superlocalmemory 3.7.8 → 3.8.1

package/.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
       "author": {
         "name": "Qualixar"
       },
-      "description": "Local-first agent memory + reversible context compression and KV cache, as an MCP server. 20-tool code profile with graph intelligence.",
+      "description": "Local-first agent memory + reversible context compression and KV cache, as an MCP server. 21-tool code profile with graph intelligence.",
       "name": "superlocalmemory",
       "source": "./plugin"
     }

package/ATTRIBUTION.md

@@ -74,10 +74,8 @@ The Implementer verified each arXiv ID against arxiv.org before citation.
 | **CacheAttack** (86% response hijack, 90.6% agentic) | arXiv:2601.23088 | Threat model. The 90.6% figure is [UNVERIFIED — body-only, RA-18]; the 86% figure is verified from the abstract. |
 | **SAFE-CACHE** (centroid-based adversarial defense) | Nature Scientific Reports 2026 | [CITATION-NEEDED-ONLINE — exact paper verified, but exact defense figures are body-only.] Defense reduced attack success from 52.77% to 14.27% per the paper. |
 | **ContextCache** (multi-turn context-aware keys) | arXiv:2506.22791 | §3 — context-aware cache keys prevent false reuse across semantically overlapping but conversationally distinct turns. |
-| **LLMLingua-2** (prose compression) | arXiv:2403.12968 (Microsoft Research) | MIT license. Models: `microsoft/llmlingua-2-bert-base-multilingual-cased-meetingbank` (default) and `microsoft/llmlingua-2-xlm-roberta-large-meetingbank`. Off by default, opt-in via `compress_prose=True` + `compress_mode="aggressive"`. **Originals always stored in CCR before lossy compression — reversible via `headroom_retrieve`.** |
+| **LLMLingua-2** (prose compression) | arXiv:2403.12968 (Microsoft Research) | MIT license. Models: `microsoft/llmlingua-2-bert-base-multilingual-cased-meetingbank` (default) and `microsoft/llmlingua-2-xlm-roberta-large-meetingbank`. Off by default, opt-in via `compress_prose=True` + `compress_mode="aggressive"`. **Originals always stored in CCR before lossy compression — reversible via `slm_ccr_retrieve`.** |
 | **LongLLMLingua** (RAG compression) | arXiv:2310.06839 | Not used in Phase 3. Documented for Phase 4 RAG integration. |
-| **Headroom** (router, JSON handler, code handler, aligner) | github.com/qualixar/headroom (Apache-2.0) | Patterns adapted with attribution: `ContentRouter`, `JSONStructureHandler`, `CodeLanguage` enum, `CacheAligner._classify_token`. |
-| **omnicache-ai** (test fixture patterns) | github.com/qualixar/omnicache-ai (Apache-2.0) | Deterministic hash-to-vector embedding fixture for tests. |
 
 **Two fabricated arXiv IDs were caught and fixed during the LLD-10 audit:**
 `2501.05064` and `2404.12693` (both previously wrong vCache labels). The

package/CHANGELOG.md

@@ -5,6 +5,135 @@ 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).
 
