Documentation
Configuring AI providers
The Magic Pointer AI in Pointus is supported using your own API credentials or local models. Below are example configuration instructions for each of the supported providers.
Please note!
AI providers release and retire models frequently. The models suggested below represent dependable, low-latency choices at the time of writing, but newly published models or alias tags may offer faster response times, higher accuracy and/or better pricing. Always check the provider's official model catalog linked in each section.
Google Gemini
- Visit the Google AI Studio API Keys page.
- Sign in with your Google Account and click Create API Key.
- Select an existing Google Cloud project or generate a new one, then copy the key.
- Check the Gemini Model Documentation to verify the latest model IDs.
| Setting field | Recommended value/example |
|---|---|
| API Key | AIzaSy... |
| Preferred model(s) * | gemini-3.5-flash-lite, gemini-3.8-flash |
*
Google regularly updates Gemini aliases and preview tags. Faster, lower-latency multimodal releases may become available—verify the latest IDs on Google AI Studio.
OpenAI
- Navigate to the OpenAI API Keys Dashboard.
- Sign in, verify your phone number if prompted, and ensure you have an active prepaid balance in Billing.
- Click Create new secret key, set your permissions, and store the token safely.
- See the OpenAI Models Overview for specific snapshot and alias names.
| Setting field | Recommended value/example |
|---|---|
| API Key | sk-proj-... |
| Preferred model(s)* | gpt-4o-mini, gpt-4o |
*
OpenAI releases frequent point updates and snapshot snapshots. Review the OpenAI model catalog to substitute lighter, newer endpoints as they become available.
Anthropic Claude
- Go to the Anthropic Console API Keys page.
- Sign in or create a developer account, and make sure credits are added under Billing.
- Click Create Key, name it (e.g., Pointus), and copy the resulting string.
- Review the Anthropic Models Guide for current generation identifiers.
| Setting field | Recommended value/example |
|---|---|
| API Key | sk-ant-api03-... |
| Preferred model(s)* | claude-3-7-sonnet-latest,claude-3-5-haiku-latest |
*
Using the -latest suffix automatically points to Anthropic's current stable build, but newer model families may supersede these. Check the Anthropic documentation for newly launched models.
xAI (Grok)
- Visit the xAI Cloud Console.
- Log in with your X/xAI account and configure a payment method.
- Head to the API Keys tab and click Create API Key.
- Consult the xAI Model Documentation for available model endpoints.
| Setting field | Recommended value/example |
|---|---|
| API Key | xai-... |
| Preferred model(s)* | grok-2-vision-1212 |
*
xAI endpoints and beta aliases change rapidly as newer Grok generations exit training. Inspect the xAI docs for active production identifiers.
Ollama
- Download and install Ollama from ollama.com.
- Start Ollama on your machine or private server. By default, it listens at http://localhost:11434.
- Pull any model you want to run locally via your terminal (e.g.,
ollama run llama3.2orollama pull mistral). - Check the Ollama Model Library for all supported local model tags.
| Setting field | Recommended value/example |
|---|---|
| Host Endpoint | http://localhost:11434 |
| Preferred model(s)* | minicpm-v, llama3.2:latest, mistral:latest |
*
Local model availability depends entirely on the models you have pulled locally. Check the Ollama library for lightweight quantized weights optimized for low-latency desktop execution.
Apple Vision
Basic on-device visual intelligence: OCR, QR decoding & scene classification.
Ready to go - zero setup required.*
Ready to go - zero setup required.*
| Setting field | Recommended value/example |
|---|---|
| n/a | n/a |
*
Only available on macOS