squish-memory 1.0.2 → 1.2.0

package/packages/mcp/tsconfig.json

@@ -0,0 +1,20 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ESNext",
+    "moduleResolution": "Bundler",
+    "lib": ["ES2022"],
+    "outDir": "./dist",
+    "rootDir": "./src",
+    "strict": true,
+    "esModuleInterop": true,
+    "skipLibCheck": true,
+    "forceConsistentCasingInFileNames": true,
+    "declaration": true,
+    "declarationMap": true,
+    "sourceMap": true,
+    "noEmit": false
+  },
+  "include": ["src/**/*"],
+  "exclude": ["node_modules", "dist"]
+}

package/skills/squish-memory/SKILL.md

@@ -1,114 +1,107 @@
----
-name: squish-memory
-description: Squish - Persistent memory system for AI assistants. Store facts, decisions, and context with semantic search. Works with Claude Code, OpenCode, OpenClaw, and any MCP client.
-version: 1.0.2
-author: michielhdoteth
-tags: [memory, persistence, search, semantic-search, ai-assistant, mcp, cli]
-emoji: brain
----
-
-# Squish Memory v1.0.2
-
-Persistent memory system for AI coding assistants. Store facts, decisions, context, and preferences with semantic search.
-
-## What is Squish?
-
-Squish provides persistent memory for AI agents with:
-- **Hybrid Search**: BM25 + vector embeddings with graph expansion
-- **Memory Types**: observation, fact, decision, context, preference
-- **Graph Associations**: Link related memories together
-- **Core Memory**: Always-visible context sections
-- **Importance Scoring**: Auto-score memories with temporal decay
-- **Consolidation**: Summarize old, low-importance memories
-- **Namespace Support**: Hierarchical organization of memories
-- **Maintenance Scheduling**: Automated cleanup and optimization tasks
-
-## Install
-
-```bash
-npm install -g squish-memory
-# or
-bun add -g squish-memory
-```
-
-## Quick Setup
-
-### Interactive Wizard (Recommended)
-```bash
-squish
-```
-Launches the interactive installer menu.
-
-### Direct Server Start
-```bash
-# MCP server (for Claude Code, etc.)
-squish run mcp
-
-# Web UI only
-squish run web
-```
-
-### MCP Clients
-Configure in your client's MCP settings:
-- **Command**: `squish-mcp`
-- **Args**: (none needed)
-- **Environment**: 
-  - `SQUISH_MODE=local` or `team`
-  - `SQUISH_DATA_DIR=~/.squish`
-
-## Skills Available
-
-This repo contains 4 skills for different use cases:
-
-| Skill | Description | Install |
-|-------|-------------|---------|
-| `squish-memory` | Overview & quick start | `npx skills add michielhdoteth/squish --skill squish-memory` |
-| `squish-mcp` | MCP tools (16 tools) | `npx skills add michielhdoteth/squish --skill squish-mcp` |
-| `squish-cli` | CLI commands | `npx skills add michielhdoteth/squish --skill squish-cli` |
-| `memory-guide` | Usage guide & best practices | `npx skills add michielhdoteth/squish --skill memory-guide` |
-
-## MCP Tools (16)
-
-- squish_remember - Store memory
-- squish_search - Hybrid search
-- squish_recall - Get by ID
-- squish_forget - Delete memory
-- squish_update - Update memory
-- squish_associate - Link memories
-- squish_related - Find related
-- squish_context - Project context
-- squish_observe - Record observation
-- squish_qmd_search - Search docs
-- squish_embed - Generate embeddings
-- squish_health - Check status
-- squish_stats - View stats
-- squish_projects - List projects
-
-## CLI Commands
-
-- `squish` - Interactive wizard/menu
-- `squish run mcp` - Start MCP server
-- `squish run web` - Start Web UI only
-- `squish remember` - Store memory
-- `squish search` - Find memories
-- `squish recall` - Get by ID
-- `squish health` - Check service status
-- `squish stats` - View memory statistics
-- `squish core_memory` - Manage always-visible context
-- `squish add` - Interactive installer (alias)
-- `squish rm` - Remove memory
-
-## Configuration
-
-| Env Variable | Description | Default |
-|--------------|-------------|---------|
-| SQUISH_MODE | local or team | local |
-| SQUISH_DATA_DIR | Data directory | ~/.squish |
-| SQUISH_EMBEDDINGS_PROVIDER | local or openai | local |
-| SQUISH_QMD_ENABLED | Enable QMD search | true |
-
-## More Info
-
-- GitHub: https://github.com/michielhdoteth/squish
-- npm: https://www.npmjs.com/package/squish-memory
