Help center

Getting started

CerePilot is a workbench for product data collection + central database + AI copy optimization + one-click export, covering Amazon, Temu, Fruugo, AliExpress, eBay and 1688.

Typical uses: sourcing research, supplier costing via 1688 image search, listing prep (export Excel + images), AI copy rewriting, and team collaboration with role-based visibility.

  1. Sign in with your account.
  2. Install WebPick from the extension page (download ZIP, load unpacked at chrome://extensions).
  3. Sign in inside the extension popup.
  4. Collect your first item on any supported product detail page.
  5. View results in the central data pool.
  1. Collect 5–10 products with the extension.
  2. Open the central data pool and find them.
  3. Click "Export" → check "embed cover image" → "all URLs".
  4. The system refreshes each row in the background, then downloads the .xlsx.

Need original images? Choose "download image folder" to export Excel plus an images/ folder.

Left navigation (top to bottom):

  • Smart Sourcing: bulk reverse-image search on 1688 to find sourcing options, purchase price, packaging size and weight
  • Product library (default landing page): manage products by platform — AI copy optimization and risk scanning happen here
  • Central data pool: everything you've collected across platforms, with multi-dimension filtering and bulk export

There is no separate "collection" page — collecting is initiated from the WebPick extension and lands straight into the two libraries above.

User entry at the bottom-left (click your username):

  • Your username: opens the account center (profile, preferences, my teams, invites, security, close account)
  • Organization (owner / manager only): members, roles, data sharing
  • Task manager: collection, export and AI task progress
  • Settings, Help, Download latest WebPick, Sign out

Not seeing an entry usually isn't a bug: some modules are gated by plan features or role, and a few platform-operations pages are visible only to platform super admins.

What AI can do

What you collect is raw data — listing it as-is hits three traps: copy that looks like the original seller's, hot-linked images, and unchecked risk words. CerePilot's AI fills all three:

  • Write your own original copy: two-stage rewrite of title / description / bullets into a differentiated version instead of copying the original seller — exactly what marketplaces' anti duplicate/clone rules care about most.
  • Swap in your own image links: at export, convert images to stable links on your own domain, avoiding "cloned image" flags and broken images.
  • Screen before listing: IP risk scanning checks brand words and ™®© symbols to stop risk before it goes live.

In short: AI turns "someone else's collected data" into "your own compliant, original, safe-to-list listing".

Fruugo treats duplicate / cloned listings with zero tolerance: copying another seller's title, description or images gets flagged as a clone — unrecoverable, with a per-SKU fine, at the same severity as prohibited goods. AI addresses the common triggers:

  • Keyword-stuffed titles → AI rewrites titles, strips redundant keywords, leads with the core category term.
  • Copied / thin descriptions → AI rewrites around real selling points and adds info the title doesn't cover, not just swapping a few words.
  • Hot-linked images → use "Self-hosted image links" to serve your own.

Note: AI significantly lowers duplicate-flag risk but doesn't replace your own judgement of platform rules; compliance is subject to the platform's review.

Two stages: a vision model reads the product, then a text model rewrites. It can optimize titles (no keyword stuffing, lead with selling points, match marketplace tone), descriptions / bullets (deep, differentiated rewrite around real selling points), and multiple languages. Results are isolated per team and form a rewrite → re-scan loop with IP scanning.

  • Compliance: marketplaces like Fruugo treat hot-linking others' images as clone evidence and require your own images. Links on your own domain keep the image source independent.
  • Stability: once listed you no longer depend on the origin platform, so images won't expire, break or get blocked.

At export, pick one of three: original, CerePilot-hosted, or Aliyun OSS public link (over CDN, faster). Links carry no account keys.

WebPick extension

Browsers: Chrome and all Chromium-based browsers (Edge, Brave, Arc, etc.). Firefox and Safari are not supported.

Platforms: Amazon, Temu, Fruugo, AliExpress, eBay and 1688, on product / search / category pages.

