Sample trace
Every tool call, in the open.
The agent is a pipeline, not a chatbot. Each step is a named tool with its input, its output, how long it took and whether it failed — shown while it runs, and kept afterwards. Nothing hides behind a spinner.
Import · 10 screenshots — a sample run
8 steps · 3.9 s · 1 failed- cloudinary.uploadOK · 812 ms
{"public_id":"pull/jobs/oscar-campaign/celine_look_27","bytes":184320}
Input
{ "file": "celine_look_27_aw26.png", "folder": "pull/jobs/oscar-campaign" }Output
{ "public_id": "pull/jobs/oscar-campaign/celine_look_27", "bytes": 184320 } - cloudinary.trimOK · 306 ms
{"removed_pct":62,"width":1200,"height":1600,"url":"res.cloudinary.com/…/c_fill,w_1200,h_1…
Input
{ "public_id": "pull/jobs/oscar-campaign/celine_look_27", "effect": "trim:20", "target": "3:4" }Output
{ "removed_pct": 62, "width": 1200, "height": 1600, "url": "res.cloudinary.com/…/c_fill,w_1200,h_1600" } - tesseract.readOK · 1.6 s
{"text":"vogue.com/runway/celine\nLook 27 — AW26\nCELINE - runway","confidence":92}
Input
{ "public_id": "pull/jobs/oscar-campaign/celine_look_27", "psm": 6 }Output
{ "text": "vogue.com/runway/celine\nLook 27 — AW26\nCELINE - runway", "confidence": 92 } - parse.lookOK · 3 ms
{"lookNo":27,"season":"AW26","brand":"CELINE","readSource":"image","confidence":0.92}
Input
{ "text": "vogue.com/runway/celine\nLook 27 — AW26\nCELINE - runway", "filename": "celine_look_27_aw26.png" }Output
{ "lookNo": 27, "season": "AW26", "brand": "CELINE", "readSource": "image", "confidence": 0.92 } - db.group_boardsOK · 41 ms
{"boards":5,"brands":["CELINE","BOTTEGA VENETA","JIL SANDER","LEMAIRE","PRADA"]}
Input
{ "jobId": "job_oscar", "looks": 10 }Output
{ "boards": 5, "brands": [ "CELINE", "BOTTEGA VENETA", "JIL SANDER", "LEMAIRE", "PRADA" ] } - shippo.ratesError · 243 ms
Shippo 401: Invalid token — add SHIPPO_API_KEY (test) in .env.local
Input
{ "from": "12 Rue de la Paix, Paris", "to": "Juje Studio, New York", "parcel": { "l": 40, "w": 30, "h": 12, "kg": 2.4 } }Error
Shippo 401: Invalid token — add SHIPPO_API_KEY (test) in .env.local
- shippo.ratesOK · 486 ms
{"shipment":"shp_9f21","rates":3,"cheapest":"UPS Ground $62.40 / 4 days"}
Input
{ "from": "12 Rue de la Paix, Paris", "to": "Juje Studio, New York", "parcel": { "l": 40, "w": 30, "h": 12, "kg": 2.4 } }Output
{ "shipment": "shp_9f21", "rates": 3, "cheapest": "UPS Ground $62.40 / 4 days" } - resend.sendOK · 391 ms
{"id":"re_4c8a…","status":"sent"}
Input
{ "to": "press@celine.com", "template": "request", "looks": [ 27, 31 ], "callIn": "2026-11-14", "return": "2026-11-18" }Output
{ "id": "re_4c8a…", "status": "sent" }
7 ok · 1 failed — persisted per job, so it survives a refresh and is the audit trail for what the agent did.
Why this is a requirement
A stylist is accountable for what leaves her studio. If the agent crops the wrong thing or reads “Look 27” when it is “Look 21”, she has to see that before the board goes to the client. The trace is the difference between an assistant and a black box.
What each step carries
- tool name — cloudinary.trim, tesseract.read, shippo.rates
- input — the exact arguments passed
- output — the structured result the next step uses
- duration — so a slow step is visible, not felt
- status — ok or error, with the provider's message kept verbatim
Where it shows up
Live while importing (streamed), on the job page as the activity trail, and next to each request when a label or email fails. Failed steps are retryable from the trace.