-- Docs: https://github.com/michielhdoteth/squish/tree/main/docs
+---
+name: squish-memory
+description: Squish - Persistent memory system for AI assistants. Store facts, decisions, context, and notes with CLI and MCP workflows.
+version: 1.2.0
+author: michielhdoteth
+tags: [memory, persistence, recall, semantic-retrieval, ai-assistant, mcp, cli]
+emoji: brain
+---
+
+# Squish Memory v1.2.0
+
+Persistent memory for AI coding assistants with local SQLite/Postgres storage, hybrid retrieval, and MCP integration.
+
+## Quick Setup
+
+```bash
+npm install -g squish-memory
+
+# Start the MCP server
+squish run mcp
+
+# Or start the web UI only
+squish run web
+```
+
+## Skills Available
+
+| Skill | Description | Install |
+|-------|-------------|---------|
+| `squish-memory` | Overview and quick start | `npx skills add michielhdoteth/squish --skill squish-memory` |
+| `squish-cli` | CLI command reference | `npx skills add michielhdoteth/squish --skill squish-cli` |
+| `memory-guide` | Usage patterns and best practices | `npx skills add michielhdoteth/squish --skill memory-guide` |
+
+## MCP Tools (12) - Unified Write Path
+
+**THE memory write tool for agents: `squish_remember`**
+- Auto-detects memory vs learning routing
+- Supports hot/cold tiers
+- Handles all memory types
+
+| Tool | Purpose | Typical Input |
+|------|---------|---------------|
+| `squish_remember` | **RECOMMENDED** - Unified memory write with auto-detection | `content`, `project`, `tags`, `tier`, `type`, `route` |
+| `squish_recall` | Recall memories by query or fetch by ID | `query`, optional `limit`, `project`, `type`, `place` |
+| `squish_forget` | Delete memory | `memoryId` or bulk filters |
+| `squish_update` | Update memory | `memoryId`, changed fields |
+| `squish_link` | Graph operations | `action`, memory IDs |
+| `squish_context` | Load project context | `project`, `limit` |
+| `squish_health` | Check health | - |
+| `squish_stats` | View stats | `project` |
+| `squish_confidence` | Get or set confidence | `memoryId`, `level` |
+| `squish_pin` | Pin or unpin memory | `memoryId`, `pinned` |
+| `squish_set_passphrase` | Configure encryption | `passphrase` |
+| `squish_rotate_key` | Rotate encryption key | - |
+| `squish_recent` | List recent memories | `project`, `limit` |
+| `squish_stale` | Show stale memories | `project` |
+| `squish_tag` | Bulk tag operations | - |
+| `squish_timeline` | Progressive disclosure | `query`, `depth` |
+
+## CLI Commands
+
+- `squish` - Interactive wizard
+- `squish run mcp` - Start MCP server
+- `squish run web` - Start web UI
+- `squish config` - View or update config
+- `squish remember` - **RECOMMENDED** - Store memory/learning with auto-detection
+- `squish recall` - Search or fetch by ID
+- `squish recent` - Show recent memories
+- `squish update` - Update memory
+- `squish forget` - Delete memory
+- `squish pin` - Pin or unpin memory
+- `squish confidence` - View or set confidence
+- `squish tag` - Bulk tag management
+- `squish stale` - Show stale memories
+- `squish link` - Graph operations
+- `squish context` - Load project context
+- `squish health` - Check service health
+- `squish stats` - View statistics
+
+## Example Workflow
+
+```bash
+# Unified remember - auto-detects routing
+squish remember "Failed because the API returned 404" --project /myproject
+squish remember "User prefers TypeScript over JavaScript" --type preference
+squish remember "Important decision made" --tier hot --type decision
+
+# Recall and retrieve
+squish recall "coding preferences"
+squish context --list-projects
+squish context
+```
+
+## Configuration
+
+| Env Variable | Description | Default |
+|--------------|-------------|---------|
+| `SQUISH_MODE` | local or team | `local` |
+| `SQUISH_DATA_DIR` | Data directory | `.squish` in current workspace |
+| `SQUISH_EMBEDDINGS_PROVIDER` | `local`, `openai`, `ollama`, `google`, `none`, `auto` | `local` |
+| `SQUISH_QMD_ENABLED` | Enable QMD integration | `false` unless set |
+
+## More Info
+
+- GitHub: https://github.com/michielhdoteth/squish
+- npm: https://www.npmjs.com/package/squish-memory
+- Docs: https://github.com/michielhdoteth/squish/tree/main/docs

