noon-exports

# Noon — Data Exports > **PREREQUISITE:** Read `../noon-shared/SKILL.md` for login, page > structure, and the per-store download directory. Three export surfaces, each with its own quirks: Sales (per country), Transaction View (multi-contract, long polling, status enum), Catalog Exports (typed templates). ## 1. Sales Export (Per Country) **URL**: `https://reports.noon.partners/en/sales/?project=PRJ{project_id}` ### Country Switching Use the **"My Stores" dropdown at bottom-left**, NOT the "Destination" filter in the toolbar. Each store shows a flag icon. ### Date Range for Last Month > ⚠️ **The page opens on a ROLLING 30-DAY window, not last month.** On > 3 August the inputs read `2026-07-04 → 2026-08-03`. Export without > touching them and you get a file that **misses the first days of the > month and includes days from this one** — the same trap as the Amazon > FBA-returns preset, and just as quiet: the CSV looks perfectly normal. > Observed live twice, once by hand and once in a scheduled run. > > So: set the range, then **read it back before you export**. A monthly > report whose window is off by three days at each end is worse than a > missing one, because nothing downstream can tell. > > ```bash > browser-use <<'PY' > # Read the two date inputs BY NAME. Collecting every input on the page > # returns unrelated controls too, and then it is guesswork which pair of > # values is the committed range — which is the same guessing this > # readback exists to remove. > print(js("return ['Start date','End date'].map(function(p){" > "var i=document.querySelector('input[placeholder=\"'+p+'\"]');" > "return p+'='+(i?i.value:'MISSING');});")) > # MUST show the 1st and the last day of the target month before you > # click Export. If it shows today's date or a 30-day window, the range > # never committed — set it again. > PY > ``` ```bash browser-use <<'PY' fill_input("input[name=start_date]", "01 Mar 2026") # start-date input fill_input("input[name=end_date]", "31 Mar 2026") # end-date input # Calendar popup opens; click the end date cell (td title) to confirm: js("document.querySelector('td[title=\"2026-03-31\"]')?.click()") PY ``` URL reflects dates: `?from_date=2026-03-01&to_date=2026-03-31`. ### Export Flow (Verified) 1. Click Export button → modal appears 2. Modal shows previous export file with download link (if any) 3. Click "New export" to regenerate 4. Wait ~10-30s for "Exporting" spinner to finish 5. Download link appears with filename: `sales_export_{date}_{time}_{project}_{country}_{seq}.csv` 6. Click download link 7. Close modal ### Both Countries After the first country's export, close modal, switch "My Stores" to the next country, repeat. ## 2. Transaction View Export **URL**: `https://noon-payments.noon.partners/en/transaction-view?project=PRJ{project_id}` Left nav: Statements, SOA, Transaction View, Invoices & Creditnotes, Legacy Exports. ### Critical facts (read before acting) 1. **One unfiltered export covers every contract in the project.** A single Transaction View export for a project that has multiple country contracts produces **one CSV with rows for every contract** — identifiable by the `Contract` / `Contract Title` columns (e.g. `MPXXXXXXXXEG,Noon EG` and `MPXXXXXXXXKW,Noon KW`). See "Transaction CSV Structure" below. **Do not run one export per country** — run one, then split locally with pandas / awk. To limit the export to a single country, filter the Contracts dropdown to that country's contract *before* clicking Download. 2. **Status enum is `Requested → Exporting → Processed`** (not "Processing → Completed"). When the panel row reads `Processed`, click the **Export** button on the right of that row — that is what actually writes the CSV. The first Download click only queues the export job. 2b. **There are TWO "Export" buttons — do not confuse them.** The **main** button (top of the page, near the Contracts / Transaction Type controls) *initiates* the job and its text flips to `Exporting`, then reverts to `Export` when done — clicking it again just queues a **second** job, it does NOT download. The button that actually downloads is a **separate** one inside the results panel (`ExportBox_content__*`) that renders **below the transaction table** after the job completes. Because both read "Export", a bare `find(b => /^export$/.test(b.textContent))` matches the MAIN button first — always scope the panel click to the `ExportBox_content` container (see step 4). The reverting main button is easy to misread as "done, file saved" when nothing downloaded. 