  1. Download the latest ZIP from the extension page.
  2. Unzip to a fixed folder (don't delete it later).
  3. Go to chrome://extensions and enable Developer mode.
  4. Click "Load unpacked" and choose the folder.

To upgrade, unzip the new version over the same folder and click the refresh icon on the extension card.

  1. Not signed in in the extension popup — sign in and refresh the product page.
  2. Unsupported page — the button only shows on supported product/search/category pages.
  3. Can't reach the server — check your connection.
  4. Extension disabled — enable it in browser settings.
  5. Page not refreshed after install/sign-in — refresh once.

Every collection lands in the central data pool — whether it came from the WebPick extension, bulk URL paste or file import.

Path: sign in → left menu → "Central data pool".

There you can search by keyword (title / URL / product ID / category), filter by platform, see collection time, price and thumbnails, open a row for details (edit URL, delete, view the variant group), and export to Excel from the top-right.

Visibility: you see what you collected plus what your group shares. The exact scope depends on your role and team setup — see "Teams & roles".

  1. Open any CerePilot page in the browser
  2. Sign in again with your account
  3. Go back to the marketplace page and press F5
  4. The extension picks up the new session automatically

If you're clearly signed in but the extension still says otherwise, it's usually pointed at a different CerePilot server. Open the extension popup and check the "connected to" address at the top.

The extension picks a server automatically: a local instance during development, a LAN server inside a company network, otherwise the cloud at cerepilot.net.

Check it: click the extension icon — the popup shows the current connection at the top.

Switch manually: popup → Settings → enter the correct server address. It will keep using that one from then on.

Data management

Two ways: Standard Excel (optionally embed cover image, single .xlsx) and Excel + image folder (Excel plus an images/ folder of full-size images). Export scope can be the current page or all URLs; the system refreshes each row before generating the file.

No. Products are de-duplicated so re-collecting updates the existing record instead of creating a second one. Variants of the same main product are grouped together automatically.

Usually the source link expired — click "Update" to re-collect. Strict-privacy browser modes can also block images; for empty images in Excel, re-export with "embed cover image" checked.

What it is: a variant group ties together every variant of one parent product (colour, size, bundle…). A tee with 5 colours × 3 sizes = 15 separate URLs; the group links all 15.

Where: open the parent product in the central data pool — the "variant group" panel lists every linked URL. Variants are expanded into the Excel on export.

Manual sync: if you collected a new variant and the group didn't pick it up, hit "sync variant group" on the detail page.

Every row in the central data pool has a refresh button that re-collects that product immediately, skipping the queue and any cooldown.

Typical cases: the price or stock just changed; an earlier collection failed and you don't want to wait; a final refresh before exporting; images stopped showing (usually the source swapped them).

Note: exporting to Excel already refreshes each row automatically — you don't need to do it by hand first.

Collection tasks

Open the Task Manager from the user entry at the bottom-left of the dashboard. Each URL shows a status: Pending, Processing, Completed, Failed or Paused. Failed rows include an error reason.

Four ways: single collection via the extension, paste one or more URLs, upload a .txt/.csv file, or use the extension's "collect this page" on list pages. All feed the same queue.

Yes — it retries automatically with increasing cooldown, up to 3 times. You can also click "Retry" to skip the cooldown, or "Retry all failed" for a batch.

That's a protection mechanism: when Fruugo trips anti-bot defences repeatedly in a short window, the system pauses the queue rather than burning resources on failures.

What to do: screenshot the pause notice in the task manager and email cerepilot@outlook.com. Once it's cleared on our side the queue resumes on its own — paused tasks restart automatically, you don't need to resubmit.

Temu can only be collected through the WebPick extension in your browser — bulk URL paste and file import don't work for it.

