Learn
Concepts, comparisons and integration notes for the Ragmux gateway — the reference layer behind the how-to docs.
Concepts
- What is an AI gateway?An AI gateway is a single API in front of every model provider you use, adding retrieval, auth and spend control along the way.
- Self-hosted AI gateway comparison: Ragmux, LiteLLM, Portkey and moreEvery self-hosted AI gateway alternative to Ragmux, compared side by side on license, database requirements and built-in retrieval.
Comparisons
- Ragmux vs LiteLLM: which self-hosted AI gateway fits?Ragmux and LiteLLM are both self-hosted AI gateways, but they solve different problems — a look at where each one is actually the better fit.
- Ragmux vs Portkey: open-source gateway or hosted control plane?Portkey's AI Gateway core is MIT-licensed and self-hostable like Ragmux, but its access control and SSO sit behind a hosted product now rebranded PRISMA AIRS.
- Ragmux vs OpenRouter: self-hosted gateway or hosted router?OpenRouter is a hosted model router you can't self-host; Ragmux is a self-hosted gateway with built-in retrieval. This is a "which job" comparison, not a feature race.
- Ragmux vs Cloudflare AI Gateway: edge control plane or self-hosted retrieval?Cloudflare AI Gateway adds edge caching and analytics in front of providers you already call. Ragmux self-hosts routing, retrieval and budgets in one binary.
- Ragmux vs Helicone: gateway or observability platform?Helicone instruments LLM calls after the fact; Ragmux routes and retrieves before the fact. Helicone also ships under two different licenses across two repositories.
- Ragmux vs Langfuse: gateway or LLM engineering platform?Langfuse instruments and evaluates LLM calls an application already makes; Ragmux routes, retrieves and budgets the call itself. Complementary, not competing.
Use cases
- Track LLM spend per project with RagmuxGive every team, product or customer its own Ragmux project, then use its budgets, usage counters and CSV export to see what it actually spent.
- Build private document Q&A with your own PDFsGround a chatbot in your own PDFs and DOCX files without sending them to a third-party RAG service — parsing, embedding and search all run in your own Postgres.
- Air-gapped LLM deployment with RagmuxRun an OpenAI-compatible gateway with retrieval fully inside a network with no outbound internet access, using local model servers and one self-contained PostgreSQL database.
- Migrating from LiteLLM to RagmuxMove a LiteLLM proxy config to Ragmux's model connections and projects — a field-by-field mapping, what has no equivalent, and what to test before cutting over.
Integrations
- Using Claude through an OpenAI-compatible gatewayPoint an existing OpenAI-SDK app at Claude through Ragmux — no Anthropic SDK, and the message, tool and image translation Ragmux does for you.
- Using Gemini through an OpenAI-compatible gatewayCall Gemini with the OpenAI SDK through Ragmux — how tool schemas get sanitised, how tool results are correlated, and the one token count Google's own docs don't agree on.
- Putting Ollama or vLLM behind a gatewayConnect a self-hosted Ollama or vLLM server to Ragmux, keep it off the public internet, and still get the same OpenAI-shaped API and budgets as any hosted provider.
- Ragmux as an OpenAI-compatible API gatewayPut OpenAI, DeepSeek, vLLM or any Chat Completions server behind one endpoint — how field passthrough, streaming usage and error codes actually behave.
- Connecting Vercel AI SDK, n8n, Open WebUI and LibreChat to RagmuxPoint four popular clients at one OpenAI-compatible gateway — which base URL, which key, and why the model field is a route name once a project has routes.