Skip to content

n8n

Any n8n workflow that can do HTTP can make decisions. The HTTP Request node posts the canonical body and the response comes back as ordinary JSON, so a branch condition can read answers.<id>.choice and answers.<id>.confidence directly — no code node required.

  • Transport: POST /v1/systemone
  • Status: payload verified — the requests below were executed against a local tachyone-serve; the n8n node UI itself is not run in CI.

1. Make the server reachable

n8n usually runs in Docker, so put both services on the same compose network. Add this to the repo's docker/compose.yaml (or to a sibling file you start alongside it):

  n8n:
    image: n8nio/n8n
    ports:
      - "5678:5678"
    depends_on:
      tachyone:
        condition: service_healthy

Compose puts both on one network where the service name resolves, so inside n8n the URL is http://tachyone:8000/v1/systemone. Running n8n outside that compose file? Point it at the published port on the host instead — http://host.docker.internal:8000/v1/systemone (add extra_hosts: ["host.docker.internal:host-gateway"] to the n8n service).

2. HTTP Request node

Field Value
Method POST
URL http://tachyone:8000/v1/systemone
Authentication Generic Credential → Header Auth → Authorization = Bearer change-me
Send Headers Content-Type = application/json
Send Body As JSON
Body the payload below 
{
  "state": "Hi, we were charged twice for the March invoice and finance needs the duplicate reversed before the end of the quarter. Invoice number INV-2291.",
  "model": "tachyone-latest",
  "questions": {
    "queue": {
      "type": "choice",
      "instructions": "Which queue should receive this ticket first?",
      "criteria": {
        "billing-queue": "invoices, payments, refunds",
        "technical-queue": "bugs, outages, system errors",
        "sales-queue": "pricing, new contracts",
        "general-queue": "everything else"
      }
    }
  }
}

Response (captured from a real encoder run):

{
  "model": "tachyone-latest",
  "answers": {
    "queue": {
      "type": "choice",
      "choice": "billing-queue",
      "probabilities": {
        "billing-queue": 1.0,
        "technical-queue": 0.0,
        "sales-queue": 0.0,
        "general-queue": 0.0
      },
      "confidence": 1.0
    }
  },
  "usage": {
    "input_tokens": 36,
    "output_tokens": 1
  }
}

The same payload from a shell, if you want to check the node before building it:

curl -s http://127.0.0.1:8000/v1/systemone -H "Content-Type: application/json" \
  --data @docs/assets/examples/ticket-routing.request.json

3. Read the decision in expressions

What Expression
Winning queue {{ $json.answers.queue.choice }}
Confidence {{ $json.answers.queue.confidence }}
Full distribution {{ JSON.stringify($json.answers.queue.probabilities) }}

Wire these into a Set node (queue = …) and your downstream nodes — Slack, email, a database insert — receive a plain string.

4. Branch on confidence instead of trusting it

Add an IF node: {{ $json.answers.queue.confidence }} is smaller than 0.6 →

  • true → route to a human queue (or call a large model): the decision was too close to call.
  • false → continue automatically.

That is the System-2 handoff pattern from the cookbook, expressed as a workflow branch rather than Python. Set τ per decision shape — 0.6 is illustrative, not a default.

5. Classify many items in one call

For a batch node (or a Schedule Trigger draining a queue), use the additive POST /predict/batch instead of looping:

{
  "requests": [
    { "state": "…", "model": "tachyone-latest", "questions": { "queue": { "type": "choice", "instructions": "Which queue should receive this ticket first?", "criteria": { "billing-queue": "invoices, payments, refunds", "technical-queue": "bugs, outages, system errors", "sales-queue": "pricing, new contracts", "general-queue": "everything else" } } } },
    { "state": "…", "model": "tachyone-latest", "questions": { "queue": { "type": "choice", "instructions": "Which queue should receive this ticket first?", "criteria": { "billing-queue": "invoices, payments, refunds", "technical-queue": "bugs, outages, system errors", "sales-queue": "pricing, new contracts", "general-queue": "everything else" } } } }
  ]
}

The response is {"results": [ …one canonical response per item… ]}; the canonical /v1/systemone shape is untouched.

Failure handling

Status Meaning What the node should do
401 Missing/wrong API key Fix the credential; do not retry
422 Body failed validation (bad options, >255 choices, <2 or >10 score levels) Turn on Continue (using error output) and inspect error.body.details
429 / 529 Upstream throttling / overload Retry with exponential backoff + jitter