocalmemory"
-version = "3.6.9"
+version = "3.6.10"
 description = "Information-geometric agent memory with mathematical guarantees"
 readme = "README.md"
 license = {text = "AGPL-3.0-or-later"}
@@ -69,6 +69,11 @@ dependencies = [
     "scikit-learn==1.8.0",
     # Vector KNN extension for the semantic channel.
     "sqlite-vec==0.1.9",
+    # v3.6.10: LLMLingua-2 prose compression (aggressive mode, opt-in at runtime).
+    # Hard dependency so `pip install superlocalmemory` ships compression-ready;
+    # the ~560MB model downloads on first setup/warmup (fail-open). Verified the
+    # pin does NOT disturb the transformers/torch/numpy pins above.
+    "llmlingua==0.2.2",
 ]
 
 [project.optional-dependencies]

package/src/superlocalmemory/__init__.py

@@ -32,7 +32,7 @@ if "OMP_NUM_THREADS" not in os.environ:
     os.environ["OMP_NUM_THREADS"] = "2"
 # ---------------------------------------------------------------------------
 
-__version__ = "3.6.9"
+__version__ = "3.6.10"
 
 _REQUIRED_VERSIONS = {
     "sentence_transformers": "5.3.0",

package/src/superlocalmemory/cli/compress_cmd.py

@@ -2,7 +2,7 @@
 # Licensed under AGPL-3.0-or-later - see LICENSE file
 # Part of SuperLocalMemory V3 | https://qualixar.com | https://varunpratap.com
 
-"""Handlers for ``slm compress status|mode|code|prose|ccr``."""
+"""Handlers for ``slm compress status|mode|prose``."""
 
 from __future__ import annotations
 
@@ -19,11 +19,6 @@ def _get_store():
     return ConfigStore()
 
 
-def _get_cache_db():
-    from superlocalmemory.optimize.storage.db import CacheDB
-    return CacheDB()
-
-
 def _write_config(**fields) -> None:
     """5-step immutable config-write."""
     store = _get_store()
@@ -44,18 +39,31 @@ def cmd_compress(args: Namespace) -> None:
     sub = getattr(args, "compress_command", None)
     _dispatch = {
         "status": cmd_compress_status,
-        "mode": cmd_compress_mode,
-        "code": cmd_compress_code,
-        "prose": cmd_compress_prose,
-        "ccr": cmd_compress_ccr,
-        "align": cmd_compress_align,
+        "mode":   cmd_compress_mode,
+        "prose":  cmd_compress_prose,
+        # removed in v3.6.10: code, ccr, align (extractive compressors removed)
+        "code":   _cmd_compress_removed("code"),
+        "ccr":    _cmd_compress_removed("ccr"),
+        "align":  _cmd_compress_removed("align"),
     }
     handler = _dispatch.get(sub or "")
     if handler:
         handler(args)
     else:
-        print("Usage: slm compress status|mode|code|prose|ccr [options]")
+        print("Usage: slm compress status|mode|prose [options]")
+        sys.exit(0)
+
+
+def _cmd_compress_removed(name: str):
+    def handler(args: Namespace) -> None:
+        print(
+            f"slm compress {name}: removed in SLM v3.6.10. "
+            f"Extractive {name} compression has been replaced by "
+            f"Layer 1 (lossless whitespace) + Layer 2 (LLMLingua-2 prose). "
+            f"Use 'slm compress prose on' to enable prose compression."
+        )
         sys.exit(0)
+    return handler
 
 
 def cmd_compress_status(args: Namespace) -> None:
@@ -68,21 +76,19 @@ def cmd_compress_status(args: Namespace) -> None:
             "status": "ok",
             "compress_enabled": cfg.compress_enabled,
             "compress_mode": cfg.compress_mode,
-            "compress_code": cfg.compress_code,
             "compress_prose": cfg.compress_prose,
-            "compress_ccr": cfg.compress_ccr,
+            "compress_protect_recent": cfg.compress_protect_recent,
         }
         print(json.dumps(data, indent=2))
         return
 
