JFYI - Your code also works with Ultra and Embeddings!

[palladius]

I've tried to substitute 'gemini-pro' with other strings provided by the Gemini Vertex AI playground code, and JFYI they both works:

GeminiModel = "gemini-1.0-pro-001" # This works as 'gemini-pro'
GeminiModel = "gemini-1.0-ultra-001" # this also works! Yuppie!

JFYI - Looking fwd to speaking to you in person (Laurencio is arranging a chat).

Riccardo

[palladius]

See scripts using your gem here: https://github.com/palladius/genai-googlecloud-scripts/tree/main/09-langchainrb-playground

[inspire22]

Nice! I was trying to get this working but had failed. Where are you setting these? Is ultra even a thing still, I thought they renamed it to advanced, but maybe the model name stayed the same. Are you in the beta program?

[palladius]

I work for Google. I found also embeddings work:

# Take @gbaptista code in README and change this: 

 client = Gemini.new(
    credentials: {
      service: 'vertex-ai-api',
      file_path: 'riccardo.json' ,
      region: 'us-central1', 
    },
    options: {
      model: 'text-embedding-preview-0409',
      service_version: 'v1',
    }
  )

  result = client.request(
    'predict',
    {"instances": [
      { "content": 'Ford Prefect vs Zaphod'}
    ]}
  )

I got the model name from these docs: https://cloud.google.com/vertex-ai/generative-ai/docs/embeddings/get-text-embeddings

[inspire22]

Ouch! So changing to vertex-ai-api, though it uses the same models, wasn't free and now I have a $120 bill lol. Google's having two very similar AI APis, with one being free, is super frustrating.

Any idea how to see which embeddings are available in generative-language-api? I can't figure out how to get it to list them, and the old 'embedding-001' is the only one I can find a reference to online. The new ones in vertex-ai-api like text-embedding-preview-0409 aren't there.

Kinda wish I'd found this before, but https://ollama.com/ is pretty amazing for running one of the free models locally.

[gbaptista]

@inspire22 Ouch, sorry to hear that! Yeah, Ollama is great; I support a gem for using it: ollama-ai

About embeddings: I added new methods for embedding in version 4.0.0 of the gem.

I tested, and I can get embedding working with the model text-embedding-004, both in Generative Language API:

result = client.embed_content(
  { content: { parts: [{ text: 'What is life?' }] } }
)

Or Vertex AI API:

result = client.predict(
  { instances: [{ content: 'What is life?' }],
    parameters: { autoTruncate: true } }
)

@palladius It was great to meet you! Thanks for reaching out. I generated a script to test which model I have access to. Here's my result:

Model	Vertex AI	Generative Language
gemini-pro-vision	✅	🔒
gemini-pro	✅	✅
gemini-1.5-pro-preview-0514	✅	🔒
gemini-1.5-pro-preview-0409	✅	🔒
gemini-1.5-pro	✅	✅
gemini-1.5-flash-preview-0514	✅	🔒
gemini-1.5-flash	✅	✅
gemini-1.0-pro-vision-latest	🔒	🔒
gemini-1.0-pro-vision-001	✅	🔒
gemini-1.0-pro-vision	✅	🔒
gemini-1.0-pro-latest	🔒	✅
gemini-1.0-pro-002	✅	🔒
gemini-1.0-pro-001	✅	✅
gemini-1.0-pro	✅	✅
gemini-ultra	🔒	🔒
gemini-1.0-ultra	🔒	🔒
gemini-1.0-ultra-001	🔒	🔒
text-embedding-preview-0514	🔒	🔒
text-embedding-preview-0409	🔒	🔒
text-embedding-004	✅	✅
embedding-001	🔒	✅
text-multilingual-embedding-002	✅	🔒
textembedding-gecko-multilingual@001	✅	🔒
textembedding-gecko-multilingual@latest	✅	🔒
textembedding-gecko@001	✅	🔒
textembedding-gecko@002	✅	🔒
textembedding-gecko@003	✅	🔒
textembedding-gecko@latest	✅	🔒

@inspire22 This may help you with "what model works with what API" ☝️

I will do some benchmarks with gemini-1.5-pro-preview-0514 and gemini-1.5-flash-preview-0514. 😄

[inspire22]

Brilliant list, thanks! How'd you get the list of models/embeddings to test?

I've been using text-embedding-preview-0409 since (was, still quite high) at the top of the MTEB https://huggingface.co/spaces/mteb/leaderboard

Ollama - oh nice, I hadn't noticed you were behind both, thanks for the great projects!

[gbaptista]

Unfortunately, there's no API to list models in Vertex AI API to my knowledge.

With Generative Language API, this endpoint works for me:

https://generativelanguage.googleapis.com/v1/models?key=YOUR_KEY

In the gem, you can use it like this:

client = Gemini.new(
  credentials: {
    service: 'generative-language-api',
    api_key: ENV.fetch('GOOGLE_API_KEY', nil)
  },
  options: { model: 'text-embedding-004', server_sent_events: true }
)

models = client.models

{ 'models' =>
  [{ 'name' => 'models/gemini-1.0-pro',
     'version' => '001',
     'displayName' => 'Gemini 1.0 Pro',
     'description' => 'The best model for scaling across a wide range of tasks',
     'inputTokenLimit' => 30_720,
     'outputTokenLimit' => 2048,
     'supportedGenerationMethods' => %w[generateContent countTokens],
     'temperature' => 0.9,
     'topP' => 1 },
   { 'name' => 'models/gemini-1.0-pro-001',
     'version' => '001',
     'displayName' => 'Gemini 1.0 Pro 001 (Tuning)',
     'description' => 'The best model for scaling across a wide range of tasks. This is a stable model that supports tuning.',
     'inputTokenLimit' => 30_720,
     'outputTokenLimit' => 2048,
     'supportedGenerationMethods' => %w[generateContent countTokens createTunedModel],
     'temperature' => 0.9,
     'topP' => 1 }] }

The others I manually extracted from reading these pages:

• Google models
  • Model versions and lifecycle