Configure your first AI model

Add an OpenAI-compatible provider, test a model, and select defaults for individual features.

NoteGen can use included models or connect to your own OpenAI-compatible service, Ollama, or LM Studio. Provider setup and feature-specific model selection are separate steps.

Model Platforms

Manage AI platforms, credentials, and models for chat, organization, knowledge, and audio.

Model platforms

ChatGPTCurrent

Gemini

Ollama

LM Studio

Custom platform

ChatGPT

Manage AI platforms, credentials, and models for chat, organization, knowledge, and audio.

Configured · 4 models

Connection

ConnectionOptions

API Key

sk-••••••••••••

Base URL

https://api.openai.com/v1

Proxy mode

Models

GPT-5

Chat · Streaming

text-embedding-3-small

Embedding

Add a provider

Open Settings → Model Services.
Choose a built-in template or create a custom provider.
Enter a name, Base URL, and API key.
Add the model name and choose the correct type, such as chat, embedding, rerank, stt, or tts.
Run the connection test and save.

A Base URL normally ends at the version path, such as https://api.openai.com/v1, without /chat/completions. Ollama commonly uses http://localhost:11434/v1 and LM Studio http://localhost:1234/v1, but the local service must be running.

Select models for features

After adding models, select them in the relevant settings:

Use	Setting	Typical type
Chat, Agent, organization	model selector or defaults	`chat`
Image understanding	Image Recognition	vision-capable `chat`
Knowledge-base indexing	Knowledge Base	`embedding`
Retrieval reranking	Knowledge Base	`rerank`
Transcription	Audio Settings	`stt`
Read aloud	Read Aloud	`tts`

Proxy and compatibility

A provider can inherit the global proxy, force a direct connection, or use a custom proxy. If an older compatible endpoint rejects max_completion_tokens, switch to max_tokens. Leave context window, temperature, and output limits at their defaults unless the provider documents different values.

Security and troubleshooting

API keys are stored in local configuration and are only uploaded if settings sync is enabled and permits them.
Never publish keys in issues, screenshots, diagnostics, or shared configuration.
401/403 usually indicates a key, permission, or account problem. 404 often means an incorrect Base URL or model name. For timeouts, check the network and proxy.
A phone cannot reach a model running on a computer through localhost; use the computer's LAN address and check firewall rules.

See Model configuration and Model selection and fallback for every field.