mirror of
https://github.com/usestrix/strix.git
synced 2026-08-28 05:25:00 +00:00
44 lines
1.9 KiB
Text
44 lines
1.9 KiB
Text
---
|
|
title: "OrcaRouter"
|
|
description: "Configure Strix with models via OrcaRouter"
|
|
---
|
|
|
|
[OrcaRouter](https://www.orcarouter.ai) routes each request to the best model across 150+ models from OpenAI, Anthropic, Google, DeepSeek, Qwen, and more, through a single OpenAI-compatible API.
|
|
|
|
## Setup
|
|
|
|
```bash
|
|
export STRIX_LLM="openai/openai/gpt-5.5"
|
|
export LLM_API_BASE="https://api.orcarouter.ai/v1"
|
|
export LLM_API_KEY="sk-orca-..."
|
|
```
|
|
|
|
## Available Models
|
|
|
|
Access any model on OrcaRouter using the format `openai/<provider>/<model>`:
|
|
|
|
| Model | Configuration |
|
|
|-------|---------------|
|
|
| OrcaRouter Auto | `openai/orcarouter/auto` |
|
|
| GPT-5.5 | `openai/openai/gpt-5.5` |
|
|
| Claude Sonnet 4.6 | `openai/anthropic/claude-sonnet-4.6` |
|
|
| Gemini 3 Pro | `openai/google/gemini-3.1-pro-preview` |
|
|
| DeepSeek V4 Flash | `openai/deepseek/deepseek-v4-flash` |
|
|
|
|
## Get API Key
|
|
|
|
1. Go to [orcarouter.ai](https://www.orcarouter.ai)
|
|
2. Sign in and navigate to **API Keys**
|
|
3. Create a new API key (starts with `sk-orca-`)
|
|
|
|
## Notes
|
|
|
|
- Model IDs are namespaced by provider (`openai/`, `anthropic/`, `google/`, ...). Keep the full `<provider>/<model>` id after the `openai/` prefix — OrcaRouter rejects bare model names.
|
|
- `orcarouter/auto` is an adaptive router, not a model: it picks an upstream per request. Its candidate pool can include models without tool-calling support, so for agentic scans pin a specific model such as `openai/gpt-5.5`.
|
|
|
|
## Benefits
|
|
|
|
- **Adaptive routing** — `orcarouter/auto` selects the best upstream per request with no client-side logic
|
|
- **Single API** — Access OpenAI, Anthropic, Google, DeepSeek, and more
|
|
- **OpenAI-compatible** — Drop-in replacement using `LLM_API_BASE`
|
|
- **Gateway-level security** — It also runs gateway-level, zero-trust security for AI agents on the same endpoint — screening every prompt/response and governing every tool call on a default-deny basis, with no application code changes.
|