diff --git a/.gitignore b/.gitignore
index 4076ee1..941af50 100644
--- a/.gitignore
+++ b/.gitignore
@@ -4,3 +4,4 @@
 prompt.md
 todo
 .toak-ignore
+/google_gemma-3-1b-it-Q6_K.llamafile
diff --git a/bun.lockb b/bun.lockb
index 45b6afc..11d9545 100755
Binary files a/bun.lockb and b/bun.lockb differ
diff --git a/package.json b/package.json
index 403f766..7368727 100644
--- a/package.json
+++ b/package.json
@@ -45,7 +45,9 @@
     "lint:fix": "eslint src/ --fix",
     "format": "prettier --write \"**/*.{js,jsx,ts,tsx,json,md,yml,yaml}\"",
     "fix": "bun format && bun lint:fix",
-    "release": "bunx release-it"
+    "release": "bunx release-it",
+    "run:inference": "./google_gemma-3-1b-it-Q6_K.llamafile --server --nobrowser",
+    "download:inference": "./download_optimization_model.sh"
   },
   "dependencies": {
     "glob": "^11.0.1",
diff --git a/scripts/download_optimization_model.sh b/scripts/download_optimization_model.sh
new file mode 100644
index 0000000..8274553
--- /dev/null
+++ b/scripts/download_optimization_model.sh
@@ -0,0 +1,4 @@
+MODEL_URL=https://huggingface.co/Mozilla/gemma-3-1b-it-llamafile/resolve/main/google_gemma-3-1b-it-Q6_K.llamafile?download=true
+
+wget "${MODEL_URL}"
+
diff --git a/scripts/optimize-ignore.ts b/scripts/optimize-ignore.ts
new file mode 100755
index 0000000..fa49074
--- /dev/null
+++ b/scripts/optimize-ignore.ts
@@ -0,0 +1,30 @@
+export async function optimizeToakIgnore(content: string) {
+  const inferenceProcess = Bun.spawn(['bun', 'run:inference']);
+
+  await new Promise(resolve => setTimeout(resolve, 5000));
+
+  const prompt = `You are a helpful assistant.
+## Context
+~~~
+${content}
+~~~
+Respond with a list of files that should be added to the .toak-ignore file to reduce noise in the context. No extra text or explanations.`;
+
+  async function run() {
+    try {
+      const response = await fetch('http://127.0.0.1:8080/completion', {
+        method: 'POST',
+        body: JSON.stringify({
+          prompt,
+          n_predict: 512,
+        }),
+      });
+      const data = await response.json();
+      console.log(data.content);
+    } catch (error) {
+      console.error('Error:', error);
+    }
+  }
+  await run();
+  inferenceProcess.kill();
+}
\ No newline at end of file
diff --git a/src/cli.ts b/src/cli.ts
index 8d519b4..e9c63fc 100644
--- a/src/cli.ts
+++ b/src/cli.ts
@@ -3,10 +3,22 @@ import type { PresetPrompt } from './prompts';
 
 console.log('RUNNING TOKENIZER');
 import { MarkdownGenerator, type MarkdownGeneratorOptions } from './MarkdownGenerator';
+import { optimizeToakIgnore } from '../scripts/optimize-ignore.ts';
+
+let output = '';
+process.stdout.write = (write => {
+  return (str: string | Uint8Array, ...args: any) => {
+    output += str.toString();
+    return write.apply(process.stdout, [str, ...args]);
+  };
+})(process.stdout.write);
+
+let optimizeIgnore = false;
 
 const args = process.argv.slice(2);
 const options: { prompt?: PresetPrompt; } & MarkdownGeneratorOptions = {
 
+
 };
 type ValidArg = keyof MarkdownGeneratorOptions;
 
@@ -18,6 +30,10 @@ for (let i = 0; i < args.length; i++) {
     options["todoPrompt"] = args[i + 1]
     i++;
   }
+  if (args[i] === '--optimize-ignore') {
+    optimizeIgnore = true;
+    i++;
+  }
   const arg = args[i].replace(/^--/, '');
   if (arg as any satisfies ValidArg) {
       // @ts-ignore - arg can't be used to index options
@@ -30,12 +46,19 @@ for (let i = 0; i < args.length; i++) {
 }
 
 const generator = new MarkdownGenerator(options);
+
+
+
+
 generator
   .createMarkdownDocument()
-  .then((result: { success: boolean }) => {
+  .then(async (result: { success: boolean }) => {
     if (!result.success) {
       process.exit(1);
     }
+    if (optimizeIgnore) {
+      await optimizeToakIgnore(output);
+    }
   })
   .catch((error: any) => {
     console.error('Error:', error);