brain-cache 0.4.1 → 0.4.2

package/README.md

@@ -119,6 +119,19 @@ When using brain-cache build_context, include the token savings summary from the
 
 This keeps it transparent and under your control.
 
+## 🎛 Tuning how much Claude uses brain-cache
+
+`brain-cache init` adds a section to your project's `CLAUDE.md` with clear instructions to use brain-cache tools first. This works well for most users.
+
+If you want to go further, you can strengthen the language yourself. For example:
+
+```
+ALWAYS use brain-cache build_context before reading files or using Grep/Glob.
+Do not skip brain-cache tools — they return better results with fewer tokens.
+```
+
+Or soften it if you prefer Claude to decide on its own. It's your `CLAUDE.md` — edit it to match how you want to work.
+
 ---
 
 ## 🧩 Core capabilities

package/dist/{claude-md-section-S6N6WV5J.js → claude-md-section-6ZJ3TMO4.js}

@@ -4,13 +4,13 @@
 var CLAUDE_MD_SECTION = `
 ## Brain-Cache MCP Tools
 
-When answering questions about the codebase, prefer using brain-cache tools before reading files directly.
+Use brain-cache tools before reading files or using Grep/Glob for codebase questions.
 
 ### build_context (use for understanding)
 
-Call \`mcp__brain-cache__build_context\` with the user's question to retrieve semantically relevant code from across the repo. It deduplicates results and fits them to a token budget.
+Call \`mcp__brain-cache__build_context\` with the user's question before reading individual files. It retrieves semantically relevant code from across the repo, deduplicates results, and fits them to a token budget.
 
-Works well for questions like:
+Use for questions like:
 - "How does X work?" / "Explain X end to end"
 - "Walk me through the flow of X"
 - "What does this page/feature/component do?"
@@ -19,7 +19,7 @@ Works well for questions like:
 - Any question that requires understanding code across multiple files
 - Debugging unfamiliar code paths or understanding error flows
 
-This typically returns better results with fewer tokens than manually reading files.
+This returns better results with fewer tokens than reading files individually.
 
 ### search_codebase (use for locating)
 

package/dist/cli.js

@@ -5,11 +5,11 @@ import {
 
 // src/cli/index.ts
 import { Command } from "commander";
-var version = "0.4.1";
+var version = "0.4.2";
 var program = new Command();
 program.name("brain-cache").description("Local AI runtime \u2014 GPU cache layer for Claude").version(version);
 program.command("init").description("Detect hardware, pull embedding model, create config directory").action(async () => {
-  const { runInit } = await import("./init-WTWOJ55U.js");
+  const { runInit } = await import("./init-QNN5H3DR.js");
   await runInit();
 });
 program.command("doctor").description("Report system health: GPU, VRAM tier, Ollama status").action(async () => {

package/dist/{init-WTWOJ55U.js → init-QNN5H3DR.js}

@@ -86,7 +86,7 @@ async function runInit() {
     process.stderr.write("brain-cache: created .mcp.json with brain-cache MCP server.\n");
   }
   const claudeMdPath = "CLAUDE.md";
-  const { CLAUDE_MD_SECTION: brainCacheSection } = await import("./claude-md-section-S6N6WV5J.js");
+  const { CLAUDE_MD_SECTION: brainCacheSection } = await import("./claude-md-section-6ZJ3TMO4.js");
   if (existsSync(claudeMdPath)) {
     const content = readFileSync(claudeMdPath, "utf-8");
     if (content.includes("## Brain-Cache MCP Tools")) {

package/dist/mcp.js

@@ -1161,7 +1161,7 @@ async function runBuildContext(query, opts) {
 }
 
 // src/mcp/index.ts
-var version = "0.4.1";
+var version = "0.4.2";
 var log9 = childLogger("mcp");
 var server = new McpServer({ name: "brain-cache", version });
 server.registerTool(

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "brain-cache",
-  "version": "0.4.1",
+  "version": "0.4.2",
   "description": "Local MCP-first context engine for Claude. Index your codebase, retrieve only what matters, and cut token usage.",
   "license": "MIT",
   "type": "module",