--- title: "Using supermemory local with different providers" sidebarTitle: "Providers" description: "Copy-paste .env setup for Ollama, OpenAI, Anthropic, Gemini, and OpenRouter" icon: "route" --- Supermemory local needs one model provider to power summaries, contextual chunking, and memory extraction. Pick the tab for whichever one you already have a key for. For the full variable reference (fast/text model overrides, offline setup, tuning), see [Configuration](/self-hosting/configuration). Fully offline — no API key leaves your machine. Any OpenAI-compatible local runner works the same way (LM Studio, vLLM, llama.cpp server); this is the Ollama version. ```bash ollama pull gpt-oss:20b ``` ```bash .env OPENAI_BASE_URL=http://localhost:11434/v1 OPENAI_API_KEY=ollama # any non-empty string — Ollama doesn't check it OPENAI_MODEL=gpt-oss:20b ``` `gpt-oss:20b` is a good default for a laptop-class GPU. Bigger models work if you have the VRAM — set `OPENAI_MODEL` to whatever you've pulled. ```bash .env OPENAI_API_KEY=sk-... ``` That's it — defaults to `gpt-5.1`. Override with `OPENAI_MODEL` if you want a different one. ```bash .env ANTHROPIC_API_KEY=sk-ant-... ``` Runs on `claude-haiku-4-5` — this one isn't currently configurable via env var. ```bash .env GEMINI_API_KEY=... ``` Runs on `gemini-3.1-flash-lite-preview` — also not currently overridable. This is the only key that also unlocks image, video, and high-fidelity PDF understanding (see the [full provider table](/self-hosting/configuration#llm-providers)). OpenRouter isn't a native provider — it's OpenAI-compatible, so it slots into the same `OPENAI_BASE_URL` path as Ollama: ```bash .env OPENAI_BASE_URL=https://openrouter.ai/api/v1 OPENAI_API_KEY=sk-or-... OPENAI_MODEL=openai/gpt-4o-mini ``` Set `OPENAI_MODEL` to any model slug from [OpenRouter's model list](https://openrouter.ai/models) — routing, fallback, and pricing all follow OpenRouter's own rules from there.