Skip to main content
This page answers the most common questions about using Portrix. If you don’t find what you’re looking for here, reach out to our support team at [email protected] or use the in-dashboard chat.
Yes. Portrix implements the OpenAI API specification, so any code that works with the OpenAI SDK or a direct HTTP client works with Portrix after two changes:
  1. Set your base URL to https://api.portrix.ai/v1 instead of https://api.openai.com/v1.
  2. Replace your OpenAI API key with your Portrix API key.
That’s it. Your existing prompt logic, parameter names, response parsing, and tool/function calling code all continue to work without modification. To use a non-OpenAI model, simply change the model field to a Portrix model ID like anthropic/claude-3-5-sonnet.
Portrix provides access to 400+ models from dozens of providers including OpenAI, Anthropic, Google, Mistral, Meta, Cohere, Groq, and more. The catalog is continuously updated as providers release new models.You can retrieve the full, up-to-date list programmatically at any time:
See the Supported Models page for a curated overview of popular models.
No. When you sign up for Portrix, you immediately get access to all supported models through Portrix’s shared provider keys — no separate accounts with OpenAI, Anthropic, or Google are required.Optionally, you can add your own provider API keys in Settings > Providers. When you do, Portrix routes your requests through your accounts instead of the shared pool. This is useful if you have existing provider credits, prefer direct billing relationships, or need dedicated provider-level quota. See Providers for setup instructions.
Portrix charges you based on the tokens you consume — input tokens plus output tokens — with pricing that varies by model and provider. Generally, smaller, faster models cost less per token than larger, more capable ones.You can view current per-model pricing on the Billing page in the dashboard at app.portrix.ai. The Usage page shows your running token consumption and estimated cost for the current billing period.If you add your own provider API keys, those requests are billed directly to you by the provider; Portrix charges only a platform fee for routing and features.
Yes. Portrix is designed for production workloads. Paid plans (Pro and Enterprise) include:
  • Uptime SLAs
  • Priority support
  • Higher rate limits and daily quotas
  • Fallback routing to minimize downtime
  • Audit logs and usage analytics
The Free plan is suitable for prototyping and low-volume testing. For production use, we recommend upgrading to Pro or Enterprise.
Portrix’s fallback routing lets you define a list of models to try in sequence. If the primary model returns a 5xx error or times out, Portrix automatically retries the request with the next model in your fallback chain — transparently, with no changes required in your application code.You configure fallback chains per request (using the fallbacks parameter) or as account-level defaults in the dashboard. You can also monitor provider status on the Portrix status page.
Portrix collects request logs for platform operations, debugging, and abuse prevention. By default, logs include metadata (model used, token counts, timestamps, error codes) but not the content of your messages.Full content logging (storing your actual prompts and completions) is available as an opt-in feature for teams that want it for debugging or compliance purposes. It can also be explicitly disabled at the API key level. Portrix does not use your data to train or fine-tune any models. See the Privacy Policy for complete details.
Yes. Any framework or library that supports OpenAI-compatible APIs works with Portrix out of the box. This includes:
  • LangChain — use ChatOpenAI with openai_api_base="https://api.portrix.ai/v1"
  • LlamaIndex — configure OpenAI with a custom api_base
  • Instructor — pass a patched OpenAI client pointing to Portrix
  • Vercel AI SDK — use the openai provider with a custom base URL
  • AutoGen, CrewAI, and other agent frameworks — configure the OpenAI client as usual
Because Portrix is API-compatible, you typically only need to set base_url and api_key on the underlying OpenAI client your framework uses.
The Free plan includes 60 requests per minute (RPM) and 100,000 tokens per minute (TPM), with a daily token quota of 500,000 tokens.For full details on limits across all plans and guidance on handling 429 errors, see the Rate Limits page.
You can reach the Portrix support team through any of these channels:When reporting an API issue, include the x-portrix-request-id value from the response headers — this lets the support team locate your specific request in the logs and diagnose the problem much faster.
Self-hosted deployment is available on the Enterprise plan. With a self-hosted deployment, the Portrix gateway runs in your own infrastructure (on-premises or your cloud account), giving you full data residency control and the ability to connect directly to provider APIs from your own network.Contact [email protected] to discuss self-hosted options and Enterprise pricing.
For most use cases, openai/text-embedding-3-small offers an excellent balance of performance and cost. If you need maximum embedding quality for complex semantic search or retrieval tasks, use openai/text-embedding-3-large.Both models are available through Portrix and are called via the /v1/embeddings endpoint using the same OpenAI-compatible request format: