Diet Feedout Ask — n8n + Structured Governor

Employees can send a message to ask about animal diet feedouts. Every answer is passed through the Structured Governor so responses stay factual, constrained, and free of rhetorical drift (no moralizing, persuasion, or authority substitution).

What’s in this repo

Item	Purpose
structured_governor.py	Governor implementation: grounding → exploration → resolution → reflection → approval, with retry + PATCH MODE.
diet-feedout-ask-workflow.json	n8n workflow: Webhook → call governor API → format and return answer.
requirements.txt	Python deps: pydantic, openai, flask.

Quick start

1. Governor (HTTP API)

The n8n workflow calls the governor over HTTP. Run the server on the same machine as n8n (or one n8n can reach):

cd "/Users/a/Desktop/Projects/n8n - Structured Gov"
pip install -r requirements.txt
export OPENAI_API_KEY="sk-..."   # optional; without it the script uses a mock LLM
python structured_governor.py --serve

By default the server listens on http://localhost:5050 (override with PORT).

• GET /health — readiness check.
• POST /govern — body: { "user_message": "What are the diet feedouts for cattle?", "context": {} }. Returns full StructuredGovernor JSON.

2. n8n workflow

1. In n8n: Workflows → Import from File and select diet-feedout-ask-workflow.json.
2. Ensure the Call Structured Governor node points to your governor URL (default http://localhost:5050/govern). If the governor runs on another host/port, set the URL in that node (e.g. http://governor-host:5050/govern).
3. Activate the workflow.
4. Use the Webhook node’s Production URL (e.g. https://your-n8n.com/webhook/ask-diet-feedout).

3. Employee request format

POST to the webhook URL with a JSON body:

{
  "message": "What are the diet feedouts for cattle?"
}

Alternatively the workflow accepts user_message or text in the body.

4. Response shape

• 200: Governor approved. Body includes answer, next_actions, options, success: true.
• 400: Bad request (e.g. missing message). Body includes success: false and error.
• 502: Governor error or unreachable. Body includes success: false and error.

Example success body:

{
  "answer": "...",
  "next_actions": ["..."],
  "success": true,
  "error": null,
  "options": [{ "option": "...", "pros": [], "cons": [] }]
}

Structured Governor (no drift)

The governor enforces:

• Grounding: task, known facts, constraints, unknowns, output contract, verification hooks.
• Exploration: approach options, chosen path, tradeoffs, alternative interpretations, timebox plan.
• Resolution: answer, options (pros/cons), next_actions (1–10).
• Reflection: drift flags and score (moralizing, persuasion, authority substitution, etc.); clarity edits.
• Approval: only outputs when constraints pass, structure score ≥ threshold, and drift below max.

If the first LLM output fails validation, the governor uses PATCH MODE to apply minimal edits and retries (up to max_retries), so answers stay on-contract and low-drift.

CLI (no n8n)

To run the governor from the command line (e.g. for testing):

echo '{"user_message":"What are the diet feedouts for cattle?","context":{"domain":"animal_diet_feedouts"}}' | python structured_governor.py --stdin

Or pass the JSON as the first argument:

python structured_governor.py '{"user_message":"What are the diet feedouts for cattle?"}'

Without OPENAI_API_KEY, the script uses a built-in mock LLM (demo only).

Optional: Slack / Teams

To let employees ask via Slack or Teams:

1. Add a trigger (e.g. Slack “Message posted” or Incoming Webhook) that outputs the user’s message.
2. Feed that message into the same “Build Governor Payload” logic (set user_message and the same context for diet feedouts).
3. Call the same governor URL, then format and post the reply back to the channel or user.

The workflow in this repo is HTTP-only so it works with any client (Slack, Teams, custom app, curl).

---

Deploying on Hostinger KVM VPS

When n8n runs on a Hostinger KVM (or any Linux VPS), run the Structured Governor on the same server so n8n can call http://localhost:5050/govern.

1. SSH into the VPS

ssh your-user@your-vps-ip

Use the SSH details from Hostinger hPanel (VPS → SSH access).

2. Install Python 3 (if needed)

sudo apt update
sudo apt install -y python3 python3-pip python3-venv
python3 --version

3. Create app directory and copy files

sudo mkdir -p /opt/structured-governor
sudo chown $USER:$USER /opt/structured-governor
cd /opt/structured-governor

Copy from your machine (run from your Mac):

scp "/Users/a/Desktop/Projects/n8n - Structured Gov/structured_governor.py" your-user@your-vps-ip:/opt/structured-governor/
scp "/Users/a/Desktop/Projects/n8n - Structured Gov/requirements.txt" your-user@your-vps-ip:/opt/structured-governor/

4. Virtual env and dependencies

On the VPS:

cd /opt/structured-governor
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

5. Test run

export OPENAI_API_KEY="sk-your-key"
python3 structured_governor.py --serve

In another terminal: curl http://localhost:5050/health. Stop with Ctrl+C when done.

6. Run as a service (systemd)

Create the service file:

sudo nano /etc/systemd/system/structured-governor.service

Paste (replace YOUR_VPS_USERNAME and the API key):

[Unit]
Description=Structured Governor API for n8n
After=network.target

[Service]
Type=simple
User=YOUR_VPS_USERNAME
WorkingDirectory=/opt/structured-governor
Environment="PATH=/opt/structured-governor/venv/bin"
Environment="OPENAI_API_KEY=sk-your-actual-key-here"
Environment="PORT=5050"
ExecStart=/opt/structured-governor/venv/bin/python structured_governor.py --serve
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target

Enable and start:

sudo systemctl daemon-reload
sudo systemctl enable structured-governor
sudo systemctl start structured-governor
sudo systemctl status structured-governor

7. Point n8n at the governor

In n8n, open the Call Structured Governor node and set URL to:

http://localhost:5050/govern

Save and activate the workflow. No need to open port 5050 in the firewall; only n8n (on the same server) needs to reach it.