brain-cache 0.1.0 → 0.3.0

package/README.md

@@ -14,19 +14,7 @@ brain-cache is an MCP server that gives Claude local, indexed access to your cod
 
 ## Use inside Claude Code (MCP)
 
-The primary way to use brain-cache is as an MCP server. Once configured, Claude automatically calls brain-cache tools instead of reading raw files — no prompting required.
-
-```json
-// .claude/settings.json
-{
-  "mcpServers": {
-    "brain-cache": {
-      "command": "brain-cache",
-      "args": ["mcp"]
-    }
-  }
-}
-```
+The primary way to use brain-cache is as an MCP server. Run `brain-cache init` once — it auto-configures `.mcp.json` in your project root so Claude Code connects immediately. No manual JSON setup needed.
 
 Claude then has access to:
 
@@ -99,26 +87,24 @@ brain-cache init
 brain-cache index
 ```
 
-`brain-cache init` appends MCP tool instructions to your `CLAUDE.md` — so Claude knows when and how to prefer brain-cache tools over built-in file reading. Runs once; won’t duplicate.
-
-**Step 3: Add MCP server to Claude Code**
+`brain-cache init` sets up your project: configures `.mcp.json` so Claude Code connects to brain-cache automatically, and appends MCP tool instructions to `CLAUDE.md`. Runs once; idempotent.
 
-```json
-// .claude/settings.json
-{
-  "mcpServers": {
-    "brain-cache": {
-      "command": "brain-cache",
-      "args": ["mcp"]
-    }
-  }
-}
-```
-
-**Step 4: Use Claude normally**
+**Step 3: Use Claude normally**
 
 brain-cache tools are called automatically. You don’t change how you work — the context just gets better.
 
+> **Advanced:** `init` creates `.mcp.json` automatically. If you need to customise it manually, the expected shape is:
+> ```json
+> {
+>   "mcpServers": {
+>     "brain-cache": {
+>       "command": "brain-cache",
+>       "args": ["mcp"]
+>     }
+>   }
+> }
+> ```
+
 ---
 
 ## 🧩 Core capabilities

package/dist/{askCodebase-ECDSSTQ6.js → askCodebase-DTII3Y6P.js}

@@ -4,16 +4,16 @@ import {
 } from "./chunk-GGOUKACO.js";
 import {
   runBuildContext
-} from "./chunk-7JLSJNKU.js";
-import "./chunk-OKWMQNH6.js";
-import "./chunk-ZLB4VJQK.js";
-import "./chunk-WCNMLSL2.js";
-import "./chunk-P7WSTGLE.js";
-import "./chunk-XXWJ57QP.js";
-import "./chunk-PA4BZBWS.js";
+} from "./chunk-ABKGOJTC.js";
+import "./chunk-5FXXZBZV.js";
+import "./chunk-BF5UDEIF.js";
+import "./chunk-GR6QXZ4J.js";
+import "./chunk-V4ARVFRG.js";
+import "./chunk-6MACVOTO.js";
+import "./chunk-MSI4MDIM.js";
 import {
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/workflows/askCodebase.ts
 import Anthropic from "@anthropic-ai/sdk";

package/dist/buildContext-JKYV7CCP.js

@@ -0,0 +1,14 @@
+#!/usr/bin/env node
+import {
+  runBuildContext
+} from "./chunk-ABKGOJTC.js";
+import "./chunk-5FXXZBZV.js";
+import "./chunk-BF5UDEIF.js";
+import "./chunk-GR6QXZ4J.js";
+import "./chunk-V4ARVFRG.js";
+import "./chunk-6MACVOTO.js";
+import "./chunk-MSI4MDIM.js";
+import "./chunk-3SFDFUEX.js";
+export {
+  runBuildContext
+};

package/dist/{chunk-PDQXJSH4.js → chunk-3SFDFUEX.js}

@@ -64,6 +64,9 @@ var logger = pino(
 function childLogger(component) {
   return logger.child({ component });
 }
+function setLogLevel(level) {
+  logger.level = level;
+}
 
 export {
   GLOBAL_CONFIG_DIR,
@@ -83,5 +86,6 @@ export {
   DIAGNOSTIC_SEARCH_LIMIT,
   DEFAULT_TOKEN_BUDGET,
   FILE_HASHES_FILENAME,
-  childLogger
+  childLogger,
+  setLogLevel
 };