     print("Compression status:")
-    print(f"  Enabled:  {'yes' if cfg.compress_enabled else 'no'}")
-    print(f"  Mode:     {cfg.compress_mode}")
-    print(f"  Code:     {'ON' if cfg.compress_code else 'OFF'}"
-          "  (extractive JSON/code — lossless structure)")
-    print(f"  Prose:    {'ON' if cfg.compress_prose else 'OFF'}")
-    print(f"  CCR:      {'ON' if cfg.compress_ccr else 'OFF'}"
-          " (reversible context retrieval)")
+    print(f"  Enabled:         {'yes' if cfg.compress_enabled else 'no'}")
+    print(f"  Mode:            {cfg.compress_mode}")
+    print(f"  Prose (Layer 2): {'ON' if cfg.compress_prose else 'OFF'}"
+          "  (LLMLingua-2, aggressive mode only)")
+    print(f"  Protect recent:  {cfg.compress_protect_recent} user turns")
+    print("  Layer 1 (lossless whitespace normalization) is always ON when enabled.")
 
 
 def cmd_compress_mode(args: Namespace) -> None:
@@ -103,23 +109,8 @@ def cmd_compress_mode(args: Namespace) -> None:
     print("Daemon hot-reload: active within 2s. No restart required.")
 
 
-def cmd_compress_code(args: Namespace) -> None:
-    """Enable or disable code/JSON compression."""
-    use_json = getattr(args, "json", False)
-    value = getattr(args, "code_value", "on")
-
-    _write_config(compress_code=(value == "on"))
-
-    if use_json:
-        print(json.dumps({"status": "ok", "compress_code": value == "on"}))
-        return
-
-    print(f"Code compression: {'ENABLED' if value == 'on' else 'DISABLED'}.")
-    print("Daemon hot-reload: active within 2s.")
-
-
 def cmd_compress_prose(args: Namespace) -> None:
-    """Enable or disable prose compression."""
+    """Enable or disable prose compression (Layer 2, LLMLingua-2)."""
     use_json = getattr(args, "json", False)
     value = getattr(args, "prose_value", "off")
 
@@ -141,39 +132,10 @@ def cmd_compress_prose(args: Namespace) -> None:
         print(json.dumps({"status": "ok", "compress_prose": value == "on"}))
         return
 
-    print(f"Prose compression: {'ENABLED' if value == 'on' else 'DISABLED'}.")
+    print(f"Prose compression (LLMLingua-2): {'ENABLED' if value == 'on' else 'DISABLED'}.")
+    if value == "on":
+        print("  Requires: compress_mode=aggressive and llmlingua package installed.")
+        print("  Run: slm compress mode aggressive  (if not already set)")
     if value == "on" and "compress_enabled" in fields:
         print("  (also enabled global compress)")
     print("Daemon hot-reload: active within 2s.")
-
-
-def cmd_compress_align(args: Namespace) -> None:
-    """Enable or disable alignment compression."""
-    use_json = getattr(args, "json", False)
-    value = getattr(args, "align_value", "on")
-
-    _write_config(compress_align=(value == "on"))
-
-    if use_json:
-        print(json.dumps({"status": "ok", "compress_align": value == "on"}))
-        return
-
-    print(f"Alignment compression: {'ENABLED' if value == 'on' else 'DISABLED'}.")
-    print("Daemon hot-reload: active within 2s.")
-
-
-def cmd_compress_ccr(args: Namespace) -> None:
-    """Enable or disable CCR (Compressed Context Retrieval)."""
-    use_json = getattr(args, "json", False)
-    value = getattr(args, "ccr_value", "off")
-
-    _write_config(compress_ccr=(value == "on"))
-
-    if use_json:
-        print(json.dumps({"status": "ok", "compress_ccr": value == "on"}))
-        return
-
-    print(f"CCR (Compressed Context Retrieval): {'ENABLED' if value == 'on' else 'DISABLED'}.")
-    if value == "on":
-        print("Originals stored in llmcache.db for reversible retrieval.")
-    print("Daemon hot-reload: active within 2s.")