+## [3.8.1] - 2026-07-23 — Existing-install stability
+
+### Fixed
+
+- **Large existing databases no longer delay daemon readiness for a full tier
+  rebalance.** Startup now initializes bounded backend state and leaves the
+  scheduled full-database tier evaluation to maintenance.
+- **Failed enrichment can no longer retry forever.** Automatic durable-ingestion
+  materialization stops after ten attempts; an operator can still request an
+  explicit retry after correcting the underlying issue.
+- **Background enrichment no longer monopolizes SQLite.** Durable
+  materialization now uses short saga checkpoints instead of holding one
+  database write transaction across model extraction and embedding work, so
+  remember, update, delete, and dashboard actions remain writable while
+  enrichment continues.
+- **Durable enrichment retries are idempotent after partial failure.** Fact,
+  evidence, provenance, graph, temporal, and entity-association effects are
+  checkpointed or keyed so retries and process recovery cannot inflate them.
+- **Entity counts stay O(1) on mature databases.** A normalized, indexed
+  fact/entity association ledger replaces per-entity scans of the full facts
+  table; existing associations are backfilled once during upgrade.
+- **Newly queryable memories are immediately visible to recall.** A strong
+  exact BM25/FTS hit now retains one bounded result slot even when semantic
+  candidates would otherwise fill a small `limit`, preserving the
+  queryable-before-enrichment contract.
+- **Mesh heartbeat writes no longer block the async HTTP event loop.** SQLite
+  work runs in FastAPI's worker thread pool, keeping health, token, dashboard,
+  recall, and remember endpoints responsive during peer activity.
+- **Dashboard navigation preserves mounted pane state.** Brain, Knowledge Graph,
+  Memories, and Entity Explorer do not remount and refetch on every return.
+  Successful writes and configuration changes invalidate pane state once so
+  the next visit refreshes deliberately.
+- **Brain telemetry is historical and provenance-backed.** Existing reward
+  outcomes are repaired into source-quality observations in resumable bounded
+  batches, and transient database errors can no longer be reported as a
+  completed repair.
+- **Company-mode learning controls enforce workspace permissions.** Learning
+  and behavioral reads and mutations now apply the same read, write, delete,
+  and manage gates as the rest of the control plane.
+- **Large graph views have a fixed browser-compute budget.** The dashboard can
+  render the requested graph result while force simulation is bounded, with a
+  short initial settle and finite animation window; selecting a large graph can
+  no longer monopolize the browser main thread.
+- **Interactive model workers stay warm instead of repeatedly cold-starting.**
+  Embedding and reranker workers now remain resident for a 30-minute working
+  session by default, and the embedding peak guard matches the daemon's
+  2.5 GB per-worker watchdog. Low-RAM installations can retain shorter idle
+  windows and stricter recycling through the existing environment overrides.
+- **CLI mutations use the resident daemon as the single database writer.**
+  Exact delete and update commands no longer initialize a competing engine
+  beside a running daemon, preventing lock failures on active existing
+  installations.
+- **Skill Evolution repairs incomplete historical schemas.** The migration
+  runner verifies a completed migration's promised end-state before skipping
+  it and reapplies additive DDL when an older upgrade left a required table
+  missing.
+- **Release packages are deterministic again.** Canonical data-root routing,
+  generated plugin parity checks, truthful integration instructions, and the
+  bounded npm package surface are restored.
+
+## [3.8.0] - 2026-07-23 — Teams, bounded loops, and nine framework adapters
+
+### Added
+
+- **Team & company memory with real access control.** A workspace (profile) can
+  now have multiple users, each with a role — **admin**, **member**, or
+  **viewer**. Admins manage people and settings, members read and write, viewers
+  read only. Everything is managed from the dashboard's new **Team & access**
+  panel: add a person, set their role, sign in, or remove them — no config files
+  or command line needed.
+- **Sign-in for shared workspaces.** Turn on *"require sign-in"* (company mode)
+  and everyone must log in before reading or writing memory, so every action is
+  attributable to a person. Single-user setups are unchanged and need no login.
+- **Private, shared, and global memories across a mesh.** Memories keep their
+  personal / shared / global visibility consistently when several SuperLocalMemory
+  instances are connected, and one team's coordination never bleeds into another's.
+- **Optional personal-data scrubbing on save.** Enable PII redaction and emails,
+  phone numbers, national IDs, payment cards, and IP addresses are stripped from
+  content before it is ever stored.
+- **Bounded loops — gate-verified iteration.** A loop terminates only when an
+  independent gate passes (a test exit code, a linter, a JSON-schema check, or
+  an SLM-recall condition) — never when the agent reports completion. Every lap
+  is persisted to SLM memory under the tag `loop:<name>`, making runs auditable
+  and resumable across sessions. Ships on three surfaces: the `slm loop` CLI
+  (`demo` / `history` / `show`), the `/slm-loop` skill with the
+  `slm-loop-runner` agent, and the MCP tools `slm_loop_run` /
+  `slm_loop_history` / `slm_loop_show` (available in the `code` and `full`
+  profiles).
+- **Nine framework adapters.** SLM now ships adapters under `ide/integrations/`
+  for LangGraph, Semantic Kernel, Microsoft Agent Framework, LangChain,
+  LlamaIndex, CrewAI, AutoGen, Google ADK, and OpenAI Agents. Each adapter
+  wires SLM as the memory and history provider without replacing the framework's
+  agent runtime. Pydantic AI is intentionally not included — it does not expose
+  a formal memory interface for external providers.
+- **Multi-Agent Memory dashboard page.** A new dashboard workspace shows
+  per-agent write activity and attribution for environments where multiple
+  agents share one SLM deployment. Memory entries are stamped via
+  `SLM_AGENT_ID`; the page surfaces per-agent write counts, recent activity,
+  and agent trust signals.
+
+### Changed / Fixed
+
+- **Strict tenant isolation across the whole product.** Coordination between
+  agents (peers, messages, shared state, file locks, activity log) is now scoped
+  per workspace, so different teams or companies sharing one deployment can never
+  see each other's activity.
+- **A person can only enter a workspace they belong to.** Switching into a
+  workspace now requires membership.
+- **Tighter file permissions.** Memory, audit, and learning databases are now
+  owner-only on disk.
+
+## [3.7.9] - 2026-07-20 — Dashboard, skill-evolution, and security hardening
+
+### Fixed
+
+- **Skill evolution now produces output.** A token-limit mismatch made every evolution attempt fail silently and produce nothing; the ceiling is corrected and the underlying misconfiguration is logged instead of masked.
+- **Manual evolution respects its cost caps.** Triggering `evolve_skill` directly now runs under the same per-cycle, wall-time, and per-day LLM caps as automatic evolution (previously it could run uncapped).
+- **Dashboard landing page loads on first open.** The Operating Mode, LLM Provider, Memories, and Version cards now populate immediately instead of staying on "Loading…".
+- **Operating mode is consistent everywhere.** The active mode is read from a single source of truth, so the CLI, daemon, and dashboard always agree and the chosen mode and provider are actually used.
+- **Code-graph updates are safe against hostile repositories.** `update_code_graph` rejects paths outside your home directory and runs `git` with hook and config execution disabled.
+- **LanceDB tier updates escape identifiers**, closing a filter-injection path.
+- **Optimize savings** no longer errors when no model is configured.
+
+### Added
+
+- **Configurable, lowest-cost skill-evolution models.** Each step (generate / verify / confirm) defaults to the cheapest capable model for your backend, keeps the blind verifier independent of the generator, and is settable from the CLI or dashboard. Enabling evolution shows a cost advisory.
+- **Dashboard write endpoint** (`POST /api/evolution/config`) for evolution settings, validated against the same allow-list as the CLI.
+- **Test coverage** for the Gmail and Calendar ingestion adapters.
+
 ## [3.7.8] - 2026-07-20 — Profile-isolation leak fix, loopback auth opt-in, hardening
 
