@romandp/context-memory-mcp 1.0.0 → 1.0.1

package/README.md

@@ -1,335 +1,112 @@
-# MCP Memory Server
+# Context Memory MCP
 
-Servidor de memoria persistente para asistentes de IA vía **Model Context Protocol (MCP)**. Almacena y organiza planes SDD (Software Design Document) con la estructura **Proyecto → Entradas → Tareas**, todo en una base de datos **SQLite** local.
+Persistent local memory server for AI assistants over the Model Context Protocol (MCP).
 
-Expone tanto un transporte **MCP sobre STDIO** como una **API REST HTTP**.
+It stores project context in SQLite and exposes:
 
----
+- MCP over STDIO
+- HTTP REST API
+- CLI for start/stop/status and maintenance
 
-## Arquitectura
+## What It Stores
 
-```
-           ┌──────────────────────────────────┐
-Cliente    │         MCP Memory Server        │
-MCP ◄────► │  ┌──────────┐  ┌──────────────┐  │
-(stdio)    │  │ MCP over │  │ Express REST │  │
-           │  │   STDIO   │  │ :3001        │  │
-           │  └────┬─────┘  └──────┬───────┘  │
-           │       └──────┬────────┘           │
-           │              ▼                    │
-           │    ┌─────────────────┐            │
-           │    │   better-sqlite3 │            │
-           │    │   (WAL mode)    │            │
-           │    └────────┬────────┘            │
-           └─────────────┼─────────────────────┘
-                         ▼
-       ~/.local/share/mcp-memory/memory.db
-```
-
-### Estructura de datos
-
-```
-projects
- ├── id, name, description, status
- │
- ├── sdd_entries
- │    ├── id, project_id
- │    ├── section (plan | design | tasks | general)
- │    ├── title, content, status, parent_id
- │    └── metadata (JSON)
- │
- ├── tasks
- │    ├── id, project_id, sdd_entry_id
- │    ├── title, description
- │    └── status, priority
- │
-    ├── design_decisions
-    │    ├── id, entry_id
-    │    ├── decision, rationale
-   │    └── alternatives_considered
-   │
-   ├── entry_relationships
-   │    ├── id, source_entry_id, target_entry_id
-   │    └── relationship_type (depends_on|implements|related_to|supersedes)
-   │
-   ├── classifications (polymorphic)
-   │    ├── classifiable_type (project | entry | task)
-   │    └── tag, confidence
-   │
-   ├── audit_log
-   │    ├── id, entity_type, entity_id
-   │    ├── action (created | updated | deleted)
-   │    └── changes (JSON), project_id, timestamp
-   │
-   └── fts_entries (FTS5 virtual table)
-        ├── entry_id, section, title, content
-        └── auto-sincronizada via triggers
-
-_migrations (tracking de cambios de schema)
-```
+- Projects
+- Structured entries (`plan`, `design`, `tasks`, `general`)
+- Tasks
+- Design decisions and relationships
+- Derived summaries and compact memory facts for LLM-friendly retrieval
 
-### Tecnologías
+## Install
 
-| Capa | Tecnología |
-|------|-----------|
-| Lenguaje | TypeScript (ESM) |
-| MCP SDK | `@modelcontextprotocol/sdk` |
-| HTTP Server | Express 5 |
-| Base de datos | SQLite vía `better-sqlite3` (WAL mode) |
-| Validación | Zod |
-| Tests | Vitest + Supertest |
+Requirements:
 
----
+- Node.js 18+
+- npm 9+
 
-## Instalación
-
-### Requisitos
-
-- Node.js ≥ 18
-- npm ≥ 9
-
-### Instalación global desde release/npm
+Install globally:
 
 ```bash
-# Instalar última release publicada
 npm install -g @romandp/context-memory-mcp
-
-# Ver rutas resueltas y versión
-mcp-memory paths
-mcp-memory version
-
-# Iniciar servidor
-mcp-memory start
-
-# Ver estado
-mcp-memory status
 ```
 
-### Uso sin instalación global
+Or run without installing globally:
 
 ```bash
 npx -y @romandp/context-memory-mcp@latest stdio
 ```
 
-### Upgrade / rollback
-
-```bash
-# Actualizar a última release
-npm install -g @romandp/context-memory-mcp@latest
-
-# Instalar versión concreta
-npm install -g @romandp/context-memory-mcp@1.0.0
-```
-
-### Releases
+## Quick Start
 