package/src/superlocalmemory/cli/optimize_cmd.py

@@ -97,9 +97,7 @@ def cmd_optimize_status(args: Namespace) -> None:
           f" semantic: {'OFF' if not cfg.semantic_enabled else f'{cfg.ttl.semantic_seconds}s'})")
     print(f"  Compress:  {'enabled' if cfg.compress_enabled else 'disabled'}"
           f"  (mode: {cfg.compress_mode},"
-          f" code: {'ON' if cfg.compress_code else 'OFF'},"
-          f" prose: {'ON' if cfg.compress_prose else 'OFF'},"
-          f" CCR: {'ON' if cfg.compress_ccr else 'OFF'})")
+          f" prose/L2: {'ON' if cfg.compress_prose else 'OFF'})")
     proxy_status = f"running on :{OPTIMIZE_DEFAULT_PORT}" if proxy_running else "not running"
     print(f"  Proxy:     {proxy_status}")
     print(f"  Config:    ~/.superlocalmemory/optimize.json  (version {cfg.config_version})")

package/src/superlocalmemory/cli/setup_wizard.py

@@ -32,6 +32,8 @@ _SLM_HOME = Path(os.environ.get("SL_MEMORY_PATH", Path.home() / ".superlocalmemo
 _SETUP_MARKER = _SLM_HOME / ".setup-complete"
 _EMBED_MODEL = "nomic-ai/nomic-embed-text-v1.5"
 _RERANKER_MODEL = "cross-encoder/ms-marco-MiniLM-L-12-v2"
+# v3.6.10: compulsory LLMLingua-2 prose compression model (~560MB, aggressive mode).
+_COMPRESSOR_MODEL = "microsoft/llmlingua-2-xlm-roberta-large-meetingbank"
 
 
 # ---------------------------------------------------------------------------
@@ -178,6 +180,49 @@ def _download_reranker(model_name: str) -> bool:
         return False
 
 
+def _download_compressor(model_name: str) -> bool:
+    """Download the LLMLingua-2 prose compression model (v3.6.10).
+
+    Mirrors _download_reranker: a subprocess forces the HF download with visible
+    progress. Fail-open — a network hiccup must NOT break setup; the model also
+    lazy-downloads on first use in prose_llmlingua.py.
+    """
+    print(f"\n  Downloading compression model: {model_name}")
+    print(f"  (LLMLingua-2 prose compressor, ~560MB — aggressive mode only)\n")
+
+    script = (
+        "from llmlingua import PromptCompressor; "
+        f"PromptCompressor(model_name='{model_name}', use_llmlingua2=True, "
+        "device_map='cpu'); "
+        "print('OK')"
+    )
+
+    try:
+        result = subprocess.run(
+            [sys.executable, "-c", script],
+            timeout=900,  # 560MB on a slow link can exceed 5 min
+            capture_output=False,
+            text=True,
+            env={
+                **os.environ,
+                "CUDA_VISIBLE_DEVICES": "",
+                "TOKENIZERS_PARALLELISM": "false",
+                "TORCH_DEVICE": "cpu",
+            },
+        )
+        if result.returncode == 0:
+            print(f"  ✓ Compression model ready")
+            return True
+        print(f"  ✗ Compression model download failed (will lazy-download on first use)")
+        return False
+    except ImportError:
+        print(f"  ⚠ llmlingua not installed — compression model will download on first use")
+        return False
+    except Exception as exc:
+        print(f"  ✗ Compression model error: {exc}")
+        return False
+
+
 # ---------------------------------------------------------------------------
 # Verification
 # ---------------------------------------------------------------------------
@@ -393,6 +438,10 @@ def run_wizard(auto: bool = False) -> None:
     else:
         _download_reranker(_RERANKER_MODEL)
 
+    print()
+    print("─── Step 4c/10: Download Compression Model (LLMLingua-2) ───")
+    _download_compressor(_COMPRESSOR_MODEL)
+
     # -- Step 5: Daemon Configuration (v3.4.3) --
     print()
     print("─── Step 5/10: Daemon Configuration ───")