 ### Fixed

package/README.md

@@ -5,10 +5,10 @@
   </picture>
 </p>
 
-<h1 align="center">SuperLocalMemory V3.7.8</h1>
-<p align="center"><strong>Cache. Compress. Remember. Three surfaces — proxy, MCP tools, or skill. Every setup covered.</strong><br/>
-<em>Local-first agent memory with explicit operating modes, auditable retrieval, and optional Optimize tools.</em></p>
-<p align="center"><code>v3.7.8</code> — <strong>Profile-isolation leak fix, opt-in loopback write auth, and fresh-install acceptance verification.</strong><br/>
+<h1 align="center">SuperLocalMemory V3.8.1</h1>
+<p align="center"><strong>Enterprise-grade, local-first memory for AI agents and teams.</strong><br/>
+<em>A persistent, auditable long-term brain for your agents that runs on your own infrastructure — with multi-workspace isolation, role-based access, and GDPR + EU AI Act governance controls built in.</em></p>
+<p align="center"><code>v3.8.1</code> — one control plane: auditable retrieval · multi-scope memory (personal / shared / global) · Cache · Compress · trusted-peer Mesh · bounded loops — across CLI, MCP, dashboard, the <strong>Claude plugin</strong>, the <strong>Codex add-on</strong>, and documented IDE integrations.<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 public research preprints</strong> (arXiv + Zenodo archives) · <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>
 
@@ -19,6 +19,7 @@ Proxy: <code>slm wrap claude</code> &nbsp;·&nbsp; MCP: add <code>slm_compress</
   <a href="https://www.npmjs.com/package/superlocalmemory"><img src="https://img.shields.io/npm/v/superlocalmemory?style=for-the-badge&logo=npm&logoColor=white" alt="npm"/></a>
   <a href="https://www.gnu.org/licenses/agpl-3.0"><img src="https://img.shields.io/badge/License-AGPL_v3-blue.svg?style=for-the-badge" alt="AGPL v3"/></a>
   <a href="#privacy-controls-and-operating-modes"><img src="https://img.shields.io/badge/Privacy-Deployment_Assessed-brightgreen?style=for-the-badge" alt="Privacy controls require deployment assessment"/></a>
+  <a href="#teams-and-enterprise-memory-v380"><img src="https://img.shields.io/badge/Enterprise-GDPR_%7C_EU_AI_Act-0b5394?style=for-the-badge" alt="Enterprise governance: GDPR and EU AI Act controls"/></a>
   <a href="https://superlocalmemory.com"><img src="https://img.shields.io/badge/Web-superlocalmemory.com-ff6b35?style=for-the-badge" alt="Website"/></a>
   <a href="#dual-interface-mcp--cli"><img src="https://img.shields.io/badge/MCP-Native-blue?style=for-the-badge" alt="MCP Native"/></a>
   <a href="#dual-interface-mcp--cli"><img src="https://img.shields.io/badge/CLI-Agent--Native-green?style=for-the-badge" alt="CLI Agent-Native"/></a>
@@ -29,14 +30,29 @@ Proxy: <code>slm wrap claude</code> &nbsp;·&nbsp; MCP: add <code>slm_compress</
 
 ## Why SuperLocalMemory?
 
+SuperLocalMemory is an enterprise-grade, local-first memory control plane for AI agents. Your team's agent memory lives on infrastructure you control, with per-workspace isolation, role-based access, and GDPR / EU AI Act governance controls — built for organizations, and for EU data-residency obligations where agent context must not leave your environment by default.
+
 Agent-memory systems make different storage, model-provider, and deployment trade-offs. SuperLocalMemory starts with a local runtime and makes provider-backed enrichment, cloud backup, connectors, and proxy use explicit choices.
 
