Very slow in requests to the groq model using ia agent or Basic llm chain

[Cesar_Sanchez]

I have a problem integrating the groq model with the agent (tool agent and conversational agent and basic llm chain), the query to the groq model (any model to use is very slow, takes up to 21 seconds and sometimes gives errors), I have used without problems other open router, gemini and openia models and the queries take a matter of 4 or 5 seconds, in the case of groq the queries take up to 21 seconds to answer easy questions.

If your question hasn’t been asked before, please follow the template below. Skip the questions that are not relevant to you. →

Additionally, I get an error in the portainer log where I have n8n installed indicating the following:

Other images:

• **n8n version: 1.80.4
• **Database (default: SQLite): Default SQLite and Postgres (local)
• **n8n EXECUTIONS_PROCESS setting (default: own, main):own
• **Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
• **Operating system: debian 12

[Knight]

I have copied you’re workflow into my own n8n instance, after adding the necessary credentials it worked:

Are you sure that you’ve added the correct credentials to the Groq node?

[Cesar_Sanchez]

Yes, the query is correctly made to the groq model. The problem is that the groq model takes up to 21 seconds to respond, taking into account that the groq models are faster than normal. A delay of such a long time may be the result of a bug in n8n.

[Cesar_Sanchez]

[Yo_its_prakash]

The delay in response might totaly be from the groq side, if you feel it is from n8n try updating to 1.80.5 latest,

Further debugging go to you groq playground and check with similar prompts in same model
Create another groq account and test it with that API

[Cesar_Sanchez]

-Make the same request towards the same groq model, case 1 made from n8n version 1.80.4, response delay is 21.73 seconds.

-Case 2 make the request through postman using the same model, the response took 2.6 seconds.

[Knight]

This might be a far stretch, but are the servers for Groq and your n8n instance in 2 different regions?

[Cesar_Sanchez]

I have the n8n installed locally on a PC that I have exclusively for these purposes, both queries that I showed in the images come from my same network

[Cesar_Sanchez]

I was testing several versions of n8n with respect to queries in groq and their respective response time using an llmchain chat, here are the results:

I was testing several versions of n8n with respect to queries in groq and their respective response time using an llmchain chat, here are the results:

1.80.4 takes 21 seconds
1.69.2 takes 21 seconds
1.68.1 takes 21 seconds
1.68.0 takes 21 seconds
1.67.1 takes 1.1 seconds
1.65.2 takes 1.2 seconds
1.63.1 takes 1.2 seconds