What's Going Wrong?

I am experiencing a persistent HTTP 400: API_KEY_INVALID error when trying to use the native Gemini provider with Google's OpenAI-compatible endpoint.

Important Context: * My GEMINI_API_KEY is 100% valid. It works perfectly in OpenClaw and direct curl requests.

It worked perfectly fine this morning. The issue only started happening immediately after I updated to the newest version of Hermes.

I have already completely cleared my WSL environment, deleted ~/.hermes, did a fresh install, and ensured there are no conflicting keys (like AI_GATEWAY_API_KEY) in the .env file. The issue persists.

It seems the latest update introduced a bug in how Hermes formats or passes the API key to the generativelanguage.googleapis.com/v1beta/openai endpoint (possibly an issue with the Bearer token header or URL query construction).

Environment & Setup:

Provider: gemini

Model: gemini-3.1-flash-lite-preview

Endpoint: https://generativelanguage.googleapis.com/v1beta/openai API call failed (attempt 1/3): BadRequestError [HTTP 400] 🔌 Provider: gemini Model: gemini-3.1-flash-lite-preview 🌐 Endpoint: https://generativelanguage.googleapis.com/v1beta/openai 📝 Error: HTTP 400: Error code: 400 - [{'error': {'code': 400, 'message': 'API key not valid. Please pass a valid API key.', 'status': 'INVALID_ARGUMENT', 'details': [{'@type': 'type.googleapis.com/google.rpc.ErrorInfo', 'reason': 'API_KEY_INVALID', 'domain': 'googleapis.com', 'metadata': {'service': 'generativelanguage.googleapis.com'}}, {'@type': 'type.googleapis.com/google.rpc.LocalizedMessage', 'locale': 'en-US', 'message': 'API key not valid. Please pass a valid API key.'}]}}] ❌ Non-retryable client error (HTTP 400). Aborting.

I followed someone‘s instructions and ran the curl test. It worked perfectly and returned a valid response. I tested it with the model gemma-4-31b-it. Here is the curl output (using the Bearer token authorization): code JSON {"choices":[{"finish_reason":"stop","index":0,"message":{"content":"The user said "hi".\nGreeting.\nRespond politely, welcomingly, and offer assistance.\n\n* "Hello! How can I help you today?"\n* "Hi there! What's on your mind?"\n* "Greetings! How can I assist you?"Hello! How can I help you today?","extra_content":{"google":{"thought":true}},"role":"assistant"}}],"created":1776509595,"id":"m2LjaadQ2Z7aug_34uPJDw","model":"gemma-4-31b-it","object":"chat.completion","usage":{"completion_tokens":9,"prompt_tokens":2,"total_tokens":70}} To answer your questions: Model Name: I didn't type the model name manually. The model I used was directly selected from the Hermes installation wizard under the "Google AI Studio" option. Version Before Update: Unfortunately, I don't remember the exact commit hash. It was a version I downloaded about 4-5 days ago. That old version was working perfectly fine this morning. Current Setup: Today, I completely uninstalled that older version and did a fresh install using the installation command, which brought me to the latest version. This is exactly when the API_KEY_INVALID issue started. Other Providers: Other APIs, like Nvidia, are working flawlessly on this newly installed version. Since the curl command works perfectly with the exact same API key and endpoint, and other providers are working fine in Hermes, this heavily points to a regression in how the newest Hermes version handles or passes the API key specifically for the Google AI Studio / Gemini OpenAI-compatible endpoint.

Steps Taken

Here is the curl output (using the Bearer token authorization): code JSON {"choices":[{"finish_reason":"stop","index":0,"message":{"content":"The user said "hi".\nGreeting.\nRespond politely, welcomingly, and offer assistance.\n\n* "Hello! How can I help you today?"\n* "Hi there! What's on your mind?"\n* "Greetings! How can I assist you?"Hello! How can I help you today?","extra_content":{"google":{"thought":true}},"role":"assistant"}}],"created":1776509595,"id":"m2LjaadQ2Z7aug_34uPJDw","model":"gemma-4-31b-it","object":"chat.completion","usage":{"completion_tokens":9,"prompt_tokens":2,"total_tokens":70}} To answer your questions:

Installation Method

Install script (curl | bash)

Operating System

Ubuntu

Python Version

No response

Hermes Version

No response

Debug Report

Full Error Output

Here is the curl output (using the Bearer token authorization):
code
JSON
{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"The user said "hi".\nGreeting.\nRespond politely, welcomingly, and offer assistance.\n\n* "Hello! How can I help you today?"\n* "Hi there! What's on your mind?"\n* "Greetings! How can I assist you?"Hello! How can I help you today?","extra_content":{"google":{"thought":true}},"role":"assistant"}}],"created":1776509595,"id":"m2LjaadQ2Z7aug_34uPJDw","model":"gemma-4-31b-it","object":"chat.completion","usage":{"completion_tokens":9,"prompt_tokens":2,"total_tokens":70}}
To answer your questions:

What I've Already Tried

No response