-SuperLocalMemory V3 combines conventional dense and lexical retrieval with graph, temporal, associative, and Fisher-informed scoring. The default local runtime does not require Docker, a separately operated graph database, or an API key.
+Different products solve different boundaries. The published benchmark evidence carried into V3.8.0 is protocol-scoped evidence from the published V3 research, not a claim of a newly rerun V3.8.0 package benchmark.
+
+SuperLocalMemory V3 combines conventional dense and lexical retrieval with graph, temporal, associative, and statistical relevance scoring. The default local runtime does not require Docker, a separately operated graph database, or an API key.
+
+**Memory with a sense of time.** SLM does not only store *what* an agent learned — it records *when*. Every fact carries ingestion timing and provenance; recall runs a dedicated temporal candidate channel alongside semantic, lexical, and associative retrieval; scenes and entity timelines reconstruct sequence; and the lifecycle lets neglected memory decay and self-archive instead of growing without bound. Time is a first-class ranking and lifecycle signal rather than a timestamp column an agent never reads — which is what lets a long-lived agent reason about how its context changed, not only what it currently holds.
+
+**What V3.8.0 added.** The 3.8.0 capability release introduced the following
+foundation; 3.8.1 is the existing-install stability patch for it:
+
+- **Temporal depth** — the time-aware retrieval and lifecycle described above.
+- **Governance & EU compliance** — [team roles, workspace isolation, a login gate, multi-scope memory, GDPR access/erasure/portability rights, a hash-chained audit trail, and per-mode EU AI Act self-assessment](#teams-and-enterprise-memory-v380).
+- **Framework adapters** — [drop-in, engine-backed memory for nine agent frameworks](#framework-adapters-v380).
+- **Bounded loops** — [gate-verified agent loops where an independent check, not the agent's own claim, decides when a task is done](#bounded-loops-v380).
+- **Stronger cache and compression** — exact-match caching with tagged invalidation plus opt-in reversible compression, across proxy, MCP, and skill surfaces.
+- **Stability** — a long defect-and-audit sweep across ingestion, retrieval, mesh, and the dashboard hardens the everyday path.
+
+SLM is one strand of Qualixar's work on AI reliability engineering: making agent behavior observable, bounded, and reproducible instead of best-effort.
 
-**Published benchmark evidence carried into V3.7:** the architecture evaluated
-in the V3 paper remains the foundation of this release. The figures below keep
-their original LoCoMo protocol, answer-construction, model, and sample scope;
-they are not a claim of a newly rerun 3.7 package benchmark.
+The architecture evaluated in the V3 paper remains the foundation of this release. The figures below keep their original LoCoMo protocol, answer-construction, model, and sample scope.
 
 ### How SLM fits beside other memory systems
 
@@ -61,7 +77,7 @@ for current primary sources and protocol-scoped benchmark evidence. A LoCoMo
 percentage is comparable only when the dataset scope, answer model, judge,
 retrieval stack, and release artifact match.
 
-### The V3.7 capability architecture
+### The V3.8.0 capability architecture
 
 SuperLocalMemory is one local control plane for persistent agent context. It is
 not just a vector store: the same runtime can accept evidence, build and govern
@@ -69,7 +85,7 @@ memory, retrieve bounded evidence for an agent, and expose cache, compression,
 and peer-coordination controls through a CLI, MCP, dashboard, and supported
 IDE integrations.
 
-![SuperLocalMemory V3.7 capability architecture: modes, seven operating layers, Scale Engine, Mesh, delivery surfaces, and opt-in adapters](docs/assets/slm-v37-capability-architecture.png)
+![SuperLocalMemory V3 capability architecture: modes, seven operating layers, Scale Engine, Mesh, delivery surfaces, and opt-in adapters](docs/assets/slm-v37-capability-architecture.png)
 
 *Architecture boundary: SQLite + sqlite-vec remain canonical; CozoDB and
 LanceDB are parity-gated projections; Mesh coordinates trusted peers rather
@@ -128,6 +144,7 @@ health surfaces expose the stages actually completed by the installed runtime.
 | Knowledge Graph and Memories | graph neighborhoods, entities, scenes, temporal evidence, memory inspection and mutation |
 | Operations | ingestion-operation state, traces, maintenance and lifecycle work |
 | Entity Explorer and Skill Evolution | compiled entity summaries/timelines; opt-in skill lineage, budgets and verification outcomes |
+| Multi-Agent Memory | per-agent write activity and attribution; memories stamped by `SLM_AGENT_ID`, agent write counts, and trust signals |
 | Mesh Peers | configured peers, inbox/outbox, pending coordination and locks |
 | Settings and Optimize | mode/provider/configuration; cache, compression and savings telemetry |
 
@@ -141,9 +158,9 @@ deployment.
 
 **[Watch the SuperLocalMemory demo on YouTube](https://www.youtube.com/watch?v=PMWW_ypsL60)** — a five-minute walkthrough of installation, setup, recall, cache, and compression. The video shows a product walkthrough; use the commands and release notes in this README as the current release contract.
 
-### Published LoCoMo evidence carried into V3.7
+### Published LoCoMo evidence (V3 architecture, carried into V3.8.0)
 
-The V3 paper evaluates the architecture carried into V3.7. Every figure below
+The V3 paper evaluates the architecture carried into V3.8.0. Every figure below
 is protocol-scoped, so a reader can distinguish local retrieval, answer
 construction, and cloud-assisted evaluation rather than treating unlike runs as
 one score.
@@ -164,8 +181,7 @@ information-geometric layers versus **58.9%** without them: **+12.7pp**.
 See [arXiv:2603.14588](https://arxiv.org/abs/2603.14588) and the [official
 LoCoMo paper](https://arxiv.org/abs/2402.17753) for the full protocol,
 ablation table, and limitations. These are published V3 architecture results
-carried into V3.7—not a substitute for a newly rerun release-artifact
-benchmark.
+carried into V3.8.0—not a substitute for a newly rerun release-artifact benchmark.
 
