squish-memory 1.1.5 → 1.2.1

package/.env.example

@@ -37,14 +37,21 @@ SQUISH_MCP_SERVER_ENABLED=false
 # auto  = Try OpenAI > Ollama > Google > local fallback
 SQUISH_EMBEDDINGS_PROVIDER=local
 
-# OpenAI (required for cloud embeddings)
-# SQUISH_OPENAI_API_KEY=sk-...
-# SQUISH_OPENAI_API_URL=https://api.openai.com/v1/embeddings
-# SQUISH_OPENAI_EMBEDDING_MODEL=text-embedding-3-small
-
-# Ollama
-# SQUISH_OLLAMA_URL=http://localhost:11434
-# SQUISH_OLLAMA_EMBEDDING_MODEL=nomic-embed-text:v1.5
+# OpenAI (example: set both key and model)
+# SQUISH_OPENAI_API_KEY=sk-...
+# SQUISH_OPENAI_API_URL=https://api.openai.com/v1/embeddings
+# SQUISH_OPENAI_EMBEDDING_MODEL=<openai-embedding-model>
+
+# Ollama (example: set URL and model served by your local Ollama)
+# SQUISH_OLLAMA_URL=http://localhost:11434
+# SQUISH_OLLAMA_EMBEDDING_MODEL=<ollama-embedding-model>
+
+# LM Studio (example: set URL and model loaded in LM Studio)
+# SQUISH_LM_STUDIO_URL=http://localhost:1234
+# SQUISH_LM_STUDIO_EMBEDDING_MODEL=<lm-studio-embedding-model>
+
+# Transformers.js local (example: set a Hugging Face ONNX embedding model)
+# SQUISH_LOCAL_MODEL=<huggingface-onnx-embedding-model>
 
 # Google Cloud
 # GOOGLE_CLOUD_PROJECT=your-project-id
@@ -52,12 +59,19 @@ SQUISH_EMBEDDINGS_PROVIDER=local
 # GOOGLE_CLOUD_API_KEY=your-api-key
 # Or use service account:
 # GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
-# SQUISH_GOOGLE_EMBEDDING_MODEL=gemini-embedding-001
+# SQUISH_GOOGLE_EMBEDDING_MODEL=<google-embedding-model>
 
-# Embedding reliability
-SQUISH_EMBEDDINGS_TIMEOUT_MS=30000
-SQUISH_EMBEDDINGS_MAX_RETRIES=3
-SQUISH_EMBEDDINGS_RETRY_DELAY_MS=1000
+# Embedding reliability
+SQUISH_EMBEDDINGS_TIMEOUT_MS=30000
+SQUISH_EMBEDDINGS_MAX_RETRIES=3
+SQUISH_EMBEDDINGS_RETRY_DELAY_MS=1000
+
+# Optional LLM graph enrichment (off by default for launch-safe local mode)
+SQUISH_LLM_ENABLED=false
+# SQUISH_LLM_PROVIDER=openai
+# SQUISH_LLM_ENDPOINT=http://localhost:1234
+# SQUISH_LLM_EXTRACTION_MODEL=<llm-extraction-model>
+# SQUISH_LLM_REASONING_MODEL=<llm-reasoning-model>
 
 # =============================================================================
 # QMD (local markdown search)
@@ -102,9 +116,11 @@ SQUISH_GOVERNANCE_ENABLED=true
 # =============================================================================
 # GRAPH BOOST (v1.1.0+)
 # =============================================================================
-# Scoring weight for graph boost in recall (default: 1.5)
-# Higher = more influence from memory associations
-# SQUISH_WEIGHT_GRAPH_BOOST=1.5
+# Scoring weight for graph boost in recall (default: 1.5)
+# Higher = more influence from memory associations
+# SQUISH_WEIGHT_GRAPH_BOOST=1.5
+# SQUISH_GRAPH_AUTO_BUILD=true
+# SQUISH_GRAPH_EXTRACTION_METHOD=auto
 
 # =============================================================================
 # MANAGED MODE (Squish Cloud)

package/CHANGELOG.md

@@ -2,6 +2,153 @@
 
 All notable changes to Squish will be documented in this file.
 
