feat(graph): GraphRAG community detection + community-summary global recall (#365)

Adds the GraphRAG global-sensemaking layer over the existing entity link
graph, closing the "local traversal only" gap:

- mimir_communities: partition the per-workspace link graph into
  communities. Pure-Rust deterministic algorithms, no new dependencies:
  label propagation (default; neighborhood-overlap edge weighting so
  bridges never tie with intra-cluster edges) or greedy one-level
  modularity optimization ("louvain"). Reports Newman modularity and
  persists the run in a new `communities` table (schema v8, #358
  BEGIN IMMEDIATE + re-check migration pattern, self-contained block).
- Community summaries: extractive by default (top members by
  in-community degree, capped at 1200 chars) so the engine stays fully
  offline; mimir_community_summary optionally polishes with the
  configured LLM and always degrades back to extractive. Summaries are
  materialized as `community_summary` entities carrying evidence_for
  links to members. Community ids are digests of the sorted member set,
  so membership changes invalidate summaries by construction (state-
  digest cache-key pattern, #256); stale summary entities are archived
  on re-detection, and summary entities are excluded from the graph so
  they can't glue communities together.
- mimir_global_recall: breadth over community summaries first, then
  drill-down into the best communities' members, round-robined so a
  query spanning multiple clusters cites entities from EACH of them.
  Deterministic token-overlap scoring (offline); optional use_llm
  synthesis. Auto-detects communities on first use.
- mimir_stats now reports total_communities + graph_modularity.
- Entity bodies are untrusted: everything spliced into summaries,
  answers, or LLM prompts goes through sanitize_prompt_field (#337).
- Member hydration uses the chunked IN(...) batch pattern (#340).

Schema: user_version 6 -> 8 (v8 block is self-contained for easy
renumbering when parallel migrations merge). Tools: 49 -> 52 (README,
server.json, manifest.json, glama.json repinned). Tests: 188 -> 204,
covering planted-cluster recovery (modularity > 0.3, category-pure
membership), byte-stable determinism across runs, summary caps and
cache reuse, membership-change invalidation, cross-cluster global
recall, workspace scoping, legacy-DB migration, and MCP dispatch
incl. mneme_*/perseus_vault_* aliases.

Closes #365

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: tcconnally

CHANGELOG.md

@@ -5,6 +5,18 @@ All notable changes to Perseus Vault (formerly Mimir/Mneme) are documented here.
 
 ## [Unreleased]
 
+### Added
+- GraphRAG over the link graph (#365): `mimir_communities` (deterministic
+  community detection — label propagation with neighborhood-overlap weighting,
+  or greedy one-level modularity "louvain"; pure Rust, no new dependencies),
+  `mimir_community_summary` (extractive by default, optional LLM polish,
+  materialized as a `community_summary` entity with `evidence_for` links,
+  cached by member-set digest), and `mimir_global_recall` (breadth over
+  community summaries, then depth into the best communities' members — cites
+  entities across clusters instead of only the nearest one). Communities are
+  persisted in a new `communities` table (schema v8); `mimir_stats` now
+  reports `total_communities` and `graph_modularity`. 49 → 52 MCP tools.
+
 ### Fixed
 - MCP Registry publish: `server.json` version/OCI identifier now synced from
   `Cargo.toml` at publish time, and the publish waits for the GHCR image tag

README.md

@@ -10,7 +10,7 @@
 [![LangGraph](https://img.shields.io/badge/integrations-LangGraph-blue)](integrations/langgraph/)
 [![CrewAI](https://img.shields.io/badge/integrations-CrewAI-orange)](integrations/crewai/)
 [![AutoGen](https://img.shields.io/badge/integrations-AutoGen-purple)](integrations/autogen/)
-[![MCP Tools](https://img.shields.io/badge/MCP%20tools-49-brightgreen)]()
+[![MCP Tools](https://img.shields.io/badge/MCP%20tools-52-brightgreen)]()
 
 Perseus Vault is a single Rust binary that gives AI agents durable memory across sessions.
 **One binary. One file. No Docker. No Postgres. No cloud.** Just persistent memory
@@ -93,15 +93,15 @@ local-first, zero-dependency, AND agent-first.
 |---|---|---|---|---|
 | **Deployment** | Single binary (~8MB) | Cloud + self-host | Docker/Postgres | Docker/Postgres |
 | **Dependencies** | None (SQLite embedded) | Python + vector DB | Postgres + Python | Postgres + Go |
-| **MCP-Native** | ✅ 49 tools | ❌ Not MCP-native | ❌ Not MCP-native | ❌ Not MCP-native |
+| **MCP-Native** | ✅ 52 tools | ❌ Not MCP-native | ❌ Not MCP-native | ❌ Not MCP-native |
 | **Offline/Local** | ✅ Fully local | Cloud-dependent | Docker needed | Docker needed |
 | **Encryption** | AES-256-GCM ✅ | ❌ | ❌ | ❌ |
 | **Hybrid Search** | BM25 + Dense + RRF | Vector only | Vector only | Vector + Graph |
 | **Entity Lifecycle** | Decay + Promote + Archive | ❌ | ❌ | ❌ |
 | **Entity Graph** | Link + Traverse | ❌ | ❌ | ✅ |
 | **Journal Audit Trail** | ✅ Immutable | ❌ | ❌ | ❌ |
 | **State Management** | ✅ Key-value + TTL | ❌ | ❌ | ❌ |
-| **MCP Tools** | 49 | 5 | 8 | 0 |
+| **MCP Tools** | 52 | 5 | 8 | 0 |
 | **GitHub Stars** | ~20 | ~55K | ~15K | ~3K |
 | **License** | MIT | Apache 2.0 | Apache 2.0 | Apache 2.0 |
 