 ---
 
@@ -255,7 +271,7 @@ retrieved at runtime rather than copied into those files.
 **Score Contract v2:** `relevance_score` is query-relative relevance;
 `ranking_score` is internal ranking utility; `memory_confidence` belongs to the
 stored assertion; and `trust_score` is an evidence-policy signal. Legacy
-`score` and `confidence` remain aliases for one compatibility release. V3.7 is
+`score` and `confidence` remain aliases for one compatibility release. V3.8.0 is
 explicitly uncalibrated: `calibration_status` is `uncalibrated` and
 `answer_confidence` is `null`. See
 [the retrieval score contract](docs/retrieval-score-contract.md).
@@ -269,8 +285,7 @@ can run without a cloud LLM:
 
 Auto-capture hooks are installed explicitly with `slm hooks install` (Claude
 Code) or `slm hooks install --agent codex` (Codex). Hook latency and capture
-quality must be evaluated for the target client and workload; V3.7 publishes
-no universal p99 claim.
+quality must be evaluated for the target client and workload; V3.8.0 publishes no universal p99 claim.
 
 **Multi-scope memory (v3.6.15, opt-in):** keep memories `personal` (default), `shared` with named profiles, or `global` across the machine. Off by default — recall only ever returns your own facts until you turn sharing on, per call or in config. See **[docs/shared-memory.md](docs/shared-memory.md)**.
 