+## [1.2.0] - 2026-04-19
+
+### Added - QMD Integration
+
+- **Dependency**: Add @tobilu/qmd@^2.1.0 as npm package
+- **Wrapper**: New core/search/qmd-wrapper.ts with search, query, vsearch, embed functions
+- **Usage as library**: Uses SDK library API (not CLI MCP wrapper)
+
+### Added - Persistent Hot Cache (Karpathy-Style)
+
+- **Persistent storage**: .squish/hot-cache.json survives restart
+- **Deduplication**: SHA256 hash-based content deduplication
+- **Stale prevention**: Entries flagged after 7 days without reference
+- **Auto-clean**: Entries older than 14 days removed on load
+- **Session integration**: Loaded on auto-load, saved on session end
+- **Size limit**: ~500 words max with trim-to-recent policy
+
+### Changed - Memory Runtime
+
+- **Simplified lifecycle**: Removed deprecated warm tier (hot/cold only now)
+- **Updated schemas**: Both SQLite and Postgres tier columns now hot|cold
+- **Updated lifecycle**: TIER_THRESHOLDS only has hot and cold thresholds
+- **Updated scorer**: Removed warm scoring bonuses
+- **Updated CLI**: Descriptions reflect simplified tier system
+
+## [1.2.0] - 2026-04-19
+
+### Added - Belief System (Derived Semantic Layer)
+
+- **Belief extraction from memories**: Automatically extracts beliefs from durable memories
+- **Belief types**: decision, preference, constraint, failure_cause, state_change, dispute
+- **Belief decay engine**: 30-day half-life, disputed beliefs decay 1.5x faster
+- **Source tracking**: Each belief tracks source memories for provenance
+- **Integration**: Beliefs extracted on both explicit and auto-capture durable writes
+
+### Changed - Schema Integration
+
+- **Canonical schema**: Belief tables now in SQLite and Postgres drizzle schemas
+- **Bootstrap migration**: Belief table creation included in db/bootstrap.ts
+- **Decay fields**: last_confirmed_at, belief_decay_rate, source_count, confidence
+
+### Changed - Scheduler Resilience
+
+- **Catch-up logic**: On scheduler init, checks for missed jobs and runs catch-up
+- **Grace period**: Jobs with elapsed > 1.5x expected interval trigger catch-up
+- **Handles sleep/wake**: No more missed jobs when machine is off/sleeping
+
+### Added - Query Functions
+
+- **getAllBeliefs()**: Query all beliefs for a project with filtering
+- **searchBeliefs()**: Search beliefs by statement content
+- **Provenance display**: Memory inspection shows source count, evidence preview
+
+### Fixed - Trust Surfaces
+
+- **Beliefs in context**: Context state now includes extracted beliefs
+- **Beliefs in trust report**: Trust report surfaces belief summaries
+- **Enhanced inspect**: Memory explain shows belief provenance
+
+### Added - Signal-Aware Memory Runtime
+- **Signal distillation engine**: Captured events are now classified as `discard`, `session-only`, `durable-distilled`, or `durable-raw+distilled`
+- **Session working set**: Active files, recent commands, failures, hypotheses, active places, and graph cues are compacted for wake-up continuity
+- **Raw fallback snapshots**: Nuance-sensitive durable events can keep linked raw artifacts without polluting normal retrieval
+- **Incremental place + graph routing**: Durable writes now feed place assignment and graph enrichment as part of the same runtime loop
+
+### Added - Trust-Oriented CLI And MCP Surfaces
+- **`squish health`**: New health command for project scope, subsystem status, and next-step guidance
+- **`squish inspect` / `squish_inspect`**: Explanation path for why a memory exists, whether raw fallback exists, and how legacy records should be read
+- **Structured MCP trust outputs**: `squish_context`, `squish_health`, and `squish_stats` now expose current project, runtime state, and recovery guidance in a predictable shape
+
+### Changed - Project Context And Wake-Up Semantics
+- **Canonical current project**: Normal context output now prefers the real workspace path and suppresses legacy placeholder project noise
+- **Wake-up order**: Session working set and active runtime state are surfaced before broad durable recall
+- **Stats semantics**: Durable memory totals are now clearly separated from capture-era signal telemetry
+
+### Fixed - Trust Consistency