package/dist/{chunk-OKWMQNH6.js → chunk-5FXXZBZV.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/services/tokenCounter.ts
 import { countTokens } from "@anthropic-ai/tokenizer";

package/dist/{chunk-XXWJ57QP.js → chunk-6MACVOTO.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   IndexStateSchema
-} from "./chunk-PA4BZBWS.js";
+} from "./chunk-MSI4MDIM.js";
 import {
   DEFAULT_EMBEDDING_DIMENSION,
   EMBEDDING_DIMENSIONS,
@@ -9,7 +9,7 @@ import {
   PROJECT_DATA_DIR,
   VECTOR_INDEX_THRESHOLD,
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/services/lancedb.ts
 import * as lancedb from "@lancedb/lancedb";

package/dist/{chunk-7JLSJNKU.js → chunk-ABKGOJTC.js}

@@ -2,29 +2,29 @@
 import {
   assembleContext,
   countChunkTokens
-} from "./chunk-OKWMQNH6.js";
+} from "./chunk-5FXXZBZV.js";
 import {
   RETRIEVAL_STRATEGIES,
   classifyQueryIntent,
   deduplicateChunks,
   searchChunks
-} from "./chunk-ZLB4VJQK.js";
+} from "./chunk-BF5UDEIF.js";
 import {
   embedBatchWithRetry
-} from "./chunk-WCNMLSL2.js";
+} from "./chunk-GR6QXZ4J.js";
 import {
   isOllamaRunning
-} from "./chunk-P7WSTGLE.js";
+} from "./chunk-V4ARVFRG.js";
 import {
   openDatabase,
   readIndexState
-} from "./chunk-XXWJ57QP.js";
+} from "./chunk-6MACVOTO.js";
 import {
   readProfile
-} from "./chunk-PA4BZBWS.js";
+} from "./chunk-MSI4MDIM.js";
 import {
   DEFAULT_TOKEN_BUDGET
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/workflows/buildContext.ts
 import { readFile } from "fs/promises";

package/dist/{chunk-ZLB4VJQK.js → chunk-BF5UDEIF.js}

@@ -5,7 +5,7 @@ import {
   DIAGNOSTIC_DISTANCE_THRESHOLD,
   DIAGNOSTIC_SEARCH_LIMIT,
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/services/retriever.ts
 var log = childLogger("retriever");

package/dist/{chunk-WCNMLSL2.js → chunk-GR6QXZ4J.js}

@@ -4,7 +4,7 @@ import {
   DEFAULT_EMBEDDING_DIMENSION,
   EMBED_TIMEOUT_MS,
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/services/embedder.ts
 import ollama from "ollama";
@@ -40,7 +40,8 @@ function isContextLengthError(err) {
 }
 async function embedBatchWithRetry(model, texts, dimension = DEFAULT_EMBEDDING_DIMENSION, attempt = 0) {
   try {
-    return await embedBatch(model, texts);
+    const embeddings = await embedBatch(model, texts);
+    return { embeddings, skipped: 0 };
   } catch (err) {
     if (attempt === 0 && isConnectionError(err)) {
       log.warn({ model }, "Ollama cold-start suspected, retrying in 5s");
@@ -50,24 +51,21 @@ async function embedBatchWithRetry(model, texts, dimension = DEFAULT_EMBEDDING_D
     if (isContextLengthError(err)) {
       log.warn({ model, batchSize: texts.length }, "Batch exceeded context length, falling back to individual embedding");
       const results = [];
+      let skipped = 0;
       for (const text of texts) {
         try {
           const [vec] = await embedBatch(model, [text]);
           results.push(vec);
         } catch (innerErr) {
           if (isContextLengthError(innerErr)) {
-            process.stderr.write(
-              `
-brain-cache: chunk too large for embedding model, skipping (${text.length} chars)
-`
-            );
+            skipped++;
             results.push(new Array(dimension).fill(0));
           } else {
             throw innerErr;
           }
         }
       }
-      return results;
+      return { embeddings: results, skipped };
     }
     throw err;
   }

package/dist/{chunk-PA4BZBWS.js → chunk-MSI4MDIM.js}

@@ -3,7 +3,7 @@ import {
   GLOBAL_CONFIG_DIR,