@@ -325,7 +340,7 @@ export SLM_MESH_SHARED_SECRET=my-secret-key
 slm init
 ```
 
-8 mesh MCP tools: `mesh_peers`, `mesh_send`, `mesh_broadcast`, `mesh_project`, `mesh_inbox`, `mesh_pending`, `mesh_state`, `mesh_lock`.
+8 mesh MCP tools: `mesh_summary`, `mesh_peers`, `mesh_send`, `mesh_inbox`, `mesh_state`, `mesh_lock`, `mesh_events`, `mesh_status`.
 
 Full docs: [docs/multi-machine.md](docs/multi-machine.md) · [docs/distributed-deployment.md](docs/distributed-deployment.md)
 
@@ -377,10 +392,10 @@ Control tool surface via `SLM_MCP_PROFILE`:
 | Profile | Tools | Use case |
 |:--------|:-----:|:---------|
 | `core` | 14 | Memory, session, and optimize core |
-| `code` | 20 | Core + code-graph tools |
+| `code` | 24 | Core + code-graph tools + profile switching + bounded loops |
 | `mesh` | 8 | Mesh-only — multi-machine coordination |
-| `full` | 38 | Memory + optimize + evolution + mesh |
-| `power` | 50 | Full + administration, lifecycle, and diagnostics |
+| `full` | 42 | Memory + optimize + evolution + mesh + bounded loops |
+| `power` | 54 | Full + administration, lifecycle, and diagnostics |
 | `whole` | all registered | Every registered MCP tool |
 
 **Precedence:** `ALL` > `TOOLS` > `PROFILE` > `default`
@@ -392,10 +407,7 @@ slm mcp
 
 For a predictable small surface, set `core` explicitly. Leaving the variable
 unset retains the compatibility default, whose mesh tools follow the local
-mesh setting. The former count-suffixed names (`core14`, `code20`, `mesh8`,
-`full38`, `power50`, `whole81`) remain temporary aliases and emit a migration
-warning. Unknown names stop startup instead of silently selecting another tool
-set.
+mesh setting. Count-suffixed aliases remain for backward compatibility and emit a migration warning: `core14`, `code20`, `code21`, `code24`, `mesh8`, `full38`, `full39`, `full42`, `power50`, `power51`, `power54`, `whole81`, `whole84`. Unknown names stop startup instead of silently selecting another tool set.
 
 Per-IDE configs available for Claude Code, Cursor, Windsurf, VS Code Copilot, Continue, Gemini CLI, JetBrains, Zed, and more (15 configs in `ide/configs/`). See [docs/ide-setup.md](docs/ide-setup.md).
 
@@ -455,6 +467,22 @@ not a byte-preserving operation; use it only when you want the MCP server
 configured. Check the result with `slm codex status`; undo SLM-owned add-ons
 with `slm codex remove`.
 
+## GitHub Copilot integration
+
+The shipped installer configures the SuperLocalMemory MCP server and additive
+agent instructions for VS Code with GitHub Copilot:
+
+```bash
+slm connect vscode-copilot --here
+```
+
+Run it from the project root. It semantically merges the SLM server into
+`.vscode/mcp.json` and adds SLM-owned guidance inside
+`.github/copilot-instructions.md`, preserving unrelated servers and existing
+instructions. The generated `copilot-plugin/` source bundle is maintained for
+parity checks, but v3.8.1 does not claim that `slm connect` installs its prompt,
+agent, or hook files.
+
 ---
 
 ## Privacy controls and operating modes
@@ -481,6 +509,141 @@ Available controls include local export and erasure commands, hash-chained audit
 
 ---
 
+## Teams and Enterprise Memory (v3.8.0)
+
+V3.8.0 adds multi-user, multi-workspace controls for teams and organizations. These are opt-in — personal single-user installs work exactly as before with no required login.
+
+### Users and roles
+
+SLM supports three role tiers within a workspace: **admin**, **member**, and **viewer**.
+
+| Role | Can read memory | Can write memory | Can manage users/config |
+|------|:---------------:|:----------------:|:-----------------------:|
+| admin | yes | yes | yes |
+| member | yes | yes | no |
+| viewer | yes | no | no |
+
+Roles are scoped per workspace (profile). A user may have different roles in different workspaces.
+
+### Workspace isolation
+
+Each workspace (profile) is a fully isolated memory namespace. One workspace cannot read another's personal memories. Shared and global scopes are opt-in and still profile-bounded at the authorization layer.
+
+### Login gate
+
+Enterprise deployments set `require_login = true` in configuration. With login enabled:
+- Every dashboard and API request requires an authenticated session.
+- First-run creates an admin account with a user-chosen password (no default credentials are shipped).
+- Session cookies use `HttpOnly` with optional `Secure` enforcement.
+- Personal installs run with `require_login = false` (loopback owner is trusted).
+
+```bash
+slm config set security.require_login true   # Enable for team/enterprise use
+```
+
+### Memory scopes
+
+| Scope | Who can recall | Set with |
+|-------|---------------|----------|
+| `personal` | Owner profile only (default) | `slm remember "..." --scope personal` |
+| `shared` | Named profiles the owner grants | `slm remember "..." --scope shared --shared-with profile-a,profile-b` |
+| `global` | Any authorized user on this machine | `slm remember "..." --scope global` |
+
+Recall is default-deny: shared and global facts are never returned unless the caller explicitly opts in (`--include-shared`, `--include-global`) or the scope policy allows it. See [docs/shared-memory.md](docs/shared-memory.md).
+
+### GDPR and data governance
+
+SLM ships built-in controls that support GDPR compliance programs:
+
+- **Export** — full profile data export as a structured JSONL bundle
+- **Erasure** — profile deletion removes data from 30+ scoped tables; erasure is logged to the tamper-proof audit chain before any data is deleted
+- **Retention rules** — time-based policies (`indefinite`, `gdpr-30d`, `hipaa-7y`, `custom`) applied per profile
+- **Audit trail** — every store, recall, mutation, and erasure produces a hash-chained audit record
+- **PII redaction** — configurable automatic redaction before memory content crosses trust boundaries
+
+These are engineering controls. Compliance depends on deployment configuration, use case, and operator responsibility. See [docs/compliance.md](docs/compliance.md).
+
+### EU AI Act mode verification
+
