From 69c447ac592b51b88586a75f227b4e53d27c19c5 Mon Sep 17 00:00:00 2001 From: Deborah Date: Thu, 16 May 2024 14:42:21 +0100 Subject: [PATCH] Update docs/integrations/builtin/app-nodes/n8n-nodes-langchain.openai.md --- .../builtin/app-nodes/n8n-nodes-langchain.openai.md | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/docs/integrations/builtin/app-nodes/n8n-nodes-langchain.openai.md b/docs/integrations/builtin/app-nodes/n8n-nodes-langchain.openai.md index 9e96f599d..86cf1078c 100644 --- a/docs/integrations/builtin/app-nodes/n8n-nodes-langchain.openai.md +++ b/docs/integrations/builtin/app-nodes/n8n-nodes-langchain.openai.md @@ -76,4 +76,6 @@ Once you add a tool connection, the OpenAI node becomes a root node, allowing it ## Using memory with OpenAI assistants -For the **Assistant** > **Message Assistant** operation, you can connect a memory sub-node to preserve and retrieve chat history. This allows the assistant to maintain context across multiple messages. The connected memory sub-node is the source of truth for the assistant's memory. This means that n8n creates a new thread on each run and pre-populates it with messages from the memory sub-node. After the run finishes, n8n updates the memory sub-node with the new messages, and deletes the thread from OpenAI. +For the **Message Assistant** operation, you can connect a memory sub-node to preserve and retrieve chat history. The assistant uses this to maintain context across multiple messages. The connected memory sub-node is the source of truth for the assistant's memory. + +To do this, n8n uses OpenAI's [threads](https://platform.openai.com/docs/assistants/how-it-works/managing-threads-and-messages){:target=_blank .external-link}. n8n creates a new thread on each time the node executes, and pre-populates it with messages from the memory sub-node. After the run finishes, n8n updates the memory sub-node with the new messages, and deletes the thread from OpenAI.