   PROFILE_PATH,
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/services/capability.ts
 import { execFile } from "child_process";

package/dist/{chunk-P7WSTGLE.js → chunk-V4ARVFRG.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   childLogger
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/services/ollama.ts
 import { execFile, spawn } from "child_process";

package/dist/cli.js

@@ -5,23 +5,23 @@ import {
 
 // src/cli/index.ts
 import { Command } from "commander";
-var version = "0.1.0";
+var version = "0.3.0";
 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-TRPFEOHF.js");
+  const { runInit } = await import("./init-SXC4MWOR.js");
   await runInit();
 });
 program.command("doctor").description("Report system health: GPU, VRAM tier, Ollama status").action(async () => {
-  const { runDoctor } = await import("./doctor-5775VUMA.js");
+  const { runDoctor } = await import("./doctor-3RIVSSNB.js");
   await runDoctor();
 });
 program.command("index").description("Index a codebase: parse, chunk, embed, and store in LanceDB").argument("[path]", "Directory to index (defaults to current directory)").option("-f, --force", "Force full reindex, ignoring cached file hashes").action(async (path, opts) => {
-  const { runIndex } = await import("./workflows-MJLEPCZY.js");
+  const { runIndex } = await import("./workflows-TWA2GDHJ.js");
   await runIndex(path, { force: opts.force });
 });
 program.command("search").description("Search indexed codebase with a natural language query").argument("<query>", "Natural language query string").option("-n, --limit <n>", "Maximum number of results", "10").option("-p, --path <path>", "Project root directory").action(async (query, opts) => {
-  const { runSearch } = await import("./search-WKKGPNLV.js");
+  const { runSearch } = await import("./search-BF7QY64J.js");
   await runSearch(query, {
     limit: parseInt(opts.limit, 10),
     path: opts.path
@@ -30,12 +30,12 @@ program.command("search").description("Search indexed codebase with a natural la
 program.command("status").description(
   "Show index stats: files indexed, chunks stored, last indexed time"
 ).argument("[path]", "Project root directory (defaults to current directory)").action(async (path) => {
-  const { runStatus } = await import("./status-2SOIQ3LX.js");
+  const { runStatus } = await import("./status-JYNMLSXZ.js");
   await runStatus(path);
 });
 program.command("context").description("Build token-budgeted context from codebase for a query").argument("<query>", "Natural language query string").option("-n, --limit <n>", "Maximum number of search results", "10").option("-b, --budget <tokens>", "Token budget for assembled context", "4096").option("-p, --path <path>", "Project root directory").option("--raw", "Output raw JSON (MCP transport compatible)").action(
   async (query, opts) => {
-    const { runBuildContext } = await import("./buildContext-6755TRND.js");
+    const { runBuildContext } = await import("./buildContext-JKYV7CCP.js");
     const result = await runBuildContext(query, {
       limit: parseInt(opts.limit, 10),
       maxTokens: parseInt(opts.budget, 10),
@@ -61,7 +61,7 @@ ${formatTokenSavings({ tokensSent: result.metadata.tokensSent, estimatedWithout,
 program.command("ask").description(
   "Ask a natural language question about the codebase \u2014 retrieves context locally, reasons via Claude"
