OpenClaw on reComputer - Privacy-First Emotion AI Agent

The core features of this AI Box are emotion perception plus emotional response.

It does not merely recognize emotions and tell you "I detect that you are feeling anxious" — that would be overly mechanical.

What it does is treat emotional state as contextual information and integrate it into daily conversations.It does not merely recognize emotions and tell you "I detect that you are feeling anxious" — that would be overly mechanical.

For example: When you look anxious and ask a code-related question, the Agent might reply:

"Don't worry, I've seen this issue many times. It's easy to fix. Let's check the error message first..."

Instead of replying like this:

"I have detected your current emotional state as anxious. I will switch my response mode to calm and patient style. After analysis, this problem may be..."

Perceive emotions, and embed them naturally into dialogue.

This is our core value: not a fancy emotion detection gimmick, but making AI truly understand people and possess human warmth.

Step 1: Flash the reComputer device image and log in for access

Download and Flash the Ubuntu Image

The reComputer Industrial R2145 supports the Ubuntu system. First, download the Ubuntu image from the official Seeed website.
Set the hardware to Boot mode, connect the device to your computer with a USB-Type C cable, and connect the power cable.

Open rpiboot and Raspberry Pi Imager, then proceed with image flashing.

First Boot and Configuration

After finishing the image flashing, set the hardware to non-Boot mode. Connect the device and your computer to the same network via an Ethernet cable, then use the following command to connect (password: 12345678).

bash

ssh reComputer@raspberrypi.local

Step 2: Environment and OpenClaw Configuration on reComputer

Install Python and Dependencies

bash

pip install tensorflow>=1.7 opencv-contrib-python==3.3.0.9 fer

Install OpenClaw

OpenClaw Download:

bash

curl -fsSL https://openclaw.ai/install.sh | bash

Download Verification:

bash

openclaw --version      # confirm the CLI is available
    openclaw doctor         # check for config issues
    openclaw gateway status # verify the Gateway is running
openclaw --version      # confirm the CLI is available
    openclaw doctor         # check for config issues
    openclaw gateway status # verify the Gateway is running

Configure the Agent Brain

Method 1: API of Mainstream Existing Large Models

If you selected to run the beginner tutorial during installation, select the corresponding model at this position in the tutorial and enter your personal API key.
If the installation is complete and you need to change the model, you can enter the openclaw onboard command to reconfigure.

Method 2: Local Model Deployment (Ollama)

Install Ollama

bash

curl -fsSL https://ollama.com/install.sh | sh
ollama --versio

Connect OpenClaw to Ollama

OpenClaw requires an API Key to identify the Ollama service. Since Ollama is a local service, this Key can be any string.

Set environment variables

bash

export OLLAMA_API_KEY="ollama-local

Configure via OpenClaw

bash

openclaw config set models.providers.ollama.apiKey "ollama-local"

Confirm that the Ollama service is running

Check the Ollama Service

bash

curl http://localhost:11434/api/tags

If it is not running, you can start it

bash

ollama serve

Check OpenClaw Models

View currently available models in OpenClaw

bash

openclaw models list

If no Ollama model is available, you can add it

bash

ollama launch openclaw
ollama launch openclaw --model qwen3:8b

Run the OpenClaw Onboarding Wizard

Run the Configuration Wizard

bash

openclaw onboard

OpenClaw will automatically detect the local Ollama service. Make selections in the wizard as follows

bash

◆  I understand this is personal-by-default and shared/multi-user use requires lock-down. Continue?
│  ● Yes / ○ No
◆  Onboarding mode
│  ● QuickStart (Configure details later via openclaw configure.)
│  ○ Manual

It will then display the currently connected model. This document takes the ollama/qwen3:8b model as an example

bash

◇  Existing config detected ─────────╮
│                                    │
│  workspace: ~/.openclaw/workspace  │
│  model: ollama/qwen3:8b            │
│  gateway.mode: local               │
│  gateway.port: 18789               │
│  gateway.bind: loopback            │
│                                    │
├────────────────────────────────────╯

You can choose either Use existing values or Update values here, but do not select Reset

bash

◆  Config handling
│  ● Use existing values
│  ○ Update values
│  ○ Reset

Select Skip for now for Model provider

bash

◆  Model/auth provider
│  ...
│  ○ Cloudflare AI Gateway
│  ○ Custom Provider
│  ● Skip for now

Select All providers

bash

◆  Filter models by provider
│  ● All providers
│  ○ amazon-bedrock
│  ○ anthropic
│  ○ azure-openai-responses