package/skills/squish-memory/install.sh

@@ -113,11 +113,11 @@ echo ""
 echo "CLI Commands:"
 echo "  squish health              - Check service health"
 echo "  squish remember \"text\"    - Store a memory"
-echo "  squish search \"query\"     - Search memories"
+echo "  squish recall \"query\"     - Recall memories"
 echo "  squish stats               - View statistics"
 echo ""
 echo "MCP Tools (via mcporter):"
-echo "  remember, search, recall, observe"
-echo "  context, health, core_memory, forget"
+echo "  remember, search, recall, learn"
+echo "  context, health, stats, note"
 echo ""
 echo "Documentation: $GITHUB_REPO"

package/skills/squish-memory/{claude-desktop.json → references/claude-desktop.json}

@@ -1,12 +1,12 @@
-{
-  "mcpServers": {
-    "squish": {
-      "command": "npx",
-      "args": ["-y", "squish-memory", "mcp"],
-      "env": {
-        "SQUISH_MODE": "local",
-        "SQUISH_DATA_DIR": "~/.squish"
-      }
-    }
-  }
-}
+{
+  "mcpServers": {
+    "squish": {
+      "command": "npx",
+      "args": ["-y", "squish-memory", "mcp"],
+      "env": {
+        "SQUISH_MODE": "local",
+        "SQUISH_DATA_DIR": "~/.squish"
+      }
+    }
+  }
+}

package/skills/squish-memory/{openclaw.json → references/openclaw.json}

@@ -1,13 +1,13 @@
-{
-  "mcpServers": {
-    "squish": {
-      "command": "npx",
-      "args": ["-y", "squish-memory", "mcp"],
-      "env": {
-        "SQUISH_MODE": "local",
-        "SQUISH_DATA_DIR": "~/.squish"
-      },
-      "transport": "stdio"
-    }
-  }
-}
+{
+  "mcpServers": {
+    "squish": {
+      "command": "npx",
+      "args": ["-y", "squish-memory", "mcp"],
+      "env": {
+        "SQUISH_MODE": "local",
+        "SQUISH_DATA_DIR": "~/.squish"
+      },
+      "transport": "stdio"
+    }
+  }
+}

package/skills/squish-memory/{opencode.json → references/opencode.json}

@@ -1,14 +1,14 @@
-{
-  "$schema": "https://opencode.ai/config.json",
-  "mcp": {
-    "squish": {
-      "type": "local",
-      "command": ["npx", "-y", "squish-memory", "mcp"],
-      "enabled": true,
-      "environment": {
-        "SQUISH_MODE": "local",
-        "SQUISH_DATA_DIR": "~/.squish"
-      }
-    }
-  }
-}
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "squish": {
+      "type": "local",
+      "command": ["npx", "-y", "squish-memory", "mcp"],
+      "enabled": true,
+      "environment": {
+        "SQUISH_MODE": "local",
+        "SQUISH_DATA_DIR": "~/.squish"
+      }
+    }
+  }
+}

package/.claude-plugin/marketplace.json

@@ -1,20 +0,0 @@
-{
-  "name": "squish-memory",
-  "owner": {
-    "name": "michielhdoteth"
-  },
-  "plugins": [
-    {
-      "name": "squish",
-      "version": "0.9.0",
-      "description": "The memory layer for AI agents. Give Claude Code persistent, intelligent memory with trigger detection, contradiction resolution, and hybrid search. Works with any MCP client.",
-      "author": {
-        "name": "michielhdoteth"
-      },
-      "license": "MIT",
-      "homepage": "https://github.com/michielhdoteth/squish",
-      "repository": "https://github.com/michielhdoteth/squish.git",
-      "source": "./"
-    }
-  ]
-}

package/.claude-plugin/plugin.json

@@ -1,32 +0,0 @@
-{
-  "name": "squish",
-  "version": "0.9.0",
-  "description": "The memory layer for AI agents - persistent, intelligent memory with trigger detection, contradiction resolution, and hybrid search",
-  "author": {
-    "name": "michielhdoteth",
-    "url": "https://github.com/michielhdoteth"
-  },
-  "license": "MIT",
-  "homepage": "https://github.com/michielhdoteth/squish#readme",
-  "repository": "https://github.com/michielhdoteth/squish.git",
-  "keywords": [
-    "claude",
-    "claude-code",
-    "claude-plugin",
-    "mcp-server",
-    "memory",
-    "persistent-memory",
-    "core-memory",
-    "context-management",
-    "semantic-search",
-    "full-text-search",
-    "local-first",
-    "privacy",
-    "agent-memory",
-    "slash-commands",
-    "two-tier-architecture",
-    "model-context-protocol"
-  ],
-  "commands": "./commands/",
-  "mcpServers": "./.mcp.json"
-}