package/src/superlocalmemory/mcp/agent_context.py

@@ -0,0 +1,111 @@
+"""Per-HTTP-request agent ID resolution — ContextVar home.
+
+Kept in a standalone module so both tools_core and tools_active can import
+it without creating circular dependencies (server.py → tools_core → here,
+and unified_daemon.py → here independently).
+
+Priority chain (HTTP-first, stdio-fallback):
+  1. ContextVar set by _AgentIDExtractorASGI middleware from /mcp/{agent_id} URL path.
+  2. SLM_AGENT_ID environment variable (stdio transport legacy).
+  3. Hard-coded "mcp_client" sentinel.
+"""
+from __future__ import annotations
+
+import contextvars
+import os
+import re
+
+_current_agent_id: contextvars.ContextVar[str] = contextvars.ContextVar(
+    "slm_agent_id", default="mcp_client"
+)
+
+# Agent ids arrive from an untrusted URL path segment. They are ATTRIBUTION
+# metadata, never an authenticated principal — but they reach loggers, the
+# agent registry, and SQL-bound attribution columns, so we hard-restrict the
+# charset at the single extraction chokepoint. This neutralises log-injection
+# (CRLF / ANSI), oversized ids, and any path-ish characters in one place.
+_AGENT_ID_SANITIZE = re.compile(r"[^A-Za-z0-9._-]")
+_AGENT_ID_MAX_LEN = 64
+
+
+def sanitize_agent_id(raw: str) -> str:
+    """Coerce an untrusted agent-id segment to a safe, bounded token."""
+    return _AGENT_ID_SANITIZE.sub("_", raw)[:_AGENT_ID_MAX_LEN]
+
+
+def get_current_agent_id(env_fallback: bool = True) -> str:
+    """Return the agent_id for the current asyncio task.
+
+    For HTTP transport the ASGI wrapper sets the ContextVar from the URL path
+    before the request reaches any MCP tool, so this returns the URL-derived id.
+    For stdio transport the ContextVar holds its default ("mcp_client") and we
+    fall through to the SLM_AGENT_ID env var instead.
+    """
+    ctx_id = _current_agent_id.get()
+    if ctx_id != "mcp_client":
+        return ctx_id
+    if env_fallback:
+        return os.environ.get("SLM_AGENT_ID", "mcp_client")
+    return "mcp_client"
+
+
+class AgentIDExtractorASGI:
+    """ASGI wrapper that maps ``/mcp/{agent_id}`` → the agent-id ContextVar.
+
+    Mounted at ``/mcp`` in unified_daemon. IMPORTANT: Starlette's ``Mount``
+    (≥0.35 / 1.x) does NOT strip the mount prefix from ``scope["path"]`` — it
+    records the prefix in ``scope["root_path"]`` and leaves ``path`` as the full
+    request path (e.g. ``/mcp/claude`` with ``root_path == "/mcp"``). So we
+    compute the mount-relative sub-path ourselves as ``path[len(root_path):]``.
+
+    Flow for ``POST /mcp/claude``:
+      sub-path ``/claude`` → agent id ``claude`` → set ContextVar → rewrite the
+      scope path to ``{root_path}/`` so the inner FastMCP app (Starlette, route
+      ``/``) sees the same mount-relative ``/`` it sees for a bare ``/mcp/``.
+
+    Backward compatible: bare ``/mcp/`` has sub-path ``/`` → no agent segment →
+    the request passes through untouched and the ContextVar keeps its
+    ``"mcp_client"`` default.
+
+    Per-request isolation is guaranteed by ContextVar + ``reset(token)`` in a
+    ``finally``, so concurrent HTTP sessions never see each other's agent id.
+    """
+
+    __slots__ = ("_app",)
+
+    def __init__(self, inner) -> None:
+        self._app = inner
+
+    async def __call__(self, scope, receive, send):
+        if scope.get("type") == "http":
+            root_path: str = scope.get("root_path", "")
+            full_path: str = scope.get("path", "/")
+            # Mount-relative sub-path (what comes AFTER /mcp). When a root_path
+            # is present the request path MUST start with it (Starlette Mount
+            # guarantees this); if it somehow does not, treat it as no-agent and
+            # pass through untouched rather than mis-parsing the full path.
+            if root_path:
+                if not full_path.startswith(root_path):
+                    await self._app(scope, receive, send)
+                    return
+                subpath = full_path[len(root_path):]
+            else:
+                subpath = full_path
+            first = subpath.lstrip("/").split("/")[0]
+            if first:
+                first = sanitize_agent_id(first)
+                token = _current_agent_id.set(first)
+                # Rewrite the path so the inner app sees the bare mount root,
+                # exactly as it would for a no-agent /mcp/ request.
+                new_full = (root_path + "/") if root_path else "/"
+                new_scope = {
+                    **scope,
+                    "path": new_full,
+                    "raw_path": new_full.encode(),
+                }
+                try:
+                    await self._app(new_scope, receive, send)
+                finally:
+                    _current_agent_id.reset(token)
+                return
+        await self._app(scope, receive, send)

