Integrating Google models with Lakehousecat enables Chat and Speech-to-Text (STT) capabilities. This page describes how to configure Google as a Provider Model.
Prerequisites
- Access to Lakehousecat as an Administrator
- A valid Google API key for Chat (Gemini API)
- For Speech-to-Text, a separate Google Cloud API key — see Speech-to-Text below. The Gemini chat key does not work for STT.
Configuration
Navigate to Workspace → Models → Provider Models and click the + icon.
Step 1: Select Provider Type
From the Provider Type dropdown, select Google.
Step 2: Select Model Type
Choose the appropriate model type:
| Model Type | Description |
|---|---|
| Chat | Conversational AI (e.g. gemini-2.0-flash-exp) |
| STT | Speech-to-Text (e.g. chirp) |
Step 3: Enter Configuration Details
All Google configurations require:
| Field | Description |
|---|---|
| Configuration Name | A unique name for this configuration (e.g. google-gemini-chat). |
| Google API Key | Your Google API key. |
Chat
| Field | Description |
|---|---|
| Model ID | The Google model identifier (e.g. gemini-2.0-flash-exp). Select from the dropdown or enter manually. |
Model ID accepts any Google model identifier as free text — see Tested Configurations below for models explicitly verified by the Lakehousecat team.
STT (Google Cloud Speech-to-Text)
Unlike other providers, Google Speech-to-Text runs on a different Google API family — Cloud Speech-to-Text (speech.googleapis.com), not the Gemini API (generativelanguage.googleapis.com) used for Chat. As a result, your Gemini chat key will not work for STT. Before configuring an STT model, complete the setup below.
One-time setup in the Google Cloud Console (not Google AI Studio):
- Use (or create) a project with billing enabled — Cloud Speech-to-Text is not part of the free Gemini API tier.
- Enable the Cloud Speech-to-Text API for that project.
- Create a separate API key dedicated to Speech-to-Text. When you set the key's API restriction, the console lets you allow either the Gemini API or Speech-to-Text — not both on the same key. A key created for Gemini/Chat therefore cannot be reused for STT, regardless of how long you wait.
Once the dedicated STT key is in place, transcription works immediately. Without it, requests fail with 403 PERMISSION_DENIED / API_KEY_SERVICE_BLOCKED, even when everything else in Lakehousecat is configured correctly.
Configuration fields:
| Field | Description |
|---|---|
| Google API Key | The dedicated Speech-to-Text key from the setup above (not the Gemini chat key). |
| STT Model | Select a recognizer model from the curated list below. |
| STT Language Codes (optional) | Comma-separated language codes (e.g. en-US, de-DE). The first is the primary language. |
Available recognizer models:
| Model | Typical use |
|---|---|
latest_long | General long-form audio (default choice) |
latest_short | Short utterances, commands |
chirp_3 / chirp_2 | Google's universal speech models |
telephony | Phone-call audio |
medical_conversation / medical_dictation | Medical speech |
Step 4: Save
Click Save. After saving, you can optionally set this configuration as the default UI model or default Backend model.
Tested Configurations
Stand: 2026-07-20. Google Chat was validated through Lakehousecat's chart generation workflow directly (no separate plain-chat test round was run — chart generation is the more demanding case and covers chat behavior along the way). This reflects what Lakehousecat has actually run and observed, not a benchmark or performance ranking — see There is no single "right" provider for why we don't publish comparative scores.
| Model ID | Chart Generation | Notes |
|---|---|---|
gemini-flash-lite-latest | ✅ Recommended (50/50 charts, 0 failures, ~13.5s median) | Best result observed across all providers tested so far in this round |
gemini-flash-latest | ⚠️ Mixed results (3/4 runs hit context issues) | Sample size still small; not yet conclusive |
gemini-2.5-pro | ❌ Not available at test time | Returned "model not found" via the API during testing |
Best Practices
- Use a descriptive configuration name that identifies the Google service and model type.
- Consult the Google Cloud documentation for valid model IDs.
- For chart generation, prefer Gemini Flash / Flash-Lite — the fastest, most reliable tier for this structured task. See Choosing a Model for Chart Generation.
- Cost visibility is delayed (~1 day): today's usage may not appear in the Google Cloud billing dashboard until the following day. Check again the next day rather than assuming usage was not recorded.
Troubleshooting
- Validation Errors: Verify the Google API Key and Model ID.
- Access Issues: Ensure the relevant Google Cloud APIs are enabled in your project.