Add scripts and documentation for local inference configuration with Ollama and mlx-omni-server

- Introduced `configure_local_inference.sh` to automatically set `.dev.vars` based on active local inference services.
- Updated `start_inference_server.sh` to handle both Ollama and mlx-omni-server server types.
- Enhanced `package.json` to include new commands for starting and configuring inference servers.
- Refined README to include updated instructions for running and adding models for local inference.
- Minor cleanup in `MessageBubble.tsx`.

README.md

@@ -6,17 +6,18 @@
   <img src="https://github.com/user-attachments/assets/620d2517-e7be-4bb0-b2b7-3aa0cba37ef0" width="250" />
 </p>
 
-## Project Status: Testing
-
 ## Table of Contents
+
 - [Stack](#stack)
 - [Installation](#installation)
 - [Deployment](#deployment)
 - [Local Inference](#local-inference)
     - [Ollama](#ollama)
+        - [Adding models for local inference (ollama)](#adding-models-for-local-inference-ollama)
     - [mlx-omni-server (Apple Silicon Only)](#mlx-omni-server-apple-silicon-only)
         - [Adding models for local inference (Apple Silicon)](#adding-models-for-local-inference-apple-silicon)
 - [Testing](#testing)
+- [Troubleshooting](#troubleshooting)
 - [History](#history)
 - [License](#license)
 
@@ -51,26 +52,33 @@
 > Local inference is achieved by overriding the `OPENAI_API_KEY` and `OPENAI_API_ENDPOINT` environment variables. See below.
 ### Ollama
 ~~~bash
-docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama ## Run Ollama (Can also be installed natively)
+bun run openai:local ollama                  # Start ollama server
-bun run openai:local                         # Start OpenAI-compatible server
+bun run openai:local:enable                  # Configure connection
-sed -i '' '/^OPENAI_API_KEY=/d' .dev.vars; echo >> .dev.vars; echo 'OPENAI_API_KEY=required-but-not-used' >> .dev.vars # Reset API key
+bun run server:dev                           # Restart server
-sed -i '' '/^OPENAI_API_ENDPOINT=/d' .dev.vars; echo >> .dev.vars; echo 'OPENAI_API_ENDPOINT=http://localhost:11434' >> .dev.vars # Reset endpoint
-bun run server:dev                           # Start dev server
 ~~~
+#### Adding models for local inference (ollama)
 
+~~~bash
+# See https://ollama.com/library for available models
+MODEL_TO_ADD=gemma3 
+docker exec -it ollama ollama run ${MODEL_TO_ADD}
+~~~  
 ### mlx-omni-server (Apple Silicon Only)
 ~~~bash
-brew tap seemueller-io/tap                   # Add seemueller-io tap
+# (prereq) install mlx-omni-server
-brew install seemueller-io/tap/mlx-omni-server # Install mlx-omni-server
+brew tap seemueller-io/tap                   
-bun run openai:local                         # Start OpenAI-compatible server
+brew install seemueller-io/tap/mlx-omni-server 
-sed -i '' '/^OPENAI_API_KEY=/d' .dev.vars; echo >> .dev.vars; echo 'OPENAI_API_KEY=required-but-not-used' >> .dev.vars # Reset API key
+
-sed -i '' '/^OPENAI_API_ENDPOINT=/d' .dev.vars; echo >> .dev.vars; echo 'OPENAI_API_ENDPOINT=http://localhost:10240' >> .dev.vars # Reset endpoint
+bun run openai:local mlx-omni-server         # Start mlx-omni-server
-bun run server:dev                           # Start dev server
+bun run openai:local:enable                  # Configure connection
+bun run server:dev                           # Restart server
 ~~~
 #### Adding models for local inference (Apple Silicon)
 
 ~~~bash
-# ensure mlx-omni-server is running in the background 
+# ensure mlx-omni-server is running
+
+# See https://huggingface.co/mlx-community for available models
 MODEL_TO_ADD=mlx-community/gemma-3-4b-it-8bit
 
 curl http://localhost:10240/v1/chat/completions \
@@ -81,15 +89,20 @@ curl http://localhost:10240/v1/chat/completions \
   }"
 ~~~  
 
-
-
-
 ## Testing
 
 Tests are located in `__tests__` directories next to the code they test. Testing is incomplete at this time.
 
