> For the complete documentation index, see [llms.txt](https://docs.borndigital.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.borndigital.ai/digital-agent/conversation-flow/nodes-explained/ai-node/migrating-from-the-old-ai-node.md).

# Migrating from the old AI node

How to switch an existing AI node to AI node 2.0 safely.

AI nodes created before AI node 2.0 keep the old version until you switch them. New AI nodes always start as 2.0.

## What changes

|                                   | Old AI node                                   | AI node 2.0                                                                               |
| --------------------------------- | --------------------------------------------- | ----------------------------------------------------------------------------------------- |
| LLM API                           | Chat Completions                              | Responses                                                                                 |
| Tools                             | One knowledge-base tool                       | Knowledge base, Web search, Redirect to human, Transfer to node, End process, MCP servers |
| Files in user input               | No                                            | Yes                                                                                       |
| Temperature                       | 0–1                                           | 0–2                                                                                       |
| Reasoning settings                | No                                            | Yes                                                                                       |
| `indexer_returned` variable       | Filled                                        | Not filled — use `mcp_call` or `kb_document_*`                                            |
| `gpt_indexer_returned` variable   | Raw indexer response                          | Not filled                                                                                |
| `kb_document_id`, `_name`, `_url` | The one document used, overwritten every time | A list of all documents used in the conversation                                          |
| `function_call` in `gpt_response` | The tool the model called                     | Not used — see `mcp_call` / `mcp_calls` (MCP servers) and `target_node` (transfer tools)  |
| Smart function in the code editor | `gpt_stream`                                  | `llm_agent`                                                                               |

## How to switch

1. Note down the knowledge-base settings of the node (index, number of snippets, adjacent snippets).
2. Turn on the **2.0** switch in the header of the editor.
3. Open **Essential → Tools** and add the **Knowledge base** tool again — the knowledge-base configuration of the old node is **not** carried over.
4. Check **Configuration → Preparing the response**: temperature (the scale is now 0–2), and **Effort** if you use a reasoning model.
5. If the flow reads `indexer_returned`, change it to `{gpt_response['mcp_call']}` (Indexer MCP server) or to the `kb_document_*` variables.
6. Train the project and test the node with real questions before deploying it — including the questions the node should refuse.

{% hint style="warning" %}
Switch one node at a time and test it. Answers can differ between the two versions even with the same Instructions and model.
{% endhint %}

To go back, turn the **2.0** switch off.
