Failure with gemini embedding

[systemz64]

Hello

i have been successfully using embedding-001 for weeks now and had been away for a week. Today I started adding more docs but now get this error when parsing:

Fail to bind embedding model: 404 models/embedding-001 is not found for API version v1beta

embedding-001 is a valid model and shows up in the models list in the config. I am using 0.23.1

Any help would be appreciated

Z

[systemz64]

BTW I tried to switch to Qwen embedding using the OpenAI compat mode (which works with rerank) but it says model not found (whether 8B or 4B). The models exist in the llm factory config.

[Magicbook1108]

#13073

[systemz64]

great thank you!

[xXMrNidaXx]

Gemini embedding failures — common issue with a few causes:

1. API key setup

export GOOGLE_API_KEY="AIza..."
# or
export GEMINI_API_KEY="AIza..."

2. Model name format

# Correct format
embedding_model: "models/embedding-001"
# or
embedding_model: "text-embedding-004"

3. Rate limits
Gemini embedding has strict limits:

• Free tier: 60 requests/minute
• Add retry logic:

from tenacity import retry, wait_exponential

@retry(wait=wait_exponential(min=1, max=60))
def embed_with_gemini(text):
    return genai.embed_content(...)

4. Region restrictions

# Some regions blocked
# Try VPN or different region config

5. Input size

# Gemini has token limits per request
max_tokens = 2048  # Check current limit
chunks = split_text(text, max_tokens)

Debug:

import google.generativeai as genai
genai.configure(api_key="...")

# Test directly
result = genai.embed_content(
    model="models/embedding-001",
    content="test"
)
print(result)

We use Gemini embeddings at RevolutionAI. What error message are you seeing?