+- **`squish doctor` coherence**: Doctor now reads as one combined trust-and-diagnostics command instead of two stacked reports
+- **Legacy inspection**: Older memories now report `legacy-durable` instead of `unknown`, with explicit provenance text
+- **Release verification path**: Added `bun run verify:mcp` for packaged MCP health verification
+
+### Docs
+- README, CLI reference, install quickstart, architecture notes, and release notes updated to match the automatic memory runtime and current command surface
+
+### Added - Unified Write Path (squish_remember)
+- **New unified memory tool**: `squish_remember` auto-detects memory vs learning routing
+- **Auto-type detection**: Detects `success`, `failure`, `fix`, `observation`, `decision`, `preference`, `note`, `reflection`
+- **Hot/cold tier support**: Automatic routing to appropriate storage tier
+- **Code pattern detection**: Recognizes code changes, errors, fixes in agent outputs
+- **Place and pin flags**: Support for place-based and pinned memories
+
+### Added - Graph Auto-Update
+- **Auto-update graph on remember**: Knowledge graph updated automatically when memories are stored
+- **Auto-update graph on link**: Graph updated when associations are created between memories
+- **Integration**: Graph module now integrated into main write path (was previously standalone)
+
+### Added - New CLI Commands
+- **`squish run web`**: Start web UI on port 37777
+- **`squish migrate`**: Unify multiple .squish folders into single location
+- **Full CLI suite**: 10+ commands (remember, search, recall, recent, context, stats, forget, link, stale, clean)
+
+### Added - Memory Decay System
+- **Enabled by default**: Memory decay now runs automatically (was broken in 1.1.x)
+- **Configurable intervals**: Sector decay intervals now configurable via env vars
+- **Auto-clean enabled**: Automatic cleanup of expired memories (was dry-run only)
+- **Fixed**: Removed dead code and consolidated decay logic
+
+### Added - add-mcp Integration
+- **Universal MCP installation**: Replaced manual config writing with `add-mcp`
+- **Multi-client support**: Works with Zed, Goose, Gemini CLI, Claude Code, Cursor, OpenCode, Windsurf
+- **Global flag**: `--global` flag for user-level installation
+- **Simplified installer**: `bin/install-interactive.mjs` rewritten to use add-mcp
+
+### Fixed - v1.1.5 Issues
+- **bin/ tracked in git**: Removed from .gitignore, now included in npm package
+- **CLI commands work**: Fixed npm publishing issue that broke CLI
+- **Type safety**: Fixed LSP errors, obsidian-vault import path, cron types
+
+### Refactored - Code Cleanup
+- **Consolidated estimateTokens**: Moved from 5 files to single source (`core/context/context-window.ts`)
+- **Consolidated MemoryDiff**: Moved from 2 files to `core/snapshots/comparison.ts`
+- **Removed stubs**: Deleted `core/autosave.ts`, `core/external-folder/`, `core/config.js`
+- **Implemented placeholders**: `readFromVault()` in obsidian-vault now functional
+
+### Refactored - Legacy Removal
+- **Removed hooks/**: No longer needed (add-mcp handles installation)
+- **Removed install scripts**: `bin/install-claude-code.sh`, `bin/install-opencode.sh` deleted
+- **Removed config/hooks/**: Hook config files for Claude Code, Cursor, OpenCode, Windsurf deleted
+- **Removed unused config files**: 10+ unused JSON configs deleted (mcp.json, plugin.json, etc.)
+- **Removed core/adapters/scripts/**: Legacy adapter install scripts deleted
+
+### Refactored - MD Docs Cleanup
+- **Removed from core/commands/**: 11 .md files deleted (docs belong in README/CHANGELOG)
+- **Updated package.json**: Removed explicit .md file entries from files array
+
+## [1.1.6] - 2026-04-13
+
+### Fixed - MCP HTTP Transport Migration
+- **Migrated MCP HTTP from SSE to Streamable HTTP**: The deprecated SSE transport wasn't working. Now uses `StreamableHTTPServerTransport` from the MCP SDK with proper session handling.
+
+### Fixed - Database Schema Migrations
+- Added `session_summaries` table
+- Added `memory_associations.metadata` column
+- Added `memories_fts.summary` column (recreates FTS table)
+
+### Added - QMD Dependency
+- Added `qmd` as bundled dependency
+
 ## [1.1.5] - 2026-04-11
 