 > `bun run test` will run all tests
 
+
+## Troubleshooting
+1.  `bun run clean`
+1.  `bun i`
+1.  `bun server:dev` 
+1.  `bun client:dev` 
+1. Submit an issue
+
 History
 ---
 A high-level overview for the development history of the parent repository, [geoff-seemueller-io](https://geoff.seemueller.io), is provided in [LEGACY.md](./LEGACY.md). 

package.json

@@ -19,6 +19,9 @@
     "tail:analytics-service": "wrangler tail -c workers/analytics/wrangler-analytics.toml",
     "tail:session-proxy": "wrangler tail -c workers/session-proxy/wrangler-session-proxy.toml --env production",
     "openai:local": "./scripts/start_inference_server.sh",
+    "openai:local:mlx": "./scripts/start_inference_server.sh mlx-omni-server",
+    "openai:local:ollama": "./scripts/start_inference_server.sh ollama",
+    "openai:local:configure": "scripts/configure_local_inference.sh",
     "test": "vitest run",
     "test:watch": "vitest",
     "test:coverage": "vitest run --coverage.enabled=true"

scripts/configure_local_inference.sh

@@ -0,0 +1,49 @@
+#!/usr/bin/env bash
+
+
+# Ensure .dev.vars file exists.
+# This prevents errors if sed tries to edit a non-existent file and ensures '>>' appends.
+# Function to configure .dev.vars with the specified API key and endpoint
+configure_dev_vars() {
+  local endpoint_url=$1
+  local api_key="required-but-not-used"
+
+  echo "Configuring .dev.vars for endpoint: ${endpoint_url}"
+
+  # Configure OPENAI_API_KEY
+  # 1. Remove any existing OPENAI_API_KEY line
+  sed -i '' '/^OPENAI_API_KEY=/d' .dev.vars
+  # 2. Append a blank line (ensures the new variable is on a new line and adds spacing)
+  # 3. Append the new OPENAI_API_KEY line
+  echo "OPENAI_API_KEY=${api_key}" >> .dev.vars
+
+  # Configure OPENAI_API_ENDPOINT
+  # 1. Remove any existing OPENAI_API_ENDPOINT line
+  sed -i '' '/^OPENAI_API_ENDPOINT=/d' .dev.vars
+  # 3. Append the new OPENAI_API_ENDPOINT line
+  echo "OPENAI_API_ENDPOINT=${endpoint_url}" >> .dev.vars
+
+  echo "Local inference is configured for $endpoint_url"
+}
+
+echo "Checking for local inference services..."
+
+# Check for Ollama on port 11434
+# nc -z -w1 localhost 11434:
+#   -z: Zero-I/O mode (port scanning)
+#   -w1: Timeout after 1 second
+#   >/dev/null 2>&1: Suppress output from nc
+if nc -z -w1 localhost 11434 >/dev/null 2>&1; then
+  echo "Ollama service detected on port 11434."
+  configure_dev_vars "http://localhost:11434"
+# Else, check for mlx-omni-server on port 10240
+elif nc -z -w1 localhost 10240 >/dev/null 2>&1; then
+  echo "mlx-omni-server service detected on port 10240."
+  configure_dev_vars "http://localhost:10240"
+else
+  echo "No active local inference service (Ollama or mlx-omni-server) found on default ports (11434, 10240)."
+  echo "If a service is running on a different port, .dev.vars may need manual configuration."
+  echo ".dev.vars was not modified by this script for OpenAI local inference settings."
+fi
+
+echo "Script finished."

scripts/start_inference_server.sh

@@ -1,8 +1,12 @@
 #!/usr/bin/env bash
 
-SERVER_TYPE="mlx-omni-server"
+if [ "$1" = "mlx-omni-server" ]; then
-
+    printf "Starting Inference Server: %s\n" "$1"
-printf "Starting Inference Server: %s\n" ${SERVER_TYPE}
-
-
     mlx-omni-server --log-level debug
+elif [ "$1" = "ollama" ]; then
+    echo "starting ollama"
+    docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
+else
+    printf "Error: First argument must be 'mlx-omni-server'\n"
+    exit 1
+fi

src/components/chat/messages/MessageBubble.tsx

@@ -1,5 +1,4 @@
 import React, { useEffect, useRef, useState } from "react";
-import { motion } from "framer-motion";
 import { Box, Flex, Text } from "@chakra-ui/react";
 import MessageRenderer from "./ChatMessageContent";
 import { observer } from "mobx-react-lite";
@@ -65,14 +64,7 @@ const MessageBubble = observer(({ msg, scrollRef }) => {
   };
 
   useEffect(() => {
-    if (
+    if (clientChatStore.items.length > 0 && clientChatStore.isLoading && UserOptionsStore.followModeEnabled) { // Refine condition
-      clientChatStore.items.length > 0 &&
-      clientChatStore.isLoading &&
-      UserOptionsStore.followModeEnabled
-    ) {
-      console.log(
-        `${clientChatStore.items.length}/${clientChatStore.isLoading}/${UserOptionsStore.followModeEnabled}`,
-      );
       scrollRef.current?.scrollTo({
         top: scrollRef.current.scrollHeight,
         behavior: "auto",