  1. Install WebPick and sign in from the popup
  2. Open temu.com and go to the product detail page
  3. Wait for the page to fully load — main image, price, specs and stock all visible (don't click while it's still spinning)
  4. Click the floating collect button — data is saved immediately
  5. Find it in the central data pool

Tips: for products with many variants, switch a variant on the Temu page first so the data loads, then collect — you'll get fuller variant info. If the button does nothing, reload and retry. Temu items don't appear in the task manager queue because the extension writes them directly — that's expected.

Smart Sourcing

It takes products you've already collected (the ones with a main image) and runs them through 1688 reverse-image search in bulk, surfacing candidate suppliers and reference purchase prices. You then confirm the final purchase price, which is written back into the system.

It answers two questions: what would this item cost me to source, and is there enough margin? and who actually supplies this bestseller, and at roughly what wholesale price?

Entry: "Smart Sourcing" in the left navigation.

Note: this is sourcing-cost analysis, not a selling-price or margin calculator. Selling price is yours to handle after export.

Smart Sourcing uses your browser as a costing node to query 1688 live, so there's a one-time setup:

  1. Keep WebPick up to date — older builds don't support costing. Grab the latest from the extension page.
  2. Sign in to 1688 in your browser (any free account works).
  3. Open the extension popup and turn on the "costing node" toggle.
  4. Keep at least one 1688 tab open — a background tab is fine, it doesn't need to be in front.

Ready check: the popup should show a green "costing node: online" badge.

  1. Open "Smart Sourcing" and click Bulk create at the top right
  2. Pick a platform (Amazon / Fruugo …) and a cap (e.g. 100)
  3. The system picks URLs that are ① visible to you ② have a main image ③ haven't been costed yet
  4. Once created, every online costing node — yours and other users' browsers — starts picking up work
  5. Rows move from "awaiting search" → "searching" → "awaiting costing" as the page refreshes itself

Rough timing: with 4 nodes running, 100 rows finish in about 1–2 minutes.

  • All: every unconfirmed task
  • Awaiting costing: image search done, waiting for you to confirm a price
  • No match: ones you marked as "no supplier found on 1688" — archived out of your to-do list
  • Done: tasks where you've confirmed the price

Each row shows 1688 candidate thumbnails on the right (the second one is selected by default — the first is usually an ad slot).

  1. Pick the candidate that genuinely matches (click to switch)
  2. Check the reference price pre-filled in the "purchase price" box on the left
  3. Enter shipping (a separate field — domestic leg)
  4. Enter weight in grams (saves automatically)
  5. Products with variants expand into sub-rows; each variant takes its own bare price, sharing the parent's shipping
  6. Click Confirm price

Price rule: the "purchase price" box takes the bare price excluding shipping. What's saved to the central data pool as "purchase cost" is bare price + shipping (true landed cost), and the "Purchase Price" column in your Excel export includes shipping too.

  • Search again: hit "re-search" on the row. Already-confirmed tasks can trigger it too — the old data stays until new results arrive.
  • Genuinely not on 1688: mark it "no match" to archive it out of your to-do list.

Common messages: "no matching product on 1688" (nothing visually similar exists), "1688 API error" (intermittent — a retry usually clears it), "invalid image format" (1688 can't read the main image), "no 1688 tab" (keep one open), "1688 session expired" (sign in again).

The extension popup shows the reason. Three common ones:

  • "No 1688 tab": open any page on 1688.com (search results or a product page) and leave it in a tab — it doesn't need to be in front. The node comes back online by itself.
  • "Verification required": 1688 triggered a risk check (slider or SMS). Complete it manually on 1688, or sign in with a different account.
  • "Repeated failures": the system suspends the node to stop dispatching work that won't succeed. Check the two items above, then hit "reset" in the popup or restart the browser.

If it permanently shows "no heartbeat", the extension is usually too old — install the latest from the extension page.

Plan & credits

Path: user entry at the bottom-left of the dashboard → click your username → Account center → "My teams" → your team card.

The card shows your remaining credits and usage for the cycle. Credits are shared across the team: every member's usage draws from the owner's single pool.