package/.env.mcp.example

@@ -1,60 +0,0 @@
-# MCP Server Configuration
-SQUISH_MCP_PORT=8767
-SQUISH_MCP_SERVER_ENABLED=true
-
-# CORS Configuration (comma-separated origins, * wildcard for port)
-# Default: localhost and 127.0.0.1 on any port
-# SQUISH_CORS_ORIGINS=http://localhost:3000,http://127.0.0.1:8080
-
-# Embeddings Provider: local | openai | ollama | google | none | auto
-# Default: local (TF-IDF, no API needed)
-SQUISH_EMBEDDINGS_PROVIDER=local
-
-# Model Selection (optional - uses defaults if not set)
-# SQUISH_OPENAI_EMBEDDING_MODEL=text-embedding-3-small
-# SQUISH_GOOGLE_EMBEDDING_MODEL=gemini-embedding-001
-# SQUISH_OLLAMA_EMBEDDING_MODEL=nomic-embed-text:v1.5
-
-# Google Cloud (for google provider)
-# Get API key from: https://console.cloud.google.com/apis/credentials
-# GOOGLE_CLOUD_PROJECT=your-project-id
-# GOOGLE_CLOUD_LOCATION=us-central1
-# GOOGLE_CLOUD_API_KEY=your-api-key
-# Or use service account:
-# GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
-
-# OpenAI (for openai provider or auto mode)
-# SQUISH_OPENAI_API_KEY=sk-...
-# SQUISH_OPENAI_API_URL=https://api.openai.com/v1/embeddings
-
-# Ollama (for ollama provider or auto mode)
-# SQUISH_OLLAMA_URL=http://localhost:11434
-
-# Embedding performance & reliability
-SQUISH_EMBEDDINGS_TIMEOUT_MS=30000
-SQUISH_EMBEDDINGS_MAX_RETRIES=3
-SQUISH_EMBEDDINGS_RETRY_DELAY_MS=1000
-# Per-provider overrides (optional)
-# SQUISH_OPENAI_TIMEOUT_MS=30000
-# SQUISH_OLLAMA_TIMEOUT_MS=30000
-# SQUISH_GOOGLE_TIMEOUT_MS=30000
-
-# QMD for local markdown search
-SQUISH_QMD_ENABLED=true
-SQUISH_QMD_COLLECTIONS=./qmd-collections
-SQUISH_QMD_FALLBACK=hybrid
-
-# Managed Mode (Squish Cloud - optional)
-# SQUISH_MANAGED_MODE=true
-# SQUISH_MANAGED_API_URL=https://api.squish.dev
-# SQUISH_MANAGED_API_KEY=your-api-key
-
-# Lifecycle & Summarization
-SQUISH_LIFECYCLE_ENABLED=true
-SQUISH_SUMMARIZATION_ENABLED=true
-SQUISH_GOVERNANCE_ENABLED=true
-
-# Core Memory Configuration (v0.9.2+)
-# Increase limits for daily agent usage (default: 16KB total, 4KB per section)
-# SQUISH_CORE_MEMORY_TOTAL_BYTES=16384
-# SQUISH_CORE_MEMORY_SECTION_BYTES=4096

package/.mcp.json

@@ -1,11 +0,0 @@
-{
-  "mcpServers": {
-    "squish": {
-      "command": "node",
-      "args": ["dist/commands/mcp-server.js"],
-      "env": {
-        "NODE_ENV": "production"
-      }
-    }
-  }
-}

