Key takeaways
- Dave by voolama (hellodave.ai) supports two model catalog modes for every provider record: Fetch from API (retrieves the model list directly from the AI service, supported by six of the seven provider types) and JSON (a manually entered array of model objects, required for Custom providers and available as a fallback for all types). Sourced from the admin-providers help documentation, DEV v0.11.0, 2026-08-30.
- The Fetch Models button in Dave by voolama (hellodave.ai) is independently verifiable: the model list it returns matches the models available in your account on the AI service's own platform. OpenAI publishes its model list at platform.openai.com/docs/models. Anthropic publishes its model list at docs.anthropic.com/en/docs/about-claude/models/overview. A successful fetch also confirms the API key and Base URI are both correct.
- Azure OpenAI providers in Dave by voolama (hellodave.ai) return zero models after a fetch if no model deployment exists in the Azure resource. The fix is to deploy at least one model in the Azure portal before fetching. The admin-providers help documentation states: 'For Azure OpenAI, ensure at least one model is deployed in your Azure resource.'
- The JSON model catalog format in Dave by voolama (hellodave.ai) is an array of model objects: [{"id": "gpt-4o", "name": "GPT-4o"}]. The id field is the model identifier sent to the AI service API. A typo in the id field causes agents to fail at runtime with a model-not-found error. Verify id values against the AI service's own model reference before saving.
- Only the Admin role can view, create, edit, or fetch models for providers in Dave by voolama (hellodave.ai). Dave by voolama has six default roles: Admin, Create, Curate, Use, Reporting, and Financial. Roles are editable. NIST SP 800-53 Rev 5 AC-6 section 6a (least privilege) requires that access to credential-bearing configuration be restricted to users who need it: the Admin-only boundary on the Providers page satisfies AC-6 section 6a.
Summary
Fetch from API versus JSON: when to use each model catalog mode in Dave by voolama
Every provider record in Dave by voolama (hellodave.ai) has a model catalog configured in one of two modes. API keys are stored encrypted with AES-256-GCM authenticated encryption, specified in NIST SP 800-38D: the authentication tag detects and rejects any tampering with the ciphertext. Keys are never returned through the API once stored.
| Mode | How it works | Supported provider types | When to use it | Independent verification |
|---|---|---|---|---|
| Fetch from API | Click Fetch Models. Dave by voolama calls the AI service's model list endpoint using the provider's API key and Base URI. The returned models appear in a searchable list with checkboxes. Check the models you want to include. Use All or None for bulk selection. | OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, Local (Ollama). All six support fetch. | Always use this mode when it is available. It avoids typos in model IDs, shows the latest models in your account, and simultaneously confirms the API key and Base URI are correct. | The model IDs returned by Fetch Models match the models in your account on the provider's platform. OpenAI: platform.openai.com/docs/models. Anthropic: docs.anthropic.com/en/docs/about-claude/models/overview. No vendor trust required. |
| JSON | Paste or type a JSON array of model objects: [{"id": "gpt-4o", "name": "GPT-4o"}]. The id field is sent to the AI service at runtime. The name field is the display label in the agent dropdown. | All seven provider types. Required for Custom; available as a fallback for the other six. | Use for Custom providers (fetch not available). Use as a fallback when fetch fails and you need to unblock agent configuration quickly. | Verify the id values against the AI service's own model reference before saving. A typo in the id field causes agents to fail at runtime with a model-not-found error. No vendor trust required. |
The toggle between modes is the Fetch from API / JSON button pair in the Model Catalog section of the provider create or edit form. Switching modes does not clear the existing catalog.
Model catalog notes for all seven provider types in Dave by voolama
Each of the seven provider types in Dave by voolama (hellodave.ai) has a specific model catalog consideration. Default Base URIs are the AI services' own published endpoints, independently verifiable against each service's documentation.
| Provider type | Default Base URI | Model catalog note |
|---|---|---|
| OpenAI | https://api.openai.com/v1 | Fetch returns all models your API key has access to. Curate to the models your team uses. Model IDs at platform.openai.com/docs/models. Example: gpt-4o, gpt-4o-mini. |
| Anthropic | https://api.anthropic.com/v1 | Fetch returns the Claude model list. Current model IDs at docs.anthropic.com/en/docs/about-claude/models/overview. Include the full model ID including the date suffix (for example, claude-sonnet-4-20250514). |
| Azure OpenAI | (you must supply your Azure endpoint) | Fetch returns zero models if no deployment exists in your Azure resource. Deploy at least one model in the Azure portal first. Include the deployment name in the provider's Name field for clarity (e.g., Azure OpenAI, eastus2 gpt-4o). |
| Google AI | https://generativelanguage.googleapis.com/v1beta | Fetch returns the Gemini model list. Curate to the models your team uses. |
| OpenAI-Compatible | (you must supply the endpoint) | Fetch is supported if the endpoint implements the OpenAI /models endpoint. If not, switch to JSON mode and enter model IDs manually. |
| Local (Ollama) | http://localhost:11434/v1 | Fetch returns the models currently pulled to the Ollama instance. The Ollama default endpoint is the Ollama server's own published default. Pointing an agent at a local Ollama model is model hosting, not product hosting: Dave by voolama remains fully managed SaaS operated by voolama LLC. |
| Custom | (you must supply the endpoint) | Fetch is not available. You must supply the model catalog using the JSON editor. |
How to curate the model catalog in Dave by voolama: NIST CM-7 section 7a and three practices
The model catalog configured on a provider record in Dave by voolama (hellodave.ai) is what Curate-role users see in the agent configuration dropdown.
