I want to extract file from ODS or XLSX, unfortunately it extracts only 2 rows, the header and first data row. What should be set in order to extract entire table from xlsx/ods ?
Just replace your expression {{Object.keys($binary)[0]}} with data, unless you want something specifically with that file.
Usually outputs the data as a series of JSON row objects.
It didn’t work. I changed to “data0” as You can see on the picture, and still te same, only 2 rows. I also attach the table from original file
Ahhh… just seen that actually have 11 items , you just need to split or item list to get individual rows?
The output is typically an array of items.
I want to insert data extracted from the table to AI agent. I want to prompt agent to analyze row by row and construct text based on the whole table. The output text would be technical description
To achieve your goal in n8n—analyzing a table row by row with an AI agent and generating a technical description based on the whole table—you can follow this general approach:
1. **Extract Data from the Table**
Use the appropriate node (e.g., Google Sheets, database, or CSV node) to fetch all rows from your table.
2. **Aggregate or Format the Data**
If you want the AI agent to consider the entire table at once, aggregate the rows into a single object or string. The [Aggregation node]( Aggregate | n8n Docs ) can help you merge all rows into one item, which is useful for summarization or analysis tasks. This is a common pattern to avoid sending excessive data to the AI model and to provide a concise context for analysis. For example, you can aggregate all rows and then summarize them before sending to the AI agent, as shown in the ERP AI chatbot example: first aggregate, then summarize, then use as context for the agent’s prompt [ERP AI chatbot guide]( ERP AI Chatbot: Advanced ERP Automation with AI – n8n Blog ).
3. **Prompt the AI Agent**
Pass the aggregated data to the AI Agent node. In your prompt, instruct the agent to analyze the data and construct a technical description. For example, you might use a prompt like:
```
Analyze the following table data and generate a technical description: {{ $json.data }}
```
This approach is also used in the ERP AI chatbot, where the context is summarized and then provided to the agent for further analysis or question answering [ERP AI chatbot implementation]( ERP AI Chatbot: Advanced ERP Automation with AI – n8n Blog ).
4. **Output the Result**
The AI Agent node will return the generated technical description, which you can then use in subsequent workflow steps (e.g., update a document, send an email, etc.).
**Tips:**
- If your table is very large, consider summarizing or chunking the data before sending it to the AI agent to avoid context window limitations.
- You can use the [Window Buffer Memory node]( Simple Memory node documentation | n8n Docs ) to help the agent remember previous interactions if you want to process the table in parts.
**Example Workflow Reference:**
This pattern—aggregate table data, summarize, and then analyze with an AI agent—is demonstrated in the [ERP AI chatbot example]( ERP AI Chatbot: Advanced ERP Automation with AI – n8n Blog ).
If you need a concrete workflow template, you can adapt the ERP AI chatbot or the [AI Data Analyst Agent]( AI Data Analyst Agent and Visualization Agent for Large Spreadsheets | n8n workflow template ) to your use case.
P.S if you achieve this, you can pass the course levels as well.
We are here to help in every way so you understand, not just “fixers” . 
Ok, but the question is what node should I use in my case? As you can see I used “Extract from file” node, and it reads only 2 rows. The xlsx file is saved on local disc, then it is uploaded to the workflow via chat message node → “extract from file” node.
Please share your workflow here as code… is strange whats happening in your case, or i dony fully understand.
I’m sorry for my misunderstanding, I’ve just realised what “11" items” mean. When I switched from “schema” to “table”, the whole table was shown. I’m very beginner. Thank You for Your tips, I have a lot of work to do. I’ll analyze Your tips in order to avoid stupid my stupid questions 
Glad that you understood!
Thats the way to learn
Mark answer as solved so the topic will be relevant to others.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.