package/QUICK-START.md

@@ -1,71 +0,0 @@
-# Squish - Quick Start
-
-Universal two-tier memory for any AI agent.
-
-## Start in 30 seconds
-
-```bash
-# 1. Install
-bun install
-
-# 2. Build
-bun run build
-
-# 3. Start MCP server (works with Claude Code, OpenClaw, or any MCP client)
-bun run mcp
-
-# ✓ Running on http://localhost:8767
-```
-
-## Test it
-
-```bash
-# Health check
-curl http://localhost:8767/health
-
-# Expected output:
-{
-  "status": "ok",
-  "server": "squish-mcp",
-   "version": "0.9.2",
-  "tools": 6
-}
-```
-
-## Use it
-
-```bash
-# Search memories
-curl -X POST http://localhost:8767/tools/call \
-  -H "Content-Type: application/json" \
-  -d '{
-    "name": "squish_search",
-    "arguments": {
-      "query": "authentication"
-    }
-  }'
-```
-
-## Configure
-
-Copy `.env.mcp.example` to `.env` and customize:
-
-```bash
-cp .env.mcp.example .env
-```
-
-Key settings:
-
-- **SQUISH_MCP_PORT**: Server port (default: 8767)
-- **SQUISH_MCP_SERVER_ENABLED**: Enable/disable MCP server (default: true)
-- **SQUISH_EMBEDDINGS_PROVIDER**: local|openai|ollama|google-multimodal|hybrid (default: local)
-- **SQUISH_MULTIMODAL_EMBEDDINGS_ENABLED**: Enable Google multimodal (default: false)
-- **SQUISH_QMD_ENABLED**: Enable markdown search (true/false)
-- **SQUISH_CORE_MEMORY_TOTAL_BYTES**: Total core memory limit in bytes (default: 16384)
-- **SQUISH_CORE_MEMORY_SECTION_BYTES**: Per-section limit in bytes (default: 4096)
-- **SQUISH_EMBEDDINGS_TIMEOUT_MS**: Timeout for embedding API calls (default: 30000)
-- **SQUISH_EMBEDDINGS_MAX_RETRIES**: Max retries for failed embedding calls (default: 3)
-
-## Full Docs
-
-See [docs/MCP-SERVER.md](docs/MCP-SERVER.md)

package/bin/squish-add.mjs

@@ -1,32 +0,0 @@
-#!/usr/bin/env node
-// Wrapper for squish commands - launches interactive mode when no args
-import { spawn } from 'child_process';
-import { fileURLToPath } from 'url';
-import { dirname, join } from 'path';
-
-const __filename = fileURLToPath(import.meta.url);
-const __dirname = dirname(__filename);
-
-const args = process.argv.slice(2);
-
-// If no args, launch interactive installer
-if (args.length === 0) {
-  const interactivePath = join(__dirname, '..', 'scripts', 'install-interactive.mjs');
-  const child = spawn('node', [interactivePath], {
-    stdio: 'inherit',
-    shell: false
-  });
-  child.on('exit', (code) => {
-    process.exit(code ?? 0);
-  });
-} else {
-  // Pass through to install-plugin.mjs
-  const scriptPath = join(__dirname, '..', 'scripts', 'install-plugin.mjs');
-  const child = spawn('node', [scriptPath, ...args], {
-    stdio: 'inherit',
-    shell: false
-  });
-  child.on('exit', (code) => {
-    process.exit(code ?? 0);
-  });
-}

package/bin/squish-rm.mjs

@@ -1,21 +0,0 @@
-#!/usr/bin/env node
-// Wrapper for squish remove command
-import { spawn } from 'child_process';
-import { fileURLToPath } from 'url';
-import { dirname, join } from 'path';
-
-const __filename = fileURLToPath(import.meta.url);
-const __dirname = dirname(__filename);
-
-const args = process.argv.slice(2);
-const scriptPath = join(__dirname, 'scripts', 'install-plugin.mjs');
-
-// Add --uninstall flag
-const child = spawn('node', [scriptPath, '--uninstall', ...args], {
-  stdio: 'inherit',
-  shell: false
-});
-
-child.on('exit', (code) => {
-  process.exit(code ?? 0);
-});

package/commands/context-paging.md

@@ -1,51 +0,0 @@
----
-description: Manage working set (load, evict, or view loaded memories)
----
-
-# Context Paging Management
-
-Manage your working set of memories for context tracking. Note: Claude manages its own context and token usage. These tools track which memories you're actively working with.
-
-## Actions
-
-### load
-Add a memory to your working set for active tracking.
-- Marks memory as currently relevant
-- Helps organize which memories are in your active consideration set
-- Does not affect Claude's actual context usage
-
-### evict
-Remove a memory from your working set.
-- Removes from your active tracking set
-- Memory is still stored and searchable
-- Useful when memory is no longer actively relevant
-
-### view
-View all memories currently in your working set.
-- Shows all memories you're actively tracking
-- Displays session ID and timestamp loaded
-- Lists memory content and metadata
-
-## Usage Examples
-
-Load a memory to working set:
-```
-/context-paging action=load sessionId=sess-123 memoryId=mem-uuid-456
-```
-
-Evict memory from working set:
-```
-/context-paging action=evict sessionId=sess-123 memoryId=mem-uuid-456
-```
-
-View all loaded memories in session:
-```
-/context-paging action=view sessionId=sess-123
-```
-
-## Important Notes
-
-- Working set is per-session - each Claude Code session has its own working set
-- Claude Code manages its own context window automatically
-- These tools are for your own organizational purposes
-- Memory removal from working set does not delete the memory permanently