-- Cada tag `v*` publica paquete npm y crea GitHub Release mediante `.github/workflows/release.yml`
-- El workflow requiere secreto `NPM_TOKEN`
-- El artefacto `.tgz` de la release permite probar instalación sin clonar repo
-
-### Desarrollo local
+Start server in background:
 
 ```bash
-git clone <repo>
-cd context-memory-mcp
-npm install
-npm run build
-node bin/mcp-memory.js start
+mcp-memory start
 ```
 
-### Migrar base de datos antigua
+Check status:
 
 ```bash
-mcp-memory migrate-db --from /ruta/antigua/memory.db
+mcp-memory status
 ```
 
----
-
-## Uso
-
-### CLI
+Show resolved paths:
 
 ```bash
-mcp-memory start      # Inicia en background
-mcp-memory stop       # Detiene el servidor
-mcp-memory status     # Muestra estado (PID, health)
-mcp-memory restart    # Reinicia
-mcp-memory logs       # Tail de logs
-mcp-memory info       # Información del proyecto
-mcp-memory paths      # Muestra rutas resueltas
-mcp-memory version    # Muestra versión instalada
-mcp-memory rebuild-memory   # Regenera summaries/facts compactos
-mcp-memory migrate-db --from /ruta/db.sqlite
+mcp-memory paths
 ```
 
-O via npm:
+Run MCP server in foreground:
 
 ```bash
-npm run start:bg   # start en background
-npm run stop       # stop
-npm run status     # status
-npm start          # start en foreground
-```
-
-### API REST
-
-El servidor HTTP corre en `http://localhost:3001` (configurable con `HTTP_PORT`).
-
-Documentación interactiva disponible en: `http://localhost:3001/api-docs`
-
-> Todos los endpoints de listado soportan paginación con `?page=1&limit=50` (max 200).  
-> La respuesta incluye `pagination: { page, limit, total, totalPages }`.
-
-#### Proyectos
-
+mcp-memory stdio
 ```
-GET    /api/projects                              → Listar proyectos (?page=&limit=)
-POST   /api/projects                              → Crear proyecto { name, description? }
-GET    /api/projects/:id                          → Proyecto con entries + tasks + clasificaciones
-PUT    /api/projects/:id                          → Actualizar proyecto { name?, status?, description? }
-DELETE /api/projects/:id                          → Eliminar proyecto (cascade)
-```
-
-#### Entradas SDD
-
-```
-GET    /api/projects/:pid/entries                     → Listar entradas (?section=&page=&limit=)
-POST   /api/projects/:pid/entries                     → Crear entrada { section, title, content? }
-GET    /api/projects/:pid/entries/search?q=texto      → Buscar entradas en el proyecto (FTS5, ?page=&limit=)
-GET    /api/projects/:pid/entries/:eid                → Obtener entrada + clasificaciones
-PUT    /api/projects/:pid/entries/:eid                → Actualizar entrada
-DELETE /api/projects/:pid/entries/:eid                → Eliminar entrada
-```
-
-#### Búsqueda global
-
-```
-GET    /api/entries/search?q=texto                    → Buscar entradas en todos los proyectos (FTS5, ?page=&limit=)
-GET    /api/entries/search?q=texto&view=compact       → Resultados compactos optimizados para LLM
-```
-
-### TOON y vistas compactas
-
-Parámetros nuevos en rutas/tools compatibles:
-
-- `view=full|summary|compact`
-- `format=json|toon-r|toon-d`
-- `max_items=<n>`
-- `max_chars=<n>`
-- `cursor=<last_id>`
 
