AI Providers¶

PokéDex Scanner supports multiple AI providers for card image recognition. All providers use a vision-capable language model — the server handles any request/response translation transparently.

Supported providers¶

Provider	`AI_PROVIDER` value	Required env var	Notes
GitHub Models (default)	`github`	`GITHUB_MODELS_TOKEN`	Free GitHub PAT, no extra scopes
OpenAI	`openai`	`OPENAI_API_KEY`
Groq	`groq`	`GROQ_API_KEY`	Free tier available
Ollama	`ollama`	(none)	Runs locally, no API key needed
Azure OpenAI	`azure`	`AZURE_OPENAI_URL`, `AZURE_OPENAI_API_KEY`
Anthropic Claude	`anthropic`	`ANTHROPIC_API_KEY`	Request/response format translated server-side

Default models per provider¶

Provider	Default model
GitHub Models	`meta/llama-4-maverick-17b-128e-instruct-fp8`
OpenAI	`gpt-4o`
Groq	`meta-llama/llama-4-scout-17b-16e-instruct`
Ollama	`llava`
Anthropic	`claude-opus-4-5`

Override with VITE_CARD_ANALYSIS_MODEL=<model-name> when running the frontend in local dev or when rebuilding it.

VITE_CARD_ANALYSIS_MODEL is a Vite frontend environment variable, so it is read at build/dev-server startup time. For Docker/production deployments with a prebuilt frontend, changing this value requires rebuilding the frontend. To change the model at runtime instead, use the Settings UI or POST /api/settings/ai.

Configuration examples¶

GitHub Models (default)OpenAIGroqOllama (local)Azure OpenAIAnthropic Claude

export AI_PROVIDER=github
export GITHUB_MODELS_TOKEN="ghp_..."
npm run dev:server

Get a free token at https://github.com/settings/tokens/new?description=Pokemon+Card+Scanner&scopes=.

export AI_PROVIDER=openai
export OPENAI_API_KEY="sk-..."
export VITE_CARD_ANALYSIS_MODEL=gpt-4o
npm run dev:server

export AI_PROVIDER=groq
export GROQ_API_KEY="gsk_..."
export VITE_CARD_ANALYSIS_MODEL=meta-llama/llama-4-scout-17b-16e-instruct
npm run dev:server

# Start Ollama with a vision model first
ollama pull llava

export AI_PROVIDER=ollama
export OLLAMA_BASE_URL=http://localhost:11434
export VITE_CARD_ANALYSIS_MODEL=llava
npm run dev:server

export AI_PROVIDER=azure
export AZURE_OPENAI_URL="https://<resource>.openai.azure.com/openai/deployments/<deployment>"
export AZURE_OPENAI_API_KEY="..."
npm run dev:server

export AI_PROVIDER=anthropic
export ANTHROPIC_API_KEY="sk-ant-..."
export VITE_CARD_ANALYSIS_MODEL=claude-opus-4-5
npm run dev:server

Choose a vision-capable model

Not all models support image input. Make sure the model you select is listed as vision-capable by the provider. The defaults above are all vision-capable.