IntegrationsAutomation

Automation Platforms

Connect Transient Intelligence to Make, Zapier, n8n, or Power Automate. All automation tools use the same /v1/answer endpoint — trigger → HTTP POST with question and input → parse citations from the response.

Scenario setup

1. Add an HTTP → Make a request module.
2. Set method to POST, URL to https://api.transientintelligence.com/api/models/v1/answer.
3. Add header: x-api-key: your_api_key.
4. Set body type to JSON (application/json).
5. Map question and input from upstream modules (e.g. Google Drive file content).
6. Use citations from the response in downstream modules.

{
"question": {{1.question}},
"input": {{2.fileContent}},
"top_k": 5
}

Zap setup

1. Choose your trigger (e.g. new file in Dropbox, new row in Google Sheets).
2. Add a Webhooks by Zapier → POST action, or use a Code by Zapier step for more control.
3. Set URL to /v1/answer, add the x-api-key header.
4. Pass question and input as data fields.
5. Map citations from the response to your output step (Slack, email, Sheets, etc.).

const res = await fetch("https://api.transientintelligence.com/api/models/v1/answer", {
method: "POST",
headers: { "x-api-key": process.env.TI_API_KEY, "Content-Type": "application/json" },
body: JSON.stringify({
question: inputData.question,
input: inputData.fileContent,
top_k: 5
})
});
const data = await res.json();
return { citations: data.citations, answer: data.answer?.summary };

Workflow setup

1. Add an HTTP Request node.
2. Method: POST, URL: https://api.transientintelligence.com/api/models/v1/answer.
3. Authentication: Generic Credential Type → Header Auth, name x-api-key.
4. Body: JSON with question and input mapped from upstream nodes.
5. Use a Set or IF node downstream to branch on empty citations.

{
"question": "{{ $json.question }}",
"input": "{{ $json.fileContent }}",
"top_k": 5
}

Response schema

Every /v1/answer response has the same shape. Map these fields in your downstream modules.

{
"answer": {
"summary": "string — synthesised answer grounded in citations",
"confidence": "number — 0 to 1"
},
"citations": [
{
"quote": "string — verbatim passage from the document",
"locator": "string — page / section / line reference",
"relevance": "number — passage relevance score"
}
],
"run_id": "string — use to retrieve full results via /v1/runs/{id}/results"
}

An empty citations array means no supporting evidence was found — branch on this in your flow rather than passing the empty response downstream.

Common mistakes

Passing a file path instead of file content

The input field expects the document text, not a path string like /drive/file.pdf. Download and extract the text in an upstream module before passing it to TI.

Missing Content-Type header

Set Content-Type: application/json on every request. Without it, some platforms (Zapier Webhooks, Power Automate HTTP) default to form-encoded and the body will be rejected.

Not branching on empty citations

If citations is empty, the model found no evidence — don't pass answer.summary downstream as if it were grounded. Add an IF/Condition node to handle the no-evidence path explicitly.

Sending binary file bytes as input

input must be plain text. For PDFs and DOCX files, use a text-extraction module (e.g. PDF.co, Adobe Extract, or Google Drive export as text/plain) before calling TI.