-Uso recomendado para agentes/LLM:
+## CLI
 
 ```bash
-curl "http://localhost:3001/api/entries/<eid>/context?view=compact&format=toon-r"
-curl "http://localhost:3001/api/entries/search?q=release&view=compact&format=toon-d"
-curl "http://localhost:3001/api/entries/batch?ids=e1,e2,e3&max_items=2"
-```
-
-Semántica:
-
-- `full`: contrato legacy completo
-- `summary`: respuesta resumida sin contenido largo
-- `compact`: respuesta basada en summaries/facts derivadas
-- `toon-r`: TOON legible para debug
-- `toon-d`: TOON denso para mínimo tamaño
-- `max_items`: corta arrays compactos y marca `truncated`
-- `max_chars`: limita volumen aproximado de texto compacto
-- `cursor`: continúa desde último `id` devuelto en respuesta previa
-- `metrics`: cada respuesta compacta JSON incluye `bytes`, `chars`, `estimated_tokens`
-
-Implementación actual:
-
-- `summary_short` y `summary_dense` derivadas por entry
-- `memory_facts` derivadas desde entries, decisiones y relaciones
-- FTS5 compacta sobre `fts_entry_summaries` para búsquedas TOON/compact
-- Fallback automático al índice legacy si faltan datos derivados
-
-#### Contexto enriquecido
-
-```
-GET    /api/entries/:eid/context                      → Entry + decisions + relationships
-GET    /api/entries/:eid/summary                      → Summary derivada compacta
-GET    /api/entries/:eid/compact                      → Contexto compacto/TOON
-GET    /api/entries/batch?ids=e1,e2                   → Batch compacto por IDs
-POST   /api/entries/:eid/decisions                    → Registrar decisión de diseño { decision, rationale, alternatives_considered? }
-POST   /api/entries/:eid/relationships                → Relacionar entries { target_entry_id, relationship_type }
-GET    /api/facts                                     → Facts derivadas (?project_id=&entry_id=&kind=&page=&limit=&max_items=&max_chars=&cursor=)
-```
-
-#### Tareas
-
-```
-GET    /api/projects/:pid/tasks                   → Listar tareas (?entry_id=&page=&limit=)
-POST   /api/projects/:pid/tasks                   → Crear tarea { title, priority?, sdd_entry_id? }
-PUT    /api/projects/:pid/tasks/:tid              → Actualizar tarea { status? }
-DELETE /api/projects/:pid/tasks/:tid              → Eliminar tarea
+mcp-memory start
+mcp-memory stop
+mcp-memory status
+mcp-memory restart
+mcp-memory logs
+mcp-memory info
+mcp-memory paths
+mcp-memory version
+mcp-memory rebuild-memory
+mcp-memory migrate-db --from /path/to/old/memory.db
 ```
 
-#### Auditoría
-
-```
-GET    /api/audit                                  → Historial de cambios (?entity_type=&entity_id=&project_id=&page=&limit=)
-```
+## HTTP API
 
-#### Clasificaciones
+Default base URL:
 
-```
-POST   /api/classify    → Añadir clasificación { classifiable_type, classifiable_id, tag, confidence }
-GET    /api/db/download → Descargar el fichero SQLite
+```text
+http://localhost:3001
 ```
 
-### MCP Tools
-
-El servidor expone estas herramientas vía MCP (STDIO):
-
-| Tool | Descripción |
-|------|-------------|
-| `project-create` | Crear proyecto |
-| `project-list` | Listar proyectos (con paginación opcional) |
-| `project-get` | Proyecto completo con entries y tasks |
-| `entry-create` | Crear entrada SDD (plan/design/tasks/general) |
-| `entry-get` | Listar entradas de un proyecto (con paginación opcional) |
-| `entry-search` | Buscar entradas por texto (FTS5, con paginación opcional) |
-| `entry-search-global` | Buscar entradas en todos los proyectos (FTS5, sin project_id) |
-| `entry-update` | Actualizar entrada (title, content, status, section, parent_id) |
-| `entry-delete` | Eliminar entrada |
-| `entry-add-decision` | Registrar decisión de diseño |
-| `entry-add-relationship` | Relacionar dos entradas |
-| `entry-batch-get` | Obtener varias entradas compactas por ID |
-| `entry-get-context` | Obtener entrada + decisions + relationships |
-| `entry-get-summary` | Obtener summary derivada compacta |
-| `task-create` | Crear tarea |
-| `task-list` | Listar tareas de un proyecto (con paginación opcional) |
-| `task-update` | Actualizar estado de tarea |
-| `memory-facts-get` | Consultar facts derivadas |
-| `audit-get` | Consultar historial de cambios (filtros: entity_type, entity_id, project_id) |
-
----
-
-## Configuración
-
-Variables de entorno (fichero `.env`):
+Swagger UI:
 
-```env
-HTTP_PORT=3001        # Puerto del servidor REST
-DB_PATH=/ruta/memory.db   # Override directo de base de datos
-MCP_MEMORY_HOME=/ruta/base
-LOG_PATH=/ruta/server.log
-PID_PATH=/ruta/server.pid
+```text
+http://localhost:3001/api-docs
 ```
 
-Rutas por defecto sin overrides:
+Typical endpoints:
 
-- Linux: `~/.local/share/mcp-memory/memory.db` y `~/.local/state/mcp-memory/`
-- macOS: `~/Library/Application Support/mcp-memory/`
-- Windows: `%LOCALAPPDATA%\mcp-memory\`
+- `/health`
+- `/api/projects`
+- `/api/projects/:id`
+- `/api/projects/:pid/entries`
+- `/api/projects/:pid/tasks`
+- `/api/entries/:eid/context`
+- `/api/entries/:eid/summary`
+- `/api/entries/batch`
+- `/api/facts`
 
----
+For full endpoint and MCP tool reference, see `docs/api.md`.
 