In Default Model, the connected Ollama model will appear at the top and serve as the default model; just select it, for example

bash

◆  Default model
│  ● Keep current (ollama/qwen3:8b)
│  ○ Enter model manually
│  ○ amazon-bedrock/anthropic.claude-3-haiku-20240307-v1:0
│  ○ amazon-bedrock/anthropic.claude-3-5-haiku-20241022-v1:0

At this point, OpenClaw has successfully connected to the local Ollama model. You can configure the subsequent steps according to your own needs

bash

◆  How do you want to hatch your bot?
│  ○ Hatch in TUI (recommended)
│  ● Open the Web UI
│  ○ Do this later

Start Using OpenClaw + Ollama

After the wizard completes, OpenClaw will automatically open the Web UI
Your robot can now run with the local Ollama model. If you want to switch models, you can

bash

ollama launch openclaw

You may also specify the model directly

bash

ollama launch openclaw --model qwen3:8b

Step 3: Develop Visual Recognition Algorithm

Emotion recognition can be implemented in three ways. Select the appropriate solution according to your resources and needs.

Method 1: Multimodal Large Model (Most convenient, but high resource consumption and slow running speed)

Adopt multimodal models such as GPT-4V, which have built-in emotion recognition functions and do not require local models.

Method 2: Small Model (Most stable)

We can implement emotion recognition with yolo and FER, or use open-source/self-built small models such as FaceNet.

This paper adopts the efficient implementation scheme of YOLO+ FER to encapsulate the inference code:

The YOLO algorithm can be quickly obtained and successfully run from the reComputer AI Lab platform through a single command:

bash

sudo docker run --rm \
--name cm5-hailo8-yolo11n \
--privileged \
--net=host \
-e PYTHONUNBUFFERED=1 \
--device /dev/hailo0:/dev/hailo0 \
-v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
-v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11n:latest \
python web_detection.py --model_path model/yolov11n.hef --video_path video/test.mp4

Step 4: Configure systemd Auto-start on Boot

To make the emotion recognition service run automatically when reComputer starts up, configure the systemd user service, modify the emotion recognition script, and save it to a fixed file readable by OpenClaw. This allows OpenClaw to obtain emotional information in real time during subsequent conversations:

Enable and start the service:

bash

mkdir -p ~/.config/systemd/user
systemctl --user daemon-reload
systemctl --user enable emotion.service
systemctl --user start emotion.service

Step 5: Configure Emotion Settings

Configure OpenClaw SOUL.md

Edit OpenClaw's SOUL.md file to add emotion perception capabilities:

bash

Emotional Awareness Mode
You have an emotional awareness AI companion mode. When the user starts the emotion detection service:
Read emotion state file:
/home/seeed/.openclaw/emotion_state.json
Before every reply, check the emotion and style fields in emotion_state.json, then adjust your tone based on the user’s emotion.

Restart OpenClaw to Apply the Configuration

Core Operations of OpenClaw

bash

Terminal:
    openclaw --Start OpenClaw
    openclaw onboard --Reconfigure OpenClaw
Chat interface:
    chat --Enter the previous conversation
    /new --Create a new conversation interface

Step 6: Verification and Testing

Reboot the Device

bash

sudo reboot

The device should now automatically run the emotion recognition service

Camera and Emotion Detection Verification

Make different facial expressions in front of the camera, and check the emotion update status via the file /home/seeed/.openclaw/emotion_state.json

bash

cd /home/seeed/.openclaw
vim emotion_state.json

Expected result:

Full Conversation Test

Start an OpenClaw conversation:

bash

openclaw
chat

Test Scenario:

Emotion Response Examples

Scenario 1: User appears anxious

bash

User (frowning): This code keeps throwing errors, please help me check it.
Agent (Anxiety emotion detected): "Don’t worry, let me take a look at this problem for you. It appears to be a type conversion error. Let me fix it for you:It should work properly after this adjustment. Feel free to tell me if you still run into issues."

Emotion Update → User’s relaxed expression detected after several seconds
Agent: "I see the issue is resolved and you look relieved~ Is there anything else I can assist you with?"

Scenario 2: User looks sad

bash

User (sighing): My coding skills seem to be getting worse lately.
Agent (Frustration emotion detected): "I understand how you feel. It’s perfectly normal to hit a bottleneck sometimes. The fact that you can recognize this means you have high standards for code quality, and that in itself is progress. Would you like me to check exactly what you’re unsatisfied with? Let’s optimize it together."