claude-memory-store 0.1.0 → 0.1.1

package/README.md

@@ -81,7 +81,7 @@ The full memory dump is only loaded when you explicitly ask for it (`?` at the s
 
 ### Prerequisites
 - Node.js 18+
-- MongoDB Atlas account ([free tier works](https://cloud.mongodb.com/register) — no credit card required)
+- MongoDB Atlas M10+ cluster ([create account](https://cloud.mongodb.com/register)) — required for vector search
 
 ### Install
 
@@ -93,7 +93,7 @@ That single command:
 1. Prompts for your Atlas connection string (auto-detects from `CLAUDE_MEMORY_ATLAS_URI` env var or `.env` file)
 2. Tests the connection
 3. Creates the `claude_memory` database and indexes
-4. Sets up Atlas vector search index (M10+ only, skipped gracefully on free tier)
+4. Sets up Atlas vector search index (requires M10+)
 5. Installs Claude Code hooks (startup, pre-tool-use, post-tool-use, stop)
 6. Registers the MCP server (recall, remember, threads, stats tools)
 7. Installs the skill file (`/recall`, `/remember`, `/threads`, `/stats`)
@@ -135,7 +135,7 @@ claude-memory clear --all             # Clear all project memory
 
 | Tool | Description |
 |------|-------------|
-| `recall(topic, limit?)` | Search memory via semantic + keyword search |
+| `recall(topic, limit?)` | Search memory via semantic vector search |
 | `remember(text, category?)` | Store with auto-classification (avoid/pattern/architecture/preference) |
 | `memory_threads()` | List open threads with status and age |
 | `memory_stats()` | Full memory statistics |
@@ -270,27 +270,26 @@ You can also set `CLAUDE_MEMORY_ATLAS_URI` as an environment variable or in a `.
 
 ---
 
-## Atlas Tier Comparison
+## Why Atlas M10+?
 
-Claude Memory works on all Atlas tiers. The difference is search precision:
+Claude Memory requires **MongoDB Atlas M10+** ($57/month) for vector search. This is a deliberate choice — not a limitation.
 
-| Feature | Free Tier (M0) | M10+ |
-|---------|---------------|------|
-| Store/recall memory | Yes | Yes |
-| Hooks (startup, pre/post-tool-use, stop) | Yes | Yes |
-| MCP tools (recall, remember, threads, stats) | Yes | Yes |
-| Keyword search (regex matching) | Yes | Yes |
-| TTL with renewal | Yes | Yes |
-| Verified commands with aliases | Yes | Yes |
-| Open threads | Yes | Yes |
-| **Semantic vector search** | No | **Yes** |
-| **Atlas Automated Embedding (Voyage AI)** | No | **Yes** |
+We tested keyword/regex search (which works on the free tier) against semantic vector search with real developer queries. The results:
 
-**Free tier** uses keyword search — your query terms need to match the words in stored memories. Works well when you search with the same terms you used when storing.
+| Query Type | Keyword Search | Vector Search |
+|------------|---------------|---------------|
+| Exact matches ("prod-cluster-east") | 80% | 80% |
+| Partial matches ("staging database") | 100% | 100% |
+| Synonym queries ("how to undo a release") | **0%** | **60%** |
+| Conceptual queries ("security best practices") | **0%** | **80%** |
+| Natural questions ("what region is our database in") | **0%** | **100%** |
+| **Overall** | **36%** | **84%** |
 
-**M10+** adds semantic vector search via Voyage AI auto-embedding. Atlas automatically generates and maintains vector embeddings for every document. This means "what's our production database?" finds "MongoDB Atlas prod-cluster-east with database api_prod" even though the words don't overlap.
+Keyword search fails completely when you phrase things differently from how they were stored. That's how humans talk. A memory store that can't understand "how do I deploy" when you stored "docker build → ECR push → webhook" is useless.
 
-**Recommendation:** Start with the free tier. It's fully functional. Upgrade to M10+ if you find keyword search isn't precise enough for your workflow.
+**$57/month is cheaper than re-explaining your deploy process every session.**
+
+Vector embeddings are generated locally via [all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) (384 dimensions) — no external API keys, no per-query fees. Atlas just stores and searches the vectors.
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-memory-store",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "description": "Persistent memory for Claude Code. Powered by MongoDB Atlas.",
   "homepage": "https://claude-memory.com",
   "repository": {