-## Configuración para opencode
+## MCP Client Setup
 
-Para conectar este servidor MCP desde **opencode**, añade la siguiente configuración en tu `opencode.json` (global en `~/.config/opencode/opencode.jsonc` o local en la raíz del proyecto):
+Example `opencode` config:
 
 ```json
 {
@@ -344,9 +121,7 @@ Para conectar este servidor MCP desde **opencode**, añade la siguiente configur
 }
 ```
 
-> Requiere tener paquete instalado globalmente con `npm install -g @romandp/context-memory-mcp`, o usar un wrapper `npx` propio.
-
-Ejemplo sin instalación global permanente:
+Without global install:
 
 ```json
 {
@@ -361,94 +136,36 @@ Ejemplo sin instalación global permanente:
 }
 ```
 
-> **Nota:** Después de guardar los cambios, reinicia opencode para que la configuración surta efecto.
-
-### Herramientas MCP disponibles
-
-Una vez conectado, opencode tendrá acceso a estas herramientas:
-
-| Tool | Descripción |
-|------|-------------|
-| `project-create` | Crear un proyecto nuevo |
-| `project-list` | Listar todos los proyectos |
-| `project-get` | Obtener un proyecto completo con entradas y tareas |
-| `entry-create` | Crear una entrada SDD (plan/design/tasks/general) |
-| `entry-get` | Obtener entradas de un proyecto |
-| `entry-search` | Buscar entradas por texto (FTS5) |
-| `entry-search-global` | Buscar entradas en todos los proyectos |
-| `entry-update` | Actualizar una entrada (title, content, status, section, parent_id) |
-| `entry-delete` | Eliminar una entrada |
-| `entry-add-decision` | Registrar decisión de diseño |
-| `entry-add-relationship` | Relacionar dos entradas |
-| `entry-batch-get` | Obtener varias entradas compactas por ID |
-| `entry-get-context` | Obtener entrada + decisions + relationships |
-| `entry-get-summary` | Obtener summary derivada compacta |
-| `task-create` | Crear una tarea |
-| `task-list` | Listar tareas de un proyecto |
-| `task-update` | Actualizar estado de una tarea |
-| `memory-facts-get` | Consultar facts derivadas |
-| `audit-get` | Consultar historial de cambios de entradas y tareas |
-
----
-
-## Tests y calidad
+## Configuration
 
-```bash
-npm test             # Ejecutar tests (Vitest, 137 tests)
-npm run test:watch   # Modo watch
-npm run lint         # ESLint
-npm run lint:fix     # ESLint con auto-fix
-npm run format       # Prettier (check)
-npm run format:fix   # Prettier (formatear)
-npm run type-check   # TypeScript strict check
-```
-
----
-
-## Migraciones
-
-El schema de la base de datos se gestiona mediante migraciones numeradas en `src/db/migrations/runner.ts`:
-
-| Migración | Descripción |
-|-----------|-------------|
-| `001_initial_schema` | Tablas base: projects, sdd_entries, tasks, classifications |
-| `002_add_fts5` | Búsqueda de texto completo (FTS5) con sincronización automática |
-| `003_add_audit_log` | Auditoría de cambios con triggers en entries y tasks |
-| `004_add_context_tables` | Tablas de contexto enriquecido: file_changes (legacy), design_decisions, entry_relationships |
-| `005_add_compact_memory` | Summaries derivadas y memory facts para views compactas / TOON |
-| `006_add_compact_memory_fts` | Índice FTS5 sobre summaries/keywords para búsqueda compacta |
+Supported environment variables:
 