 ### Added - Release Prep Enhancements

package/README.md

@@ -1,11 +1,11 @@
-# Squish - Universal Memory for AI Agents
+# Squish - Memory Runtime for AI Agents
 
 [![npm version](https://img.shields.io/npm/v/squish-memory)](https://www.npmjs.com/package/squish-memory)
 [![npm downloads](https://img.shields.io/npm/dm/squish-memory)](https://www.npmjs.com/package/squish-memory)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 [![TypeScript](https://img.shields.io/badge/TypeScript-5.7-blue.svg)](https://www.typescriptlang.org/)
 
-**Give any AI agent persistent, intelligent memory.** Without memory, agents forget everything between sessions. With Squish, they learn and adapt over time.
+**Your agent forgets. Squish fixes that.** It auto-captures useful context, derives durable beliefs like decisions and constraints, and restores that context through CLI, MCP, and a local web UI.
 
 > Squish does not have a crypto token, has no token launch planned, and nobody is authorized to launch one on behalf of the project.
 
@@ -13,7 +13,7 @@
 bun add squish-memory
 ```
 
-## Why Memory Matters
+## Why Squish
 
 | Without Squish | With Squish |
 |----------------|-------------|
@@ -22,83 +22,93 @@ bun add squish-memory
 | No project awareness | Builds understanding over time |
 | Can't track preferences | Adapts to your style |
 
-## How It Works
+## What It Does
 
-**Two-tier architecture** for optimal speed and durability:
+Squish is a forward-only memory runtime for agents:
+
+- **Auto-capture** stores durable signal without relying on the model to remember to save it.
+- **Belief derivation** turns memories into decisions, constraints, and preferences that can change future behavior.
+- **Context restore** gives a restarted agent the relevant state instead of a cold start.
+- **Inspection surfaces** let you see what was stored, what belief was derived, and why it was injected back.
+
+Under the hood, Squish uses a hybrid memory pipeline for signal quality, persistence, and wake-up continuity:
 
 ```
-User Action ──► Trigger Detection ──► Write Gate ──► Short-term (QMD)
-                                                             │
-                                                         Long-term (SQLite/PG)
-                                                             │
-                                                         Hybrid Retrieval
-                                                             │
-                                                      Agent Context
+User Action ──► Signal Distillation ──► Write Gate ──► Session Working Set
+                                                          │
+                                                          ├─ Durable Distilled Memory
+                                                          ├─ Raw Fallback Snapshot
+                                                          └─ SQLite/Postgres + Hybrid Retrieval
 ```
 
-- **Short-term (QMD)**: Lightning-fast file-based search. Instant recall for recent context.
-- **Long-term (SQLite/PG)**: Durable storage. SQLite for local, PostgreSQL for teams.
+- **Signal distillation**: Squish suppresses noisy output, keeps session-only context local, and only promotes durable signal.
+- **Session working set**: Active files, recent commands, failures, hypotheses, active places, and small graph cues are compacted for the next wake-up.
+- **Places**: Durable memories are routed into spatial buckets like `WIP`, `Sandbox`, `Board`, and `Ref` for segmented retrieval.
+- **Graph enrichment**: Durable memories strengthen entity and relationship structure used by retrieval scoring.
+- **Durable memory**: Stable facts, corrections, decisions, and fixes are stored for long-term retrieval.
+- **Raw fallback**: Nuance-sensitive output can keep an internal raw artifact for inspection without polluting normal context.
 
 ## Quick Start
 
 ### Install with add-mcp (Recommended)
-One command installs to Claude Code, OpenCode, Cursor, VS Code, Codex, and more:
+One command installs Squish into Claude Code, OpenCode, Cursor, VS Code, Codex, and other MCP-capable clients:
 
 ```bash
 npx add-mcp squish-memory
 ```
 
-Or traditional npm install:
+Or install the package directly:
 
 ```bash
 bun add squish-memory
 ```
 
-```bash
-# Store a memory
-squish remember "User prefers TypeScript over JavaScript"
-
-# Save a quick note
-squish note "Revisit caching strategy after launch"
+New installs should work on first run with the current schema. If you are upgrading an older local install, use `squish doctor --migrate` to repair it forward.
 
-# Record an observation
-squish learn observation "Updated auth flow" --action edit
+Most memory behavior is automatic once Squish is installed. The CLI remains available for explicit saves, inspection, diagnostics, and one-command demos:
 
-# Record a fix or lesson learned
-squish learn fix "Patched auth middleware regression"
+```bash
+# Zero-touch demo: show current project context and derived beliefs
+squish context --json
 
-# Search memories
-squish search "coding preferences"
+# Explicit save when you want to pin something intentionally
+squish remember "We chose PostgreSQL for team mode" --type decision
 
-# List projects, then inspect relevant context
-squish context --list-projects
-squish context
+# Inspect why a memory exists and which beliefs it supports
+squish inspect <memory-id> --json
 
-# Get relevant context or fetch by ID
-squish recall "user preferences"
+# Repair an older install forward if local schema drifted
+squish doctor --json --migrate
 ```
 
