Postman collection
Import Tracira into Postman from a link and start making calls. Includes a ready-made collection for labelling files and editing an output.
Two importable files, both public. Neither needs a login to fetch, so Postman's Import from link works directly.
The ready-made collection
https://www.tracira.com/tracira.postman_collection.jsonIn Postman: Import → Link → paste the URL → Continue. Then open the
collection's Variables tab and paste your workspace token into token. Find the token
in Tracira under Integrations. Nothing else needs filling in.
Run the requests in order. They hand values to each other: request 1 stores the new output's id, request 2 stores the file keys off that output, and everything after uses them. No copy-pasting ids between calls.
| # | Request | What it shows |
|---|---|---|
| 1 | Send an output | Files labelled at attach time. Two 1x1 PNGs inline, so it runs as-is. |
| 2 | Get the output | Reading attachments back: key, filename, label. |
| 3 | Edit metadata (merge) | Only the keys you send change. |
| 4 | Remove one metadata key | An empty value deletes that key. |
| 5 | Replace all metadata | metadataMode: "replace". |
| 6 | Label the files | Relabelling by key, after the fact. |
| 7 | Clear a label | label: null. |
| 8 | All three at once | Metadata, labels and a backfilled context ID in one call. |
| 9a-c | Large file | Create the upload, PUT the bytes, then attach it with the label. |
| 10 | Re-attach on a revision | The label and filename carry over automatically. |
Every request carries a description explaining the rule it demonstrates, so the collection doubles as a walkthrough of file labels and updating an output.
Scope
This collection covers the output lifecycle: submitting, reading, updating, and the file flows. For the decision, flag, search and instructions endpoints, use the OpenAPI import below, which contains every endpoint.
The OpenAPI spec
https://www.tracira.com/openapi.jsonSame Import → Link flow. Postman turns the spec into a collection covering every endpoint, with each request's parameters and schemas filled in from the reference. It also works in Insomnia, Bruno, Hoppscotch, and any OpenAPI 3 client generator.
The spec is the same document that renders the API reference, so it never drifts from the docs.
Set auth after importing the spec
An OpenAPI import does not carry your token. On the generated collection, set Authorization to Bearer Token and paste your workspace token, so every request inherits it.
Authentication
Every request uses the same header:
Authorization: Bearer YOUR_WORKSPACE_TOKENThere is one exception, and it matters: the PUT that uploads file bytes goes to a
presigned storage URL that carries its own signature, and storage rejects any request
that also sends an Authorization header. Request 9b in the collection has auth disabled
for exactly this reason. The same trap applies to GET /api/media/{key}, which redirects to
that signed URL: use GET /api/media-url and fetch the returned link
with no auth header.
Reading errors
Branch on code, never on the error sentence:
{ "error": "Output is still being evaluated. Retry in a few seconds.", "code": "LOG_PENDING" }code is stable; error is human prose whose wording changes whenever it can be made
clearer. Each endpoint's reference page lists the codes it returns. code is absent on plain
validation failures, where error and fields already name the offending field.