Configure LLM Providers in TEDI
TEDI uses a chat model for Assistant replies, Document Intelligence, playbook automation that needs an LLM, and Microsoft 365 auto-replies. Configure providers under Models → Providers .
When you need this
- You want Chat / Assistant to answer with a cloud model.
- Document Intelligence or playbooks need an LLM.
- Teams or Outlook inbound is set up and you want auto-reply (inbound capture alone does not require a model).
You can install and run many Vista automations without an LLM. Add a provider when you are ready for chat or auto-reply.
Before you begin
- TEDI is installed and you can open the desktop app.
- You have an API key from one of the supported cloud providers (or you will create one next):
Open Models → Providers
- Launch TEDI .
- In the left sidebar, open Models .
- Select the Providers tab.
Expected outcome: You see Active selection and provider cards (or an empty state ready to add a provider).
Add a cloud provider
- Under Add provider , choose OpenAI , Grok , or Gemini .
- Select Add .
- Expand the new provider card.
TEDI uses a fixed default endpoint for these three kinds. You normally do not need to edit the endpoint.
- OpenAI:
https://api.openai.com/v1 - Grok:
https://api.x.ai/v1 - Gemini:
https://generativelanguage.googleapis.com/v1beta/openai/
Paste the API key and choose models
- In the provider card, paste your key into API key .
- Under Models , enable one or more suggested models (or add a Custom model id if your account has a specific model name).
- Set Default model for this provider .
- Select Save on the card.
Expected outcome: The card summary shows API key set and a default model. Leave the API key field blank on later edits to keep the saved key.
Set the active provider and model
- In Active selection , choose Active provider .
- Choose Active model .
Switching provider or model saves immediately and applies to chat, playbooks, and Document Intelligence.
Expected outcome: The status banner shows the active provider as configured (for example, OpenAI is active · <model> ).
How keys are stored
- TEDI stores the API key on this Windows machine with the provider config. It is not meant to be shared or pasted into tickets.
- Copy the key when the vendor shows it (often only once), paste it into TEDI, then store a backup in your password manager if needed.
- If a key is leaked, revoke it in the vendor console and create a new one.
Validation checklist
- [ ] Provider card shows API key set
- [ ] At least one model is selected and a default is set
- [ ] Active provider / active model match what you intend to use
- [ ] Status shows the provider as active (not “API key missing”)
- [ ] Optional: send a short Chat message and confirm a reply
Common mistakes
- Looking under Settings for LLM — Cloud providers are under Models → Providers .
- Adding the provider but not Saving the card — Expand the card, paste the key, then Save .
- No models selected — Enable at least one model (or add a custom model id) before relying on Chat.
- Expecting Teams/Outlook auto-reply without a model — Inbound can capture messages; auto-reply needs an active chat model. See Configure Microsoft 365 in TEDI .
- Using a ChatGPT / consumer subscription as an API key — Consumer chat apps are separate from developer API keys. Use the provider’s developer console.