@@ -143,7 +143,7 @@ Each adapter:
 Any MCP-compatible framework works with Perseus Vault directly. See
 [Awesome Mimir](awesome-mimir.md) for the full list.
 
-## 49 MCP Tools
+## 52 MCP Tools
 
 ### Entity CRUD
 | Tool | Description |
@@ -175,6 +175,9 @@ Any MCP-compatible framework works with Perseus Vault directly. See
 | `mimir_link` | Create typed relationship links between entities. |
 | `mimir_unlink` | Remove entity links. |
 | `mimir_traverse` | Walk entity link graph up to configurable depth. |
+| `mimir_communities` | GraphRAG community detection over the link graph (deterministic label propagation or greedy-modularity "louvain"; pure Rust, offline). |
+| `mimir_community_summary` | Extractive (optionally LLM-polished) summary of one community, materialized as an entity with `evidence_for` links to members. |
+| `mimir_global_recall` | GraphRAG global search: breadth over community summaries, then depth into the best communities' members — holistic answers across clusters. |
 
 ### Journal
 | Tool | Description |

glama.json

@@ -3,7 +3,7 @@
   "maintainers": [
     "tcconnally"
   ],
-  "tools": 49,
+  "tools": 52,
   "transport": {
     "stdio": true,
     "sse": true,

manifest.json

@@ -3,8 +3,8 @@
   "name": "perseus-vault",
   "version": "2.13.0",
   "display_name": "Perseus Vault — Persistent Memory for Claude (formerly Mimir/Mneme)",
-  "description": "Persistent cross-session memory for Claude. 49 MCP tools: remember, recall, search, link entities, journal audit trail, state management, AES-256-GCM encryption. Local-first, single Rust binary, zero cloud dependencies.",
-  "long_description": "Perseus Vault (formerly \"Mimir\", then \"Mneme\") gives Claude durable memory across sessions. It is a single Rust binary (~8MB) with an embedded SQLite database — no Docker, no Postgres, no cloud. Features include:\n\n- **49 MCP tools** for entity storage, hybrid search (FTS5 + dense vector), knowledge-graph traversal, journal audit trail, and state management\n- **AES-256-GCM encryption at rest** (optional, transparent)\n- **Fully local** — never phones home, no telemetry, air-gapped by default\n- **Agent-first design** — entity lifecycle with Ebbinghaus decay, auto-promotion, archiving\n- **Works with any MCP host** — Claude Desktop, Claude Code, Cursor, Hermes Agent, Perseus, and more\n\nInstall in one line: `curl -sSf https://raw.githubusercontent.com/Perseus-Computing-LLC/perseus-vault/main/scripts/install.sh | sh`",
+  "description": "Persistent cross-session memory for Claude. 52 MCP tools: remember, recall, search, link entities, journal audit trail, state management, AES-256-GCM encryption. Local-first, single Rust binary, zero cloud dependencies.",
+  "long_description": "Perseus Vault (formerly \"Mimir\", then \"Mneme\") gives Claude durable memory across sessions. It is a single Rust binary (~8MB) with an embedded SQLite database — no Docker, no Postgres, no cloud. Features include:\n\n- **52 MCP tools** for entity storage, hybrid search (FTS5 + dense vector), knowledge-graph traversal, journal audit trail, and state management\n- **AES-256-GCM encryption at rest** (optional, transparent)\n- **Fully local** — never phones home, no telemetry, air-gapped by default\n- **Agent-first design** — entity lifecycle with Ebbinghaus decay, auto-promotion, archiving\n- **Works with any MCP host** — Claude Desktop, Claude Code, Cursor, Hermes Agent, Perseus, and more\n\nInstall in one line: `curl -sSf https://raw.githubusercontent.com/Perseus-Computing-LLC/perseus-vault/main/scripts/install.sh | sh`",
   "author": {
     "name": "Perseus Computing LLC",
     "email": "privacy@perseus.observer",

server.json

@@ -2,7 +2,7 @@
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
   "name": "io.github.Perseus-Computing-LLC/mimir",
   "title": "Perseus Vault (formerly Mimir/Mneme), persistent memory engine for AI agents",
-  "description": "Local-first MCP memory server: SQLite, FTS5 and vector search, AES-256-GCM. 49 tools, no cloud.",
+  "description": "Local-first MCP memory server: SQLite, FTS5 and vector search, AES-256-GCM. 52 tools, no cloud.",
   "version": "2.13.0",
   "packages": [{
     "registryType": "oci",