N8n Cloud 1.100.1] Persistent Issues with Google Gemini API & Code Node for External HTTP Requests

[Down_Fall]

Hi n8n Community and Support Team,

I’m encountering a very stubborn issue with my n8n Cloud instance (URL: hayatechaofen.app.n8n.cloud, Version: 1.100.1 (Cloud)) when trying to make external API requests. My goal is to build a workflow that calls both the Google Gemini Generative Language API (for script generation) and the ElevenLabs Text-to-Speech API (for voiceovers).

Good News: The ElevenLabs Text-to-Speech API call is now successfully working within my n8n workflow, and the generated audio is correctly uploaded to Google Drive. This confirms my ElevenLabs API key, account, and the Google Drive connection are functioning.

The CURRENT persistent issue is specifically with the Google Gemini API call.

Problem Description (Focus on Gemini): I am attempting to send a POST request with a JSON body to the Google Gemini API (using models/gemini-2.5-flash:generateContent) from the first HTTP Request node in my workflow. This request consistently fails.

Specific Issues Encountered:

1. Using the HTTP Request Node (for Google Gemini API):

• Error: The node fails with a 400 Bad Request and Invalid value at 'contents'. The detailed API response from Google indicates that it’s receiving the contents value as a literal string (including JSON.stringify(...) and backslashes), instead of a pure JSON array/object.
• Observation: This occurs despite configuring Body Content Type to JSON and Specify Body to Using Fields Below, and providing the raw JavaScript array literal directly in the Value field for contents (as per n8n documentation). It seems n8n is incorrectly serializing the JSON body.

2. Using the Code Node (as an alternative for HTTP requests - for future needs):

• Error 1: When attempting to use the standard JavaScript fetch API within the Code node, the workflow fails with "fetch is not defined".
• Error 2: When attempting to use n8n’s built-in n8n.httpRequest function within the Code node, the workflow fails with "n8n is not defined".
• This suggests a highly restricted or misconfigured JavaScript runtime environment for the Code node on this instance, preventing any external HTTP calls from there.

3. “Import cURL” Feature in HTTP Request Node:

• Error: When trying to import a working PowerShell Invoke-RestMethod script (which successfully calls Gemini externally), the “Import cURL” feature fails with "Couldn't import cURL command. This command is in an unsupported format". This implies it cannot parse PowerShell syntax.

Troubleshooting Steps Already Taken:

• External API Tests (Successful for both Gemini & ElevenLabs via PowerShell):
  • Successfully called ElevenLabs API via PowerShell, confirming API key, setup, and binary audio output.
  • Successfully called Google Gemini API via PowerShell (for a simple prompt) using Invoke-RestMethod, confirming API key and basic connection. Also successfully generated the full script using Invoke-RestMethod in PowerShell. This confirms both APIs work externally.
• API Keys & Accounts: Tried multiple ElevenLabs API keys (including a paid subscription key) and multiple Google Gemini API keys (newly generated from different projects).
• Voice IDs & Models: Tested with various voice IDs (Bella, Daniel) and Gemini models (gemini-pro, gemini-2.5-flash).
• Google Drive Connection (WORKING): Successfully connected and uploaded files to Google Drive, resolving 403 Forbidden errors for OAuth scopes and 403 Access Blocked for test user configuration.
• n8n UI & Cache: Performed multiple hard refreshes of the n8n UI and cleared browser cache/cookies for the n8n domain.
• n8n Logs: The live “Logs” panel at the bottom of the n8n UI consistently shows “Nothing to display yet” even when an error occurs. Detailed errors can only be retrieved from the “Executions” tab history.

It appears there is a fundamental and persistent issue with how my specific n8n Cloud instance handles external HTTP requests (especially with complex JSON bodies) from the HTTP Request node, and also its Code node’s runtime environment.

I have working API keys and can provide them privately if requested by support.

Any assistance in diagnosing and resolving these issues would be greatly appreciated.

Thank you!

Best regards,

Hayate

[fahmiiireza]

Hi there, so from my understanding the elevenlabs is now working, and the issue is nly on the Google Gemini Generative Language API?

can you give me the import curl that you use that you cannot import?

[cutecatcode]

Hi @Down_Fall! Welcome to community!

Since you want use Gemini for script generation, I assume your Gemini node called before ElevenLabs API calls.

You can use Basic LLM Chain for that case, it doesn’t need big configuration to do that, and below is the sample for your case.

Hope it helps! And if it helps, kindly mark my message as solution, thanks!

[Down_Fall]

Hi @cutecatcode,

Thank you so, so much for your solution! It worked perfectly!

The workflow you provided, especially using the Basic LLM Chain for Google Gemini, successfully bypassed the persistent issues I was having with the standard HTTP Request node’s JSON serialization.

Thanks to your help:

• Google Gemini (via Basic LLM Chain) is now successfully generating my full scripts!
• ElevenLabs is also now successfully generating voiceovers based on those scripts!

This is a huge breakthrough for my project, and I’m incredibly grateful for your clear and effective solution.

Yes, it definitely helps! I will gladly mark your message as a solution.

Thanks again for your awesome contribution to the community!

Best regards,

Hayate

[cutecatcode]

You’re welcome!

[Down_Fall]

Hi Fahmilireza,

Thanks so much for getting back to me!

Good news: I’ve made significant progress!

1. ElevenLabs (Text-to-Speech) API: Yes, my ElevenLabs integration is now successfully working within n8n! I’m able to generate voiceovers and get the audio.
2. Google Gemini Generative Language API (Script Generation): This is also now successfully working within n8n! The workflow from the community you and another user provided, using the “Basic LLM Chain” node, successfully bypasses the issue I was having with the standard HTTP Request node. Gemini is now generating my full scripts.

So, the core script generation and voiceover pipeline is now functional, thanks to the community’s help!

Regarding the curl command you asked for, this is the PowerShell script I was trying to import that n8n’s “Import cURL” feature rejected as “unsupported format.” It’s the one I use directly in PowerShell to test the full script generation from Gemini:

$gemini_api_key = “YOUR_GOOGLE_GEMINI_API_KEY_HERE”
$headers = @{
“Content-Type” = “application/json”
}
$body = @’
{
“contents”: [
{
“parts”: [
{
“text”: "You are a viral content strategist for YouTube Shorts and TikTok. Your goal is to create a compelling, actionable script for a 1-minute video that helps creators .
}
]
}
]
}
'@

try {
Invoke-RestMethod -Method POST -Uri “https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=$gemini_api_key” -Headers $headers -Body $body -ErrorAction Stop | ConvertTo-Json -Depth 100
} catch {
Write-Host “— Error from Gemini API —”
Write-Host “Status Code:” $.Exception.Response.StatusCode
Write-Host “Status Description:” $.Exception.Response.StatusDescription
Write-Host “Response Body (Detailed Error):”
(New-Object System.IO.StreamReader($_.Exception.Response.GetResponseStream())).ReadToEnd() | ConvertFrom-Json | ConvertTo-Json -Depth 100
Write-Host “-----------------------------”
}