Path: Account center → "My teams" → team card → "Renew plan" → pick a plan → complete payment in the new tab.

Note: these buttons are only visible to the team owner. If you're a manager or member they won't appear — ask your owner to do it.

Path: Account center → "My teams" → team card → "Booster" → pick a size → pay. Credits land immediately.

Two sizes are available (3,000 and 13,000 credits); the exact price is shown in the picker before you pay.

Free plans can't buy one — you'll be prompted to upgrade to Pro / Max first. That's a plan limit, not a bug.

Your balance is actually two buckets with different rules:

  • Plan credits: reset each billing cycle — unused ones don't roll over.
  • Booster credits: never expire — they stay until spent.

Spending order is "plan first, Booster second", so the credits that would expire get used up before the ones you paid for.

Collecting is free — extension, bulk import and manual paste cost no credits; they only count toward your collection quota.

These do spend credits:

  • AI copy optimization: 1 listing = 1 credit
  • Image link conversion: 1 image = 1 credit (on export)
  • AI vision costing: 1 costing task = 1 credit, plus 0.5 per variant under it

Batch actions always show a total estimate for you to confirm before spending.

Path: Account center → "My teams" → team card → "Credit ledger".

Entries are newest-first with time / type / change / balance after, and you can page further back. Team members' usage shows up here too.

Charges are invoiced as a software license / technical service fee.

There's no self-service invoicing in the app yet — email cerepilot@outlook.com with your order details, invoice title, tax ID and receiving email, and we'll issue it manually.

Teams & roles
  • Owner — team creator; can add managers and members, configure feature toggles, and see all team data.
  • Manager — can add members to their own group and control cross-group sharing.
  • Member — runs collection; sees their own data plus shared data.

Members contact their manager or owner to reset; owners contact us at cerepilot@outlook.com with the registered username.

Two ways:

  • Created or invited by your team: an owner can create managers and members under "Organization", or invite an existing user to join the team (the invitee confirms).
  • Invited into a team: accepting an invite adds that team to your account — your existing identity and data are untouched.

Setting up a new team from scratch? Email cerepilot@outlook.com and we'll get you started.

Error self-check
  1. Session expired — sign in again.
  2. Stale cache — hard refresh (Cmd/Ctrl+Shift+R).
  3. Network — check connectivity; try turning off VPN.

Your session likely expired. Sign in again; for the extension, refresh the product page so it re-reads your credentials.

Either your role lacks permission for that action, or a target site temporarily blocked the request (the system retries with increasing cooldown).

Export runs in the background and refreshes each row first (about 2–5 minutes for 100 products). Check the Task Manager for progress; re-download from there if the browser closed the download.

What it means: something broke inside the system and needs the CerePilot team to look at it — there's not much you can do from your side.

What helps: note which button you clicked and the exact message, screenshot it (include the address bar) and email cerepilot@outlook.com with the time it happened. Don't hammer the same button — it can make things worse.

  • Everything is cooling down: after a failure the system waits minutes to an hour before retrying — no movement is normal. Wait, or hit "retry now" on a row to skip the cooldown.
  • Collection is paused: check the collection toggle in the task manager, and email cerepilot@outlook.com if needed.
  • The batch was auto-halted: see "Why are Fruugo tasks paused" under Collection tasks.
  • None of the above: screenshot the task manager and email cerepilot@outlook.com.

What it means: the system temporarily can't reach a network service it needs. This is infrastructure-level and does not mark your task as finally failed — it retries automatically after a short wait.

What helps: give it a few minutes. If it keeps erroring, screenshot and email cerepilot@outlook.com. This class of error relates to server-side networking and isn't something you can fix directly.

What it means: Amazon has flagged the request as automated and wants a verification challenge.

What happens: the system automatically slows its retry cadence (10 min → 30 min → 1 hour) so Amazon lifts the block on its own. Don't retry manually straight away — it makes things worse. If a whole batch of Amazon items fails, email cerepilot@outlook.com to check the collection environment.