-Las migraciones se aplican automáticamente al iniciar el servidor.  
-La tabla `_migrations` registra cuáles se han ejecutado.
-
-> Nota: `file_changes` puede seguir existiendo en bases de datos antiguas para preservar historial, pero ya no forma parte del modelo activo ni de la API/MCP pública.
-
----
-
-## Acceso a la base de datos
-
-Primero obtén ruta real:
-
-```bash
-mcp-memory paths
+```env
+HTTP_PORT=3001
+DB_PATH=/path/to/memory.db
+MCP_MEMORY_HOME=/path/to/home
+LOG_PATH=/path/to/server.log
+PID_PATH=/path/to/server.pid
 ```
 
-En Linux/WSL, consulta SQLite así:
+Default storage locations:
 
-```bash
-sqlite3 ~/.local/share/mcp-memory/memory.db "SELECT * FROM projects;"
-```
+- Linux: `~/.local/share/mcp-memory/memory.db` and `~/.local/state/mcp-memory/`
+- macOS: `~/Library/Application Support/mcp-memory/`
+- Windows: `%LOCALAPPDATA%\mcp-memory\`
 
-Para abrirla en DBeaver desde Windows, haz una copia:
+## Development
 
 ```bash
-cp ~/.local/share/mcp-memory/memory.db /tmp/memory-view.db
+npm install
+npm run build
+npm test
 ```
 
-Y abre `\\wsl.localhost\Ubuntu-24.04\tmp\memory-view.db` en DBeaver.
-
-O descárgala vía HTTP:
+## Documentation
 
-```bash
-curl -o memory-view.db http://localhost:3001/api/db/download
-```
+- `docs/api.md`: HTTP endpoints and MCP tools
+- `docs/architecture.md`: runtime architecture and data flow
+- `docs/database.md`: schema and migrations
+- `docs/release.md`: install, packaging, and release process
+- `docs/project-history.md`: project history and milestones

package/docs/api.md

@@ -0,0 +1,104 @@
+# API Reference
+
+## HTTP API
+
+Base URL:
+
+```text
+http://localhost:3001
+```
+
+Swagger UI:
+
+```text
+http://localhost:3001/api-docs
+```
+
+List endpoints support pagination with `page` and `limit`.
+
+### Projects
+
+```text
+GET    /api/projects
+POST   /api/projects
+GET    /api/projects/:id
+PUT    /api/projects/:id
+DELETE /api/projects/:id
+```
+
+### Entries
+
+```text
+GET    /api/projects/:pid/entries
+POST   /api/projects/:pid/entries
+GET    /api/projects/:pid/entries/search?q=...
+GET    /api/projects/:pid/entries/:eid
+PUT    /api/projects/:pid/entries/:eid
+DELETE /api/projects/:pid/entries/:eid
+```
+
+### Global Search
+
+```text
+GET    /api/entries/search?q=...
+```
+
+### Context and Compact Views
+
+```text
+GET    /api/entries/:eid/context
+GET    /api/entries/:eid/summary
+GET    /api/entries/:eid/compact
+GET    /api/entries/batch?ids=e1,e2
+POST   /api/entries/:eid/decisions
+POST   /api/entries/:eid/relationships
+GET    /api/facts
+```
+
+Supported compact-response parameters:
+
+- `view=full|summary|compact`
+- `format=json|toon-r|toon-d`
+- `max_items=<n>`
+- `max_chars=<n>`
+- `cursor=<last_id>`
+
+### Tasks
+
+```text
+GET    /api/projects/:pid/tasks
+POST   /api/projects/:pid/tasks
+PUT    /api/projects/:pid/tasks/:tid
+DELETE /api/projects/:pid/tasks/:tid
+```
+
+### Audit and Utilities
+
+```text
+GET    /api/audit
+POST   /api/classify
+GET    /api/db/download
+GET    /health
+```
+
+## MCP Tools
+
+- `project-create`
+- `project-list`
+- `project-get`
+- `entry-create`
+- `entry-get`
+- `entry-search`
+- `entry-search-global`
+- `entry-update`
+- `entry-delete`
+- `entry-add-decision`
+- `entry-add-relationship`
+- `entry-batch-get`
+- `entry-get-context`
+- `entry-get-summary`
+- `task-create`
+- `task-list`
+- `task-update`
+- `memory-facts-get`
+- `audit-get`

package/docs/architecture.md

@@ -0,0 +1,40 @@
+# Architecture
+
+## Runtime Overview
+
+```text
+MCP client (STDIO)
+        |
+        v
+  MCP server layer
+        |
+        +---- HTTP REST server (Express)
+        |
+        v
+   SQLite database
+```
+
+## Main Components
+
+- `src/index.ts`: process bootstrap, tool dispatch, HTTP startup
+- `src/server/setup.ts`: MCP server and tool definitions