 ).argument("<question>", "Natural language question about the codebase").option("-b, --budget <tokens>", "Token budget for context retrieval", "4096").option("-p, --path <path>", "Project root directory").action(async (question, opts) => {
-  const { runAskCodebase } = await import("./askCodebase-ECDSSTQ6.js");
+  const { runAskCodebase } = await import("./askCodebase-DTII3Y6P.js");
   const result = await runAskCodebase(question, {
     path: opts.path,
     maxContextTokens: parseInt(opts.budget, 10)

package/dist/{doctor-5775VUMA.js → doctor-3RIVSSNB.js}

@@ -4,14 +4,14 @@ import {
   isOllamaInstalled,
   isOllamaRunning,
   modelMatches
-} from "./chunk-P7WSTGLE.js";
+} from "./chunk-V4ARVFRG.js";
 import {
   detectCapabilities,
   readProfile
-} from "./chunk-PA4BZBWS.js";
+} from "./chunk-MSI4MDIM.js";
 import {
   PROFILE_PATH
-} from "./chunk-PDQXJSH4.js";
+} from "./chunk-3SFDFUEX.js";
 
 // src/workflows/doctor.ts
 import ollama from "ollama";

package/dist/{embedder-KRANITVN.js → embedder-2UG2GDQO.js}

@@ -2,8 +2,8 @@
 import {
   embedBatch,
   embedBatchWithRetry
-} from "./chunk-WCNMLSL2.js";
-import "./chunk-PDQXJSH4.js";
+} from "./chunk-GR6QXZ4J.js";
+import "./chunk-3SFDFUEX.js";
 export {
   embedBatch,
   embedBatchWithRetry

package/dist/{init-TRPFEOHF.js → init-SXC4MWOR.js}

@@ -5,12 +5,12 @@ import {
   isOllamaRunning,
   pullModelIfMissing,
   startOllama
-} from "./chunk-P7WSTGLE.js";
+} from "./chunk-V4ARVFRG.js";
 import {
   detectCapabilities,
   writeProfile
-} from "./chunk-PA4BZBWS.js";
-import "./chunk-PDQXJSH4.js";
+} from "./chunk-MSI4MDIM.js";
+import "./chunk-3SFDFUEX.js";
 
 // src/workflows/init.ts
 import { existsSync, readFileSync, writeFileSync, appendFileSync } from "fs";
@@ -52,7 +52,7 @@ async function runInit() {
     `brain-cache: warming model ${profileWithVersion.embeddingModel} into VRAM...
 `
   );
-  const { embedBatchWithRetry } = await import("./embedder-KRANITVN.js");
+  const { embedBatchWithRetry } = await import("./embedder-2UG2GDQO.js");
   await embedBatchWithRetry(profileWithVersion.embeddingModel, ["warmup"]);
   process.stderr.write("brain-cache: model warm.\n");
   await writeProfile(profileWithVersion);
@@ -63,6 +63,28 @@ async function runInit() {
   VRAM tier: ${profileWithVersion.vramTier}
 `
   );
+  const brainCacheMcpEntry = {
+    command: "node",
+    args: ["node_modules/brain-cache/dist/mcp.js"]
+  };
+  const mcpJsonPath = ".mcp.json";
+  if (existsSync(mcpJsonPath)) {
+    const mcpContent = readFileSync(mcpJsonPath, "utf-8");
+    const parsed = JSON.parse(mcpContent);
+    const existing = parsed.mcpServers?.["brain-cache"];
+    if (existing && JSON.stringify(existing) === JSON.stringify(brainCacheMcpEntry)) {
+      process.stderr.write("brain-cache: .mcp.json already contains brain-cache MCP server, skipping.\n");
+    } else {
+      parsed.mcpServers = parsed.mcpServers ?? {};
+      parsed.mcpServers["brain-cache"] = brainCacheMcpEntry;
+      writeFileSync(mcpJsonPath, JSON.stringify(parsed, null, 2) + "\n");
+      process.stderr.write("brain-cache: added brain-cache MCP server to .mcp.json.\n");
+    }
+  } else {
+    const mcpConfig = { mcpServers: { "brain-cache": brainCacheMcpEntry } };
+    writeFileSync(mcpJsonPath, JSON.stringify(mcpConfig, null, 2) + "\n");
+    process.stderr.write("brain-cache: created .mcp.json with brain-cache MCP server.\n");
+  }
   const claudeMdPath = "CLAUDE.md";
   const brainCacheSection = `
 ## Brain-Cache MCP Tools