+SLM includes a per-mode EU AI Act self-assessment. The `EUAIActChecker` produces a compliance report for the active operating mode — risk category, whether data stays local, whether generative AI is used, and transparency / human-oversight signals:
+
+- **Mode A (Local Guardian)** and **Mode B (Smart Local)** — assessed as compliant: memory processing stays local and uses no generative AI.
+- **Mode C (Provider-assisted)** — flagged non-compliant, because query or enrichment content is sent to a cloud model provider.
+
+This is operator self-assessment tooling, not a legal certification or conformity assessment; actual EU AI Act obligations depend on your system, deployment, and role. See [docs/compliance.md](docs/compliance.md).
+
+### Deployment tiers
+
+SLM ships one binary and is configured for the appropriate tier at install or post-install time.
+
+| Tier | Login gate | PII redaction | Retention | Audit |
+|------|:---------:|:-------------:|:---------:|:-----:|
+| **Personal** | off | off | off | on |
+| **Enterprise** | on | on | on | on |
+
+The installer or `slm reconfigure` sets the tier. Each setting is independently overridable at runtime. Full tier documentation: [docs/deployment-tiers.md](docs/deployment-tiers.md).
+
+### RBAC and teams docs
+
+Full reference: [docs/rbac-teams.md](docs/rbac-teams.md) · [docs/deployment-tiers.md](docs/deployment-tiers.md)
+
+---
+
+## Bounded Loops (v3.8.0)
+
+A bounded loop terminates only when an **independent gate** passes — a test
+suite exit code, a linter, a JSON-schema check, or an SLM-recall condition.
+The agent's own "I finished" message is recorded as advisory context and never
+used as the termination signal. Every lap is persisted to SLM memory under the
+tag `loop:<name>`, so runs are auditable and resumable across sessions.
+
+Three surfaces ship together:
+
+| Surface | How you use it |
+|---------|---------------|
+| **CLI** | `slm loop demo` · `slm loop history [--name <n>]` · `slm loop show <run_id>` |
+| **Skill + agent** | `/slm-loop` skill with the `slm-loop-runner` agent — delegate a task that has a checkable acceptance condition |
+| **MCP tools** | `slm_loop_run` · `slm_loop_history` · `slm_loop_show` — call from any IDE or agent (available in the `code` and `full` MCP profiles) |
+
+```bash
+# Run the built-in convergence demo (no API key needed)
+slm loop demo
+
+# Inspect recorded runs
+slm loop history --name convergence-demo
+slm loop show <run_id>
+```
+
+Loop laps are stored as ordinary SLM memories and are visible in the dashboard
+under Knowledge Graph and Memories (filter by tag `loop:<name>`) and in the
+Multi-Agent Memory workspace.
+
+---
+
+## Framework Adapters (v3.8.0)
+
+SLM ships nine framework adapters under `ide/integrations/`. Each adapter
+wires SLM as the memory and history provider for the respective framework
+without replacing the framework's own agent runtime.
+
+| Framework | Directory |
+|-----------|-----------|
+| LangGraph | `ide/integrations/langgraph/` |
+| Semantic Kernel | `ide/integrations/semantic-kernel/` |
+| Microsoft Agent Framework | `ide/integrations/agent-framework/` |
+| LangChain | `ide/integrations/langchain/` |
+| LlamaIndex | `ide/integrations/llamaindex/` |
+| CrewAI | `ide/integrations/crewai/` |
+| AutoGen | `ide/integrations/autogen/` |
+| Google ADK | `ide/integrations/google-adk/` |
+| OpenAI Agents | `ide/integrations/openai-agents/` |
+
+Pydantic AI is not included — it does not expose a formal memory interface for
+external providers. Each adapter's `README.md` covers installation and
+configuration for that framework.
+
+---
+
 ## Advanced
 
 | Topic | Link |
@@ -495,6 +658,8 @@ Available controls include local export and erasure commands, hash-chained audit
 | MCP tools reference | [docs/mcp-tools.md](docs/mcp-tools.md) |
 | Getting started | [docs/getting-started.md](docs/getting-started.md) |
 | IDE setup (15 configs) | [docs/ide-setup.md](docs/ide-setup.md) |
+| Teams, users, and RBAC | [docs/rbac-teams.md](docs/rbac-teams.md) |
+| Deployment tiers | [docs/deployment-tiers.md](docs/deployment-tiers.md) |
 | pi.dev integration | [docs/pi-dev-integration.md](docs/pi-dev-integration.md) |
 | Skill evolution | [docs/skill-evolution.md](docs/skill-evolution.md) |
 | V2 migration | [docs/migration-from-v2.md](docs/migration-from-v2.md) |
@@ -502,82 +667,19 @@ Available controls include local export and erasure commands, hash-chained audit
 | Retrieval score contract | [docs/retrieval-score-contract.md](docs/retrieval-score-contract.md) |
 | Wiki | [github.com/qualixar/superlocalmemory/wiki](https://github.com/qualixar/superlocalmemory/wiki) |
 
-**Web dashboard:**
-```bash
-slm dashboard    # Opens at http://localhost:8765
-```
-The dashboard includes Dashboard, Brain, Knowledge Graph, Memories, Health,
-Operations, Entity Explorer, Skill Evolution, Mesh Peers, Settings, and
-Optimize workspaces. Features are populated only when their corresponding
-runtime capability is enabled and healthy.
-
-**Release history:**
-
-| Version | Codename | Key Features |
-|---|---|---|
-| **v3.6.23** | Cross-platform Patch | Windows doctor/cache stats fixes (#65), neutral SLM hook guidance (#64), pi.dev MCP docs (#31), contributor fixes for dashboard profile path resolution (#63) and tz-naive Langevin maintenance backfill (#66) |
-| **v3.6.22** | Stability | backbone.py JSONDecodeError on empty HTTP 200 body (issue #62) — retries 3× then returns "" gracefully; remaining dashboard UI audit: clusters/compliance/entities r.ok guards, math-health status badge colors |
-| **v3.6.21** | Dashboard Audit | Full UI audit across all 7 dashboard tabs — auth fix for mesh panel (issue #60 frontend), Quick Store endpoint, timeline endpoint, r.ok guards, SSE \r fix, event delegation for lazy tabs, optimize toggle revert |
-| **v3.6.20** | Mesh Auth | Remote mesh auth fix (issue #60) — `_get_broker` now accepts Bearer + X-Mesh-Secret from non-loopback callers; config settings preservation (AIDEV-86) |
-| **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 |