3. **Exports can take several minutes**, occasionally >5. Do not re-click Download while a previous export is still `Exporting` — each click queues a *new* `EXP{…}` job that competes for the same backend and usually extends the wait. 4. **Download location depends on the browser backend**, not on noon. The file is not guaranteed to land in `~/Downloads`. See "Finding the downloaded file" below. 5. **`No data` disables the Download button.** Date ranges with no activity show a greyed-out button and no panel appears — that's a no-op, not a bug. Skip that range. ### Country switching on Transaction View The top-of-page **flag widget** (little arrow next to the country flag) is the country switcher — not the Contracts dropdown. The Contracts dropdown lists only contracts for the *currently selected country*, so after switching flag the dropdown repopulates. But since unfiltered Download already covers all contracts, you usually don't need to switch country at all. ### Download flow ```bash # 1. Navigate + set the date range, confirm via page_info(). browser-use <<'PY' new_tab("https://noon-payments.noon.partners/en/transaction-view?project=PRJ{project_id}") wait_for_load() fill_input("input[name=start_date]", "YYYY-MM-DD") # Start date input fill_input("input[name=end_date]", "YYYY-MM-DD") # End date input # Optional: filter Contracts dropdown to a single country's contract. print(page_info()) PY # 2. Click the MAIN Export/Download button ONCE to queue the job. # (It is labelled "Export" on the current build — text flips to # "Exporting" — or "Download" on older builds.) NEVER re-click — # each click queues a new export job on Noon's backend, backing up # the queue and making every subsequent attempt take longer. browser-use <<'PY' print(js("var b=Array.from(document.querySelectorAll('button')).find(b=>/^(export|download)$/i.test(b.textContent.trim())); if(b){b.click(); 'clicked: '+b.textContent.trim();} else 'main export button not found';")) PY # 3. Poll for Processed status. Exports can take up to 35 min # (data-heavy months are slower). Keep waiting — do NOT # re-click Download, do NOT close the panel. for i in $(seq 1 70); do sleep 30 browser-use <<'PY' | grep -q "Processed" && break print(page_info()) PY done # 4. Click the Export button INSIDE the results panel — the one that # actually writes the file. Scope the search to the ExportBox_content # container so it does NOT match the main Export button (which would # queue another job — see critical fact 2b). Scroll down first; the # panel renders below the transaction table. browser-use <<'PY' js("window.scrollTo(0, document.body.scrollHeight)") print(js("var p=document.querySelector('[class*=ExportBox_content]'); if(!p) 'panel not rendered yet — poll longer'; else {var b=Array.from(p.querySelectorAll('button')).find(b=>/export/i.test(b.textContent)); if(b){b.click(); 'clicked panel export';} else 'no export button inside ExportBox panel';}")) PY ``` ### API fallback (if page refreshes or panel disappears) If the browser session disconnects or the page refreshes during the long wait, do NOT click Download again — the export job is still running on Noon's backend. Instead, query the status API directly and download via curl: ```bash # Poll the status API for the export code you recorded earlier. browser-use <<'PY' print(js(""" var xhr = new XMLHttpRequest(); xhr.open('POST', '/_svc/mp-partner-impex-api/export/status', false); xhr.setRequestHeader('Content-Type', 'application/json'); xhr.send(JSON.stringify({exportCode: '<EXP_CODE>'})); return xhr.responseText; """)) PY # Response: {"export": {"status_code": "COMPLETE", # "download_url": "https://storage.googleapis.com/..."