NIST SP 800-53 Rev 5 CM-7 section 7a requires: "Configure the system to provide only essential capabilities." Curating the model catalog to only the models your team actually uses satisfies CM-7 section 7a at the AI model layer: agents cannot call a model that is not in the catalog. This is verifiable at csrc.nist.gov/pubs/sp/800/53/r5/upd1/final.
Three catalog curation practices:
- Use Fetch from API, then deselect unused models. After clicking Fetch Models, use the search box to filter by model family. Deselect models your team does not use. Click Update Provider. The agent configuration dropdown now shows only the curated list.
- Keep at least one active provider with at least one model. If all providers are deleted or misconfigured, no agent can execute AI calls and all workflows that rely on agents will fail.
- Test connectivity after any catalog change. After editing a provider, click Fetch Models to verify the API key and Base URI are still correct. A successful fetch confirms the provider can reach the AI service.
Only the Admin role can edit the model catalog. Dave by voolama has six default roles: Admin, Create, Curate, Use, Reporting, and Financial. Roles are editable. The catalog change takes effect immediately. Verify: edit a provider catalog to remove a model, open the agent configuration dropdown in a Curate-role session. The removed model no longer appears. No vendor trust required.
What this article does not cover
This article covers model catalog modes, provider-type-specific catalog notes, and catalog curation practices in Dave by voolama (hellodave.ai) at DEV v0.11.0. Three things it does not cover:
- Adding a provider for the first time. The step-by-step provider creation path, including entering the API key and Base URI, is covered at Add an AI Provider.
- Provider error troubleshooting. The five documented error patterns (invalid API key, missing provider, failed model fetch, rate limits, zero models) are covered in the existing provider troubleshooting guide on hellodave.ai.
- Agent configuration. Selecting a provider and model when building or editing an agent is covered at Create an Agent. This article covers the provider catalog itself, not how agents reference it.
Frequently asked questions
What is the difference between Fetch from API and JSON mode for the model catalog in Dave by voolama?
Fetch from API mode in Dave by voolama (hellodave.ai) retrieves the model list directly from the AI service using the provider's API key and Base URI. It is supported by six of the seven provider types (all except Custom). A successful fetch also confirms the API key and Base URI are correct. JSON mode lets you paste or type a model array manually: [{"id": "gpt-4o", "name": "GPT-4o"}]. JSON mode is required for Custom providers and is available as a fallback for all types. Use Fetch from API whenever it is available: it avoids typos in model IDs and shows the latest models in your account.
Why does my Azure OpenAI provider return zero models after fetching in Dave by voolama?
An Azure OpenAI provider in Dave by voolama (hellodave.ai) returns zero models after a fetch if no model deployment exists in the Azure resource. Fix: navigate to your Azure resource in the Azure portal, deploy at least one model (for example, gpt-4o in the eastus2 region), return to Dave by voolama, and click Fetch Models again. Alternatively, switch to JSON mode and enter the model ID manually. The Azure portal is the independently verifiable source: navigate to your Azure resource, confirm at least one model deployment exists, then retry the fetch.
What JSON format does Dave by voolama use for the manual model catalog?
The JSON model catalog format in Dave by voolama (hellodave.ai) is an array of model objects: [{"id": "gpt-4o", "name": "GPT-4o"}]. The id field is the model identifier sent to the AI service API at runtime. The name field is the display label shown in the agent configuration dropdown. A typo in the id field causes agents to fail at runtime with a model-not-found error. Verify the id values against the AI service's own model reference: OpenAI at platform.openai.com/docs/models, Anthropic at docs.anthropic.com/en/docs/about-claude/models/overview.
How do I add or remove models from an existing provider catalog in Dave by voolama?
Click the pencil icon on the provider card at Admin, then API Management, then Providers in Dave by voolama (hellodave.ai). The edit form opens pre-populated with the current catalog. In Fetch from API mode: click Fetch Models, then check or uncheck models using the search box and checkboxes, or use All or None for bulk selection. In JSON mode: edit the array directly. Click Update Provider to save. The updated catalog is immediately available in the agent configuration dropdown for Curate-role users.
Which provider types require a manually supplied model catalog in Dave by voolama?
Only the Custom provider type in Dave by voolama (hellodave.ai) requires a manually supplied model catalog. The admin-providers help documentation states: 'For Custom providers, model fetch is not available: you must supply the model catalog manually using the JSON editor.' All six other provider types (OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, and Local (Ollama)) support Fetch from API mode. JSON mode is also available as a fallback for those six types.
Sources
- OpenAI Models: available model IDs and capabilities reference (platform.openai.com)
- Anthropic Claude Models: available model IDs and capabilities reference (docs.anthropic.com)
- Add an AI Provider: Learn Dave Quick-Reference (hellodave.ai)
- NIST SP 800-53 Rev 5: AC-6 Least Privilege, CM-7 Least Functionality, IA-5 Authenticator Management (NIST CSRC)
Last reviewed