-Or use as a plugin:
+Or use the other shipped surfaces directly:
 
 ```bash
-# Install for Claude Code
-npx squish-memory install-plugin --client=claude-code
+# MCP health check / manual startup surface
+squish-mcp --health
 
-# Install for OpenCode
-npx squish-memory install-plugin --client=opencode
+# Local web UI
+squish run web
 ```
 
 ## Features
 
 ### Memory Intelligence
 - Auto-detects "remember this", "important", corrections
+- Distills noisy tool output before durable writes
+- Splits events into discarded, session-only, durable, and durable-with-raw-fallback paths
 - Handles contradictions when facts change
 - Temporal facts with expiration ("until January")
 - Confidence scoring for each memory
-- **Tier lifecycle**: hot/warm/cold memory tiers with automatic decay
+- **Memory Runtime**: Hot/cold memory lifecycle with automatic decay (hot=active, cold=archived)
 - **Graph-boosted retrieval**: associations between memories boost relevance
+- **Belief System**: Derived semantic layer - decisions, preferences, constraints extracted from memories
+- **Persistent Hot Cache**: Karpathy-style wiki layer that survives restarts (not just session)
+- **Scheduler Resilience**: Jobs catch up after machine sleep/wake - no missed maintenance
 
 ### Retrieval Quality
+- Session wake-up uses compacted working-set context before broad recall
+- Place context remains attached to retrieved memories and can shape context selection
 - Hybrid search: semantic + keyword (BM25) with Reciprocal Rank Fusion
 - Multi-factor ranking: relevance, recency, importance, graph-boost
 - LLM-powered context extraction with Ollama (local)
@@ -106,46 +116,79 @@ npx squish-memory install-plugin --client=opencode
 
 ### Security & Encryption
 - **Client-side encryption**: AES-256-GCM encryption for sensitive memories
-- **Passphrase management**: `squish_set_passphrase` and `squish_rotate_key` MCP tools
-- Optional encryption via `SQUISH_ENCRYPTION_PASSPHRASE` env var
+- **Passphrase management**: Via `SQUISH_ENCRYPTION_PASSPHRASE` env var (not exposed via MCP)
+- Encryption passphrase configured in `.env` file in data directory
 
 ### Universal Compatibility
-- **CLI**: `squish config`, `squish remember`, `squish note`, `squish learn`, `squish search`, `squish context`, `squish stats`
+- **CLI**: `squish remember`, `squish recall`, `squish inspect`, `squish context`, `squish stats`, `squish doctor`
 - **MCP Server**: Works with Claude Code, OpenCode, Cursor, VS Code, OpenClaw
-- **HTTP API**: REST API + WebSocket for any agent
+- **Web UI**: Inspect memories, projects, and recent observations locally
 - **SQLite**: Local, zero-config
 - **PostgreSQL**: Team mode with Supabase/pgvector
-- **QMD Integration**: Native .md file search via QMD
-
-### Current MCP Tools
-- `squish_remember`, `squish_search`, `squish_recall`, `squish_forget`, `squish_update`
-- `squish_link`, `squish_context`, `squish_learn`, `squish_health`, `squish_stats`
-- `squish_confidence`, `squish_pin`, `squish_set_passphrase`, `squish_rotate_key`
-- `squish_recent`, `squish_stale`, `squish_note`, `squish_tag`
+- **QMD Integration**: Native .md file search via @tobilu/qmd npm package
+
+### Current MCP Tools (12 tools)
+- `squish_timeline` - 3-layer progressive disclosure
+- `squish_remember` - Store memory or learning (auto-detects type)
+- `squish_recall` - Recall memories by query or retrieve memory by ID
+- `squish_forget` - Delete memory by ID or bulk delete
+- `squish_link` - Manage memory associations
+- `squish_context` - Get project context
+- `squish_health` - Check system health
+- `squish_stats` - Get memory statistics
+- `squish_inspect` - Inspect memory retention
+- `squish_pin` - Pin/unpin memory
+- `squish_recent` - Get recent memories
+- `squish_stale` - Show stale memories
 
