fix local3 notebook

Author: MikeBirdTech

examples/local3.ipynb

@@ -4,7 +4,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "This example uses Codestral via Ollama"
+    "This notebook replicates running Open Interpreter locally and uses Llama3 via llamafile"
    ]
   },
   {
@@ -13,7 +13,13 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!ollama run codestral"
+    "## Download LLama3\n",
+    "\n",
+    "# Download the Meta-Llama-3-8B-Instruct.llamafile\n",
+    "!curl -L -o Meta-Llama-3-8B-Instruct.Q5_K_M.llamafile 'https://huggingface.co/Mozilla/Meta-Llama-3-8B-Instruct-llamafile/resolve/main/Meta-Llama-3-8B-Instruct.Q5_K_M.llamafile?download=true'\n",
+    "\n",
+    "# Make the downloaded file executable\n",
+    "!chmod +x Meta-Llama-3-8B-Instruct.Q5_K_M.llamafile"
    ]
   },
   {
@@ -22,7 +28,10 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install open-interpreter"
+    "## Install OI\n",
+    "\n",
+    "!pip install open-interpreter --quiet\n",
+    "!pip install opencv-python --quiet"
    ]
   },
   {
@@ -31,19 +40,65 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "## Configure OI\n",
+    "\n",
+    "import cv2\n",
+    "import subprocess\n",
     "from interpreter import interpreter\n",
     "\n",
     "interpreter.offline = True\n",
-    "interpreter.llm.model = \"ollama/codestral\"\n",
-    "interpreter.llm.api_base = \"http://localhost:11434\"\n",
-    "interpreter.llm.context_window = 32000\n",
-    "interpreter.llm.max_tokens = 8000\n",
-    "interpreter.auto_run = True\n",
     "\n",
-    "def run(prompt):\n",
-    "    interpreter.chat(prompt)\n",
+    "interpreter.llm.model = \"openai/local\" # Tells OI to use an OpenAI-compatible server\n",
+    "interpreter.llm.api_key = \"dummy_key\"\n",
+    "interpreter.llm.api_base = \"http://localhost:8081/v1\"\n",
+    "interpreter.llm.context_window = 7000\n",
+    "interpreter.llm.max_tokens = 1000\n",
+    "interpreter.llm.supports_functions = False"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "## Start server, run OI\n",
+    "\n",
+    "import subprocess\n",
+    "import threading\n",
+    "import os\n",
+    "\n",
+    "def read_output(process):\n",
+    "    while True:\n",
+    "        output = process.stdout.readline()\n",
+    "        if output == b'' and process.poll() is not None:\n",
+    "            break\n",
+    "        if output:\n",
+    "            print(output.decode().strip())\n",
+    "\n",
+    "# Check if the file exists and has execute permissions\n",
+    "file_path = os.path.abspath('Meta-Llama-3-8B-Instruct.Q5_K_M.llamafile')\n",
+    "\n",
+    "# Why are the arguments not being used??\n",
+    "command = [file_path, \"--nobrowser\", \"-ngl\", \"9999\"]\n",
+    "print(command)\n",
+    "\n",
+    "# Setting up the Popen call with stderr redirected to stdout\n",
+    "process = subprocess.Popen(command, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, shell=True)\n",
+    "\n",
+    "# Thread to handle the output asynchronously\n",
+    "thread = threading.Thread(target=read_output, args=(process,))\n",
+    "thread.start()\n",
+    "\n",
+    "# Here you can do other tasks concurrently\n",
+    "# For example:\n",
+    "interpreter.chat()\n",
+    "\n",
+    "# Wait for the thread to finish if the process completes\n",
+    "thread.join()\n",
     "\n",
-    "run(\"how many files are on my desktop?\")"
+    "# Ensure the process has completed\n",
+    "process.wait()"
    ]
   }
  ],