Integrations / NEWO
NEWO
Connect Feedpipe to NEWO
The Feedpipe module for NEWO lets the agent search your catalog and use the found facts in customer answers: titles, prices, stock, specs, and links.
The customer describes the product they need. The agent clarifies when required, then calls Feedpipe. Instead of the full catalog, the agent receives only the matched products.
The module talks to the Feedpipe API directly. The agent uses data from the latest successful catalog collection in Feedpipe. Shop changes become searchable after the feed refreshes and Feedpipe collects again.
Before you start
You will need:
- a NEWO project with rights to add modules and publish;
- a Feedpipe shop with products already loaded;
- the Feedpipe API base URL;
- the shop identifier — Shop IDN;
- an API key with catalog access.
If you have not added a shop and run collection yet, follow the Start guide first.
Step 1. Add the module
Download the Feedpipe module for NEWO and unpack the archive.
Step 2. Add the module to the NEWO project
Add the unpacked module to your NEWO project and publish the project.
On publish, the module creates the catalog connection fields. They belong to the Module — CatalogIntegration group.
Fill these fields before you test the agent.
Step 3. Connect to Feedpipe
Set the three required parameters:
| Field | What to enter |
|---|---|
catalog_api_base | Feedpipe API base HTTPS URL |
catalog_shop_idn | Your shop identifier in Feedpipe |
catalog_api_token | API key with catalog access |
API base URL
In catalog_api_base, enter only the server base URL: no /v1, no search path, and no trailing slash.
Example format:
https://feedpipe.app
That is the production API. If you were given a different URL for your connection, use that instead.
Shop identifier
In catalog_shop_idn, paste the exact Shop IDN for your shop. The shop name and website URL do not replace this identifier.
API key
In catalog_api_token, paste a key that can access the target catalog.
Keep the key secret: do not put it in agent instructions, customer messages, or public materials.
Step 4. Enable search and set the phrase
Check the extra module fields:
| Field | Purpose | Default |
|---|---|---|
catalog_search_enabled | Turns product search on or off | True |
catalog_code_phrase | Phrase tied to starting search | I'll check stock. |
catalog_setup_scenarios | Auto-adds catalog scenarios to Canvas on publish | True |
For a standard setup, leave catalog_search_enabled and catalog_setup_scenarios as True.
Search trigger phrase
By default the agent uses:
I'll check stock.
This is the agent’s own phrase, not a command the customer must say. It is tied to the catalog_search tool call.
If you change the phrase, keep it aligned with the agent instructions and scenarios. Casual rephrasing can block search from starting.
After you change settings, save and publish the project again.
Step 5. Test search in a dialog
Open a test dialog with your agent and ask about a product that is definitely in the loaded catalog.
For example:
Do you have a Bosch cordless screwdriver?
Expected flow:
- The agent figures out which product the customer needs.
- If details are missing, it asks a clarifying question.
- It uses the configured phrase and starts search.
- It receives the Feedpipe response.
- It tells the customer the found facts.
In the current version the module requests up to five products per search.
Search by SKU
To test an exact identifier, use a SKU from your catalog:
Check the product with SKU 44102.
In this example 44102 is a placeholder. Replace it with a real SKU.
The module treats the shop SKU and the model name as different things. For SKU search, the customer should say “SKU”, “article”, or “product code” explicitly.
Search with a budget
Add a price limit to the product description:
Find a cordless screwdriver under 3,000 UAH.
The module supports lower and upper price bounds. Numeric limits use the catalog currency, so test with that currency.
Clarifying the query
In the current version, text search starts when the formed query has at least two meaningful words. The alternative is an exact product SKU.
So after a vague request like “I need a screwdriver,” the agent may first ask for brand, model, or required specs.
How the agent uses found data
Feedpipe returns facts the agent turns into an answer. How products are presented depends on the NEWO scenario and the channel.
Check that the agent:
- names prices and specs only from the returned data;
- distinguishes in-stock, out-of-stock, and stock-not-set products;
- does not confirm stock when the status is unknown;
- says when nothing was found;
- does not invent products or specs in place of missing data.
A missing product in the search response does not always mean it is absent from the shop. The cause can be the query wording, incomplete feed data, or an unfinished catalog collection.
Pre-launch checklist
| Check | What to verify |
|---|---|
| Title or model | The agent finds the right product and names it correctly |
| Exact SKU | The found product matches the SKU |
| Price limit | Prices match the stated budget |
| Stock | The agent reports the catalog stock status correctly |
| Vague request | The agent clarifies what the customer needs |
| Product not found | The agent says so without inventing options |
| Connection error | The agent reports a problem instead of treating it as “no products” |
Test the integration on real products from your shop. Compare answers with the latest Feedpipe collection. If you compare against the shop website, allow for changes after that collection.
Updating the catalog
The NEWO module does not download or process the product feed itself. It queries the catalog already prepared in Feedpipe.
Both stages must work for updates:
- The shop refreshes the product feed.
- Feedpipe collects the updated data from that feed.
After a successful collection, the new data is available for the next agent search requests.