+- `src/http/routes.ts`: REST routes
+- `src/db/schema.ts`: data access and queries
+- `src/memory/*`: summaries, facts, TOON formatting, compact responses
+- `bin/mcp-memory.js`: CLI wrapper and process management
+
+## Retrieval Model
+
+The server stores canonical project data in SQLite and derives compact memory artifacts for LLM-friendly access:
+
+- entry summaries
+- dense summaries
+- memory facts
+- FTS indexes for compact search
+
+Compact retrieval supports:
+
+- reduced payloads
+- response budgets
+- cursor pagination
+- JSON or TOON output

package/docs/database.md

@@ -0,0 +1,49 @@
+# Database
+
+## Core Schema
+
+```text
+projects
+  id, name, description, status
+
+sdd_entries
+  id, project_id, section, title, content, status, parent_id, metadata
+
+tasks
+  id, project_id, sdd_entry_id, title, description, status, priority
+
+design_decisions
+  id, entry_id, decision, rationale, alternatives_considered
+
+entry_relationships
+  id, source_entry_id, target_entry_id, relationship_type
+
+classifications
+  classifiable_type, classifiable_id, tag, confidence
+
+audit_log
+  entity_type, entity_id, action, changes, project_id, timestamp
+```
+
+## Derived Memory Tables
+
+- `entry_summaries`
+- `memory_facts`
+- `fts_entry_summaries`
+- `fts_entries`
+
+## Migrations
+
+- `001_initial_schema`
+- `002_add_fts5`
+- `003_add_audit_log`
+- `004_add_context_tables`
+- `005_add_compact_memory`
+- `006_add_compact_memory_fts`
+
+Migrations are applied automatically on startup and tracked in `_migrations`.
+
+## Notes
+
+- SQLite runs in WAL mode
+- Some legacy databases may still contain `file_changes`, but it is not part of the active public model

package/docs/project-history.md

@@ -0,0 +1,34 @@
+# Project History
+
+## Summary
+
+- SDD-oriented memory model implemented
+- Full CRUD for projects, entries, and tasks
+- HTTP API and MCP tools available
+- Compact memory and TOON responses added
+- CLI packaging and npm release flow added
+- Test suite passing
+
+## Main Milestones
+
+### SDD Memory Refactor
+
+- replaced older schema with project, entry, task, and classification model
+- added supporting context and relationship tables
+
+### Test Coverage
+
+- Vitest coverage for database, tools, and HTTP routes
+
+### CLI Distribution
+
+- installable `mcp-memory` command
+- per-user runtime storage
+- database migration and rebuild commands
+
+### Compact Memory
+
+- derived summaries and facts
+- compact views
+- TOON formats
+- compact FTS search

package/docs/release.md

@@ -0,0 +1,43 @@
+# Release and Installation
+
+## Package
+
+Published package:
+
+```text
+@romandp/context-memory-mcp
+```
+
+## Install
+
+```bash
+npm install -g @romandp/context-memory-mcp
+```
+
+```bash
+npx -y @romandp/context-memory-mcp@latest stdio
+```
+
+## Runtime Paths
+
+Defaults:
+
+- Linux: `~/.local/share/mcp-memory/memory.db` and `~/.local/state/mcp-memory/`
+- macOS: `~/Library/Application Support/mcp-memory/`
+- Windows: `%LOCALAPPDATA%\mcp-memory\`
+
+## Maintenance Commands
+
+```bash
+mcp-memory version
+mcp-memory paths
+mcp-memory status
+mcp-memory rebuild-memory
+mcp-memory migrate-db --from /path/to/old/memory.db
+```
+
+## Release Flow
+
+- Tagging `v*` triggers `.github/workflows/release.yml`
+- Workflow installs dependencies, runs tests, builds, packs, publishes to npm, and creates a GitHub Release
+- Current workflow uses `NPM_TOKEN` for npm publish

package/package.json

@@ -1,13 +1,14 @@
 {
   "name": "@romandp/context-memory-mcp",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "Local SQLite-backed MCP server for AI memory and project context.",
   "main": "dist/index.js",
   "type": "module",
   "files": [
     "dist",
     "bin",
-    "README.md"
+    "README.md",
+    "docs"
   ],
   "bin": {
     "mcp-memory": "./bin/mcp-memory.js"