}} # Once status_code is COMPLETE, download via curl: curl -sL -o transaction_view.csv "<download_url>" ``` This avoids re-triggering the export and lets you recover a completed export even after a browser disconnect. ### Finding the downloaded file Downloads go to `~/.vibe-seller/downloads/<store-slug>/`. This is a stable per-store directory managed by the CDP proxy (it overrides browser-use's random temp dirs). File names for Transaction View exports start with `noon_financeweb_transactionviewreportonitemlevelwithcontractselection` and end with `.csv`. Repeat downloads get ` (1)`, ` (2)` suffixes — always take the newest by `mtime`, then copy into the task workspace (dropping the `(N)` suffix). See the amazon-reports skill's "Download Behavior" section for the general pattern. ### Transaction CSV Structure Header columns (in order): ```csv Contract,Contract Title,Reference Nr,Order Nr,Item Nr,Order Date, Transaction Date,Title,SKUs,Partner SKUs,Transaction Type,Currency, Net Proceeds,Referral Fee including VAT, Fullfilment & Logistics Fees including VAT, ← sic, Noon misspells this Shipping Credits including VAT,Other Order Fees including VAT, Order Subsidies including VAT,Non-Order Fees including VAT, Non-Order Subsidies including VAT,Others including VAT,Total ``` - **`Contract`** is the contract code — 12 chars, `MP` prefix, country-code suffix (e.g. `MPXXXXXXXXEG`, `MPXXXXXXXXKW`, `MPXXXXXXXXOM`). Use the last 2 chars to bucket by country. - **`Contract Title`** is the human-readable name (`Noon EG`, `Noon KW`, …). - Use either column to split a mixed-country export locally: ```python import pandas as pd df = pd.read_csv('<exported-file>.csv') for cc in df['Contract'].str[-2:].unique(): df[df['Contract'].str.endswith(cc)].to_csv( f'transactions_{cc}.csv', index=False ) ``` Observed `Transaction Type` values include `order`, `order_update`, `statement_fee`, `payment`, `balance_transfer`. Noon may add more; filter liberally when aggregating. ## 3. Catalog Exports **URL**: `https://noon-catalog.noon.partners/en/exports?project=PRJ{project_id}` Click "Add Export" → modal with Type dropdown: | Type | Description | |------|-------------| | Catalog Export | Full catalog data | | Content | Product titles, descriptions, images, attributes | | Pricing | Current + promotional prices | | Stock | Inventory levels | | Partner SKU Generate | Generate partner SKU mappings | | Reports | Performance/sales report data | | Global Catalog Export | Global catalog across markets | Select type → Create → wait → download from Result column when Status shows "Completed" (1-10 minutes depending on size). ## Tips - **Transaction View export covers all contracts by default** — one click produces one CSV with rows from every country contract. Don't export twice (see § 2). - **Transaction View status enum is `Requested → Exporting → Processed`**, not "Processing → Completed". Click the Export button on the panel only after it reads `Processed`. - **Exports can take up to 35 min** — data-heavy months are slower. Poll for at least 35 min before giving up. NEVER click Download again while a previous export is still processing — it queues a new job and makes everything slower. - **Export files land in `~/.vibe-seller/downloads/<store-slug>/`**, not `~/Downloads`. ## 4. What the Transaction View does NOT give you The Transaction View is **order-grained for sales but lumped for FBN fees**. Two things a finance/profitability task usually needs are not in it: - **Per-SKU storage / return fees.** The whole month's FBN storage settles as a single `balance_transfer` row whose `Others including VAT` is opaque. Per-SKU monthly / long-term / non-saleable storage and RTV removal fees come from the **FBN Reports** page — see `noon-fbn` § 8 and `noon-fbn/references/fee-reports.md`. That reference also carries the reconciliation proving the four reports sum to this settlement (with a **one-month lag** and a **VAT gross-up**). - **Per-SKU ad spend.** `statement_fee` rows give only an account-level `Advertising Fee`. Per-SKU spend comes from Ad Manager's `Export all campaigns` — see `noon-ads` § 7.1. **This export is, however, the SKU bridge for both.** Its `SKUs` column is the noon-internal key (`Z…Z-<n>`) and `Partner SKUs` the seller's own code, on the same row — the only place the two keyspaces meet. Keep the transaction CSV alongside the fee/ads files; without it the per-SKU numbers cannot be mapped to seller SKUs. ## See also - `noon-shared` — login, page structure (prerequisite) - `noon-fbn` § 8 + `references/fee-reports.md` — per-SKU FBN storage / RTV-removal fees, and the reconciliation against § 2's settlement line - `noon-ads` § 7.1 — per-SKU ad spend export - `amazon-reports` — Amazon equivalent + general "Download Behavior" pattern