-| **v3.5.0** | Historical scale work | Early CozoDB/LanceDB projection paths, retrieval additions, Core Memory Block, context injection v2, score normalization |
-| **v3.4.x** | Scale-Ready (foundation) | Tiered storage, graph pruning, Hopfield channel, LightGBM ranking, mDNS mesh discovery |
-| **v3.3.x** | Foundation | BM25Plus, Fisher-Rao, sqlite-vec, RRF fusion, cross-encoder rerank. 3 published papers |
-
----
+Open the web dashboard with `slm dashboard`; workspaces appear only when their
+runtime capability is enabled and healthy. See [CHANGELOG.md](CHANGELOG.md) for
+the complete release history.
 
 ## Research Papers
 
-SuperLocalMemory is backed by three published research papers (arXiv preprints + Zenodo DOIs). These are preprints — not conference-accepted or journal-published yet.
-
-### Paper 3: The Living Brain (V3.3)
-> **SuperLocalMemory V3.3: The Living Brain — Biologically-Inspired Forgetting, Cognitive Quantization, and Multi-Channel Retrieval for Zero-LLM Agent Memory Systems**
-> Varun Pratap Bhardwaj (2026)
-> [arXiv:2604.04514](https://arxiv.org/abs/2604.04514) · [Zenodo DOI: 10.5281/zenodo.19435120](https://zenodo.org/records/19435120)
-
-### Paper 2: Information-Geometric Foundations (V3)
-> **SuperLocalMemory V3: Information-Geometric Foundations for Zero-LLM Enterprise Agent Memory**
-> Varun Pratap Bhardwaj (2026)
-> [arXiv:2603.14588](https://arxiv.org/abs/2603.14588) · [Zenodo DOI: 10.5281/zenodo.19038659](https://zenodo.org/records/19038659)
-
-### Paper 1: Trust & Behavioral Foundations (V2)
-> **SuperLocalMemory: A Structured Local Memory Architecture for Persistent AI Agent Context**
-> Varun Pratap Bhardwaj (2026)
-> [arXiv:2603.02240](https://arxiv.org/abs/2603.02240) · [Zenodo DOI: 10.5281/zenodo.18709670](https://zenodo.org/records/18709670)
-
-### Cite This Work
-
-```bibtex
-@article{bhardwaj2026slmv33,
-  title={SuperLocalMemory V3.3: The Living Brain — Biologically-Inspired
-         Forgetting, Cognitive Quantization, and Multi-Channel Retrieval
-         for Zero-LLM Agent Memory Systems},
-  author={Bhardwaj, Varun Pratap},
-  journal={arXiv preprint arXiv:2604.04514},
-  year={2026},
-  url={https://arxiv.org/abs/2604.04514}
-}
-
-@article{bhardwaj2026slmv3,
-  title={Information-Geometric Foundations for Zero-LLM Enterprise Agent Memory},
-  author={Bhardwaj, Varun Pratap},
-  journal={arXiv preprint arXiv:2603.14588},
-  year={2026}
-}
-
-@article{bhardwaj2026slm,
-  title={A Structured Local Memory Architecture for Persistent AI Agent Context},
-  author={Bhardwaj, Varun Pratap},
-  journal={arXiv preprint arXiv:2603.02240},
-  year={2026}
-}
-```
+SuperLocalMemory is backed by three preprints by Varun Pratap Bhardwaj (2026):
 
----
+- **The Living Brain (V3.3):** [arXiv:2604.04514](https://arxiv.org/abs/2604.04514) · [Zenodo 19435120](https://zenodo.org/records/19435120)
+- **Information-Geometric Foundations (V3):** [arXiv:2603.14588](https://arxiv.org/abs/2603.14588) · [Zenodo 19038659](https://zenodo.org/records/19038659)
+- **Trust & Behavioral Foundations (V2):** [arXiv:2603.02240](https://arxiv.org/abs/2603.02240) · [Zenodo 18709670](https://zenodo.org/records/18709670)
+
+Use the citation metadata on the linked arXiv or Zenodo records.
 
 ## Support / License / Qualixar
 
@@ -593,34 +695,14 @@ Part of [Qualixar](https://qualixar.com) · Author: [Varun Pratap Bhardwaj](http
 
 ### Acknowledgments
 
-- **[Everything Claude Code (ECC)](https://github.com/affaan-m/everything-claude-code)** — SLM's skill observation patterns were inspired by ECC's continuous learning architecture. SLM supports direct ingestion of ECC observations via `slm ingest --source ecc`. We recommend ECC for Claude Code users who want the deepest learning experience alongside SLM.
-- **[HKUDS/OpenSpace](https://github.com/HKUDS/OpenSpace)** — The skill evolution research in SLM draws from the EvoSkills co-evolutionary verification concepts (arXiv:2604.01687). We adopted their 3-trigger evolution system and anti-loop guard patterns.
+- **[Everything Claude Code (ECC)](https://github.com/affaan-m/everything-claude-code)** inspired SLM's skill-observation patterns; SLM can ingest ECC observations with `slm ingest --source ecc`.
+- **[HKUDS/OpenSpace](https://github.com/HKUDS/OpenSpace)** informed the skill-evolution verification design (arXiv:2604.01687).
 
 ### Qualixar AI Agent Reliability Platform
 
-Qualixar is building the open-source infrastructure for AI agent reliability engineering. Seven products, one coherent platform:
-
-| Product | Purpose | Install |
-|---------|---------|---------|
-| **[SuperLocalMemory](https://github.com/qualixar/superlocalmemory)** | Persistent memory + learning | `npm install -g superlocalmemory` |
-| **[Qualixar OS](https://github.com/qualixar/qualixar-os)** | Universal agent runtime | `npx qualixar-os` |
-| **[SLM Mesh](https://github.com/qualixar/slm-mesh)** | P2P coordination across sessions | `npm i slm-mesh` |
-| **[SLM MCP Hub](https://github.com/qualixar/slm-mcp-hub)** | Federate 430+ MCP tools | `pip install slm-mcp-hub` |
-| **[AgentAssay](https://github.com/qualixar/agentassay)** | Token-efficient agent testing | `pip install agentassay` |
-| **[AgentAssert](https://github.com/qualixar/agentassert-abc)** | Behavioral contracts + drift detection | `pip install agentassert-abc` |
-| **[SkillFortify](https://github.com/qualixar/skillfortify)** | Formal verification for agent skills | `pip install skillfortify` |
-
-**Local-first architecture. Deployment-specific privacy and compliance controls.**
-
-Start here → **[qualixar.com](https://qualixar.com)** · [All papers on Qualixar HuggingFace](https://huggingface.co/Qualixar)
-
----
-
-<p align="center">
-  <sub>Built with mathematical rigor. Not in the race — here to help everyone build better AI memory systems.</sub>
-</p>
-
----
+Qualixar builds open-source infrastructure for AI reliability engineering.
+Start at **[qualixar.com](https://qualixar.com)** or browse the
+[Qualixar research archive](https://huggingface.co/Qualixar).
 
 ## Star This Project