Nataris — use P2P Android inference as a backend for your Langflow flows

[Sharrmavishal]

Built an inference network that runs open-weight models on Android phones and exposes them as an OpenAI-compatible API. Langflow supports custom OpenAI-compatible endpoints, so it drops in with a one-line config change.

What it is

Nataris is a P2P inference marketplace. Android phones run Qwen 2.5 0.5B and Llama 3.2 1B locally via llama.cpp, and serve developer requests through a standard OpenAI-compatible API. Phone owners earn per token. Developers get inference without spinning up any servers.

No prompt logging. No content filtering. No model training on your queries.

Adding Nataris to Langflow

In any Langflow component that uses an OpenAI-compatible model, set:

• API Base URL: https://api.nataris.ai/v1
• API Key: your Nataris API key (get one at nataris.ai)
• Model name: nataris-fast (Qwen 2.5 0.5B) or nataris-balanced (Llama 3.2 1B)

For the ChatOpenAI or OpenAI component, update the base URL in the model settings. Works with any Langflow flow that uses a chat or completion node.

One thing to keep in mind: latency is 5–20s depending on model and device warmth, so Nataris works best in flows where you’re not expecting sub-second responses. Keep streaming enabled where the component supports it.

Where we are

21 provider devices on the network, 2,775 inference jobs completed, 350K+ tokens processed. Just came out of closed beta. Built by a two-person team.

$5 free credits on signup, no card needed.

API: https://api.nataris.ai/v1
Docs: https://api.nataris.ai/docs
Provider app (earn by running models on your Android): https://play.google.com/store/apps/details?id=ai.nataris.app