 ## Benchmark Results
 
-Real tests using [LoCoMo](https://github.com/snap-research/locomo) benchmark (22 questions):
+Real tests using academic benchmarks with both configurations:
+
+### LoCoMo (1540 questions)
+| Configuration | Score | Notes |
+|---------------|-------|-------|
+| **Without LLM** | **65.19%** | Vector-only |
+| With LLM | 67.34% | Vector + LLM extraction |
+| Mem0 | 66.88% | Requires LLM |
+| TrueMemory | 91.5% | Uses LLM |
+
+**Squish achieves competitive scores WITHOUT requiring LLM tokens.**
 
+#### LoCoMo Breakdown (No LLM)
+| Category | Score |
+|----------|-------|
+| Single-hop | 85.14% |
+| Multi-hop | 47.78% |
+| Temporal | 95.24% |
+| Open-domain | 98.00% |
+| Common-sense | 1.79% |
+
+### LongMemEval (100 questions)
+| Configuration | Score | Breakdown |
+|---------------|-------|-----------|
+| **Without LLM** | **67.00%** | Temporal: 85%, Multi-session: 40% |
+| With LLM | 67.00% | Temporal: 85%, Multi-session: 40% |
+
+Note: LLM extraction helps during storage but scoring uses retrieved context keyword matching.
+
+### Performance
 | Metric | Result |
 |--------|--------|
-| **LoCoMo Score** | **77%** |
 | Embedding Latency | 1-5ms |
 | API Latency | 1-20ms |
 | Max Throughput | 943 ops/sec |
 | Package Size | **283 KB** |
 
-### vs Cloud Solutions
-
-| | Squish | Cloud Memory |
-|--|--------|-------------|
-| **Cost** | $0 | API fees |
-| **Local-first** | Yes | No |
-| **Setup** | 1 command | 3+ steps |
-| **API keys** | None | Required |
-| **LoCoMo** | 77% | 75-81% |
+### Local Runtime Characteristics
 
-Squish matches cloud solutions on accuracy while running 100% locally with zero API costs.
+| Characteristic | Result |
+|----------------|--------|
+| Default Cost | $0 local runtime |
+| Local-first | Yes |
+| Setup | 1 command |
+| API keys | Not required for the default path |
+| Session continuity | Built in |
 
 ## Supported Clients
 
@@ -180,9 +223,16 @@ DATABASE_URL=postgresql://user:pass@host/db
 - **QMD (Files)**: BM25 + vectors for fast recall
 - **SQLite/PostgreSQL**: ACID-compliant persistent storage
 
+### Runtime Pipeline
+- **Signal engine**: classifies captured events as discard, session-only, durable-distilled, or durable-with-raw-fallback
+- **Session working set**: persists active working context, active places, and small graph cues between sessions
+- **Places**: spatially segment durable memory for retrieval and wake-up continuity
+- **Graph**: incrementally enriches durable memories so graph boost applies to cleaner signal
+- **Inspection path**: lets you inspect why a memory was retained and whether a raw fallback artifact exists
+
 ### Interfaces
-- **MCP**: Native agent integration
-- **HTTP**: REST + WebSocket
+- **packages/mcp**: Native agent integration via Model Context Protocol
+- **HTTP**: Streamable HTTP server
 - **CLI**: Shell and scripts
 
 ### Memory Lifecycle
@@ -202,24 +252,16 @@ bun run verify:mcp
 ## Troubleshooting
 
 ```bash
-# Reset local database
-rm -rf .squish/squish.db
+# Repair an older local install forward
+squish doctor --migrate
 
-# Verify MCP setup
-bun run verify:mcp
+# Zero-touch runtime check
+squish context --json
 
-# Check health
-squish health
+# MCP health check
+squish-mcp --health
 ```
 
-## CLI Command Families
-
-- Setup/runtime: `squish`, `squish config`, `squish install`, `squish run mcp`, `squish run web`
-- Capture/retrieval: `squish remember`, `squish note`, `squish learn`, `squish search`, `squish recall`, `squish recent`
-- Memory management: `squish update`, `squish forget`, `squish pin`, `squish confidence`, `squish tag`, `squish stale`, `squish link`
-- Context/project discovery: `squish context --list-projects`, `squish context`
-- System: `squish health`, `squish stats`
-
 ## License
 
 MIT License. See [LICENSE](LICENSE).