package/src/superlocalmemory/mcp/tools_active.py

@@ -109,15 +109,14 @@ def _sqlite_emergency_recall(
 def _get_agent_id(default: str = "mcp_client") -> str:
     """Resolve the calling agent's ID for attribution.
 
-    Each MCP client (Claude Code, Codex, Gemini CLI, Kimi, etc.) can set
-    the ``SLM_AGENT_ID`` env var in its MCP server config so that memories,
-    observations, and registry entries are tagged with the actual source
-    agent — not the legacy ``"mcp_client"`` default.
-
-    v3.4.39+: enables proper per-agent attribution in ``session_init``,
-    ``observe``, and event emissions.
+    Priority chain (v3.6.10+):
+    1. ContextVar set by HTTP URL path (/mcp/{agent_id}) — HTTP transport.
+    2. SLM_AGENT_ID env var — stdio transport per-process identity.
+    3. Provided default (legacy "mcp_client").
     """
-    return os.environ.get("SLM_AGENT_ID", default)
+    from superlocalmemory.mcp.agent_context import get_current_agent_id
+    resolved = get_current_agent_id(env_fallback=True)
+    return resolved if resolved != "mcp_client" else default
 
 
 def _emit_event(event_type: str, payload: dict | None = None,

package/src/superlocalmemory/mcp/tools_core.py

@@ -110,6 +110,10 @@ def register_core_tools(server, get_engine: Callable) -> None:
         Extracts atomic facts, resolves entities, builds graph edges,
         and indexes for 4-channel retrieval.
         """
+        # v3.6.10: resolve "mcp_client" sentinel → URL path (HTTP) or env var (stdio)
+        if agent_id == "mcp_client":
+            from superlocalmemory.mcp.agent_context import get_current_agent_id
+            agent_id = get_current_agent_id()
         meta = {
             "project": project,
             "importance": importance,
@@ -171,6 +175,10 @@ def register_core_tools(server, get_engine: Callable) -> None:
         ``CLAUDE_SESSION_ID``. Omitting it degrades to "no closed-loop
         learning for this recall" — the recall itself always works.
         """
+        # v3.6.10: resolve "mcp_client" sentinel → URL path (HTTP) or env var (stdio)
+        if agent_id == "mcp_client":
+            from superlocalmemory.mcp.agent_context import get_current_agent_id
+            agent_id = get_current_agent_id()
         import asyncio
         try:
             from superlocalmemory.mcp._daemon_proxy import choose_pool
@@ -485,6 +493,10 @@ def register_core_tools(server, get_engine: Callable) -> None:
             fact_id: Exact fact ID to delete (from recall or list_recent results).
             agent_id: Identifier of the calling agent (logged for audit).
         """
+        # v3.6.10: resolve "mcp_client" sentinel → URL path (HTTP) or env var (stdio)
+        if agent_id == "mcp_client":
+            from superlocalmemory.mcp.agent_context import get_current_agent_id
+            agent_id = get_current_agent_id()
         try:
             from superlocalmemory.core.worker_pool import WorkerPool
             pool = WorkerPool.shared()
@@ -519,6 +531,10 @@ def register_core_tools(server, get_engine: Callable) -> None:
             content: New content for the memory (cannot be empty).
             agent_id: Identifier of the calling agent (logged for audit).
         """
+        # v3.6.10: resolve "mcp_client" sentinel → URL path (HTTP) or env var (stdio)
+        if agent_id == "mcp_client":
+            from superlocalmemory.mcp.agent_context import get_current_agent_id
+            agent_id = get_current_agent_id()
         try:
             if not content or not content.strip():
                 return {"success": False, "error": "content cannot be empty"}