CerebriumAI · mintlify · Jul 15, 2026
@@ -1,6 +1,6 @@
 ---
 title: "Custom Python Web Servers"
-description: "Run ASGI/WSGI Python apps on Cerebrium"
+description: "Run ASGI and WSGI Python web servers like FastAPI, Flask, and Django on Cerebrium with custom routes, middleware, and streaming responses."
 ---
 
 Cerebrium's default runtime covers most app needs. For more control, use ASGI or WSGI servers through the custom runtime feature - enabling custom authentication, dynamic batching, frontend dashboards, public endpoints, and WebSocket connections.

@@ -1,5 +1,6 @@
 ---
 title: Defining Container Images
+description: Configure custom Docker base images, Python dependencies, system packages, and shell commands to build reproducible container images for Cerebrium apps.
 ---
 
 ## Introduction

@@ -1,6 +1,6 @@
 ---
 title: Multi-Region Deployment
-description: Deploy your apps globally across multiple regions for improved latency and data residency compliance
+description: Deploy Cerebrium apps across multiple AWS regions for lower global inference latency, higher availability, and data residency compliance.
 ---
 
 Deploy apps globally across multiple continents to reduce latency through co-location and meet data residency requirements.

@@ -1,6 +1,6 @@
 ---
 title: "Async requests"
-description: "Execute calls to a Cerebrium app to be run asynchronously"
+description: "Send async requests to Cerebrium apps, poll job status, and retrieve results for long-running inference workloads without blocking your client."
 ---
 
 Some apps require asynchronous "fire-and-forget" execution. In this model, Cerebrium handles running the function, while the developer is responsible for ensuring data leaves the function (e.g. via a webhook).

@@ -1,6 +1,6 @@
 ---
 title: "REST API"
-description: "Make authenticated HTTP requests to your Cerebrium endpoints"
+description: "Call Cerebrium apps over HTTPS using the global REST inference API with JWT authentication, JSON request bodies, and low-latency responses."
 ---
 
 All functions on Cerebrium are accessible via authenticated POST requests by default, unless marked private by prefixing the function name with an underscore (e.g. `_private_function()`). Authenticate using the JWT token from the **API Keys** section of the dashboard.

@@ -1,6 +1,6 @@
 ---
 title: "OpenAI-Compatible Endpoints"
-description: ""
+description: "Serve LLMs on Cerebrium with the OpenAI-compatible chat completions and completions API so existing OpenAI SDK clients work without code changes."
 ---
 
 All Cerebrium endpoints are OpenAI-compatible, supporting both `/chat/completions` and `/embedding`. Below is a basic implementation of a streaming OpenAI-compatible endpoint.

@@ -1,5 +1,6 @@
 ---
 title: "Streaming Endpoints"
+description: "Stream tokens and partial results from Cerebrium apps using server-sent events (SSE) for low-latency LLM output and real-time inference responses."
 ---
 
 Streaming sends live output from a model over a server-sent event (SSE) stream.

@@ -1,6 +1,6 @@
 ---
 title: "Webhook Forwarding"
-description: "Forward responses to a specified webhook"
+description: "Forward async inference results from Cerebrium apps to your own webhook URL so long-running jobs can push results back without polling."
 ---
 
 Forward function response data to an external endpoint via POST by adding the `webhookEndpoint` query parameter to any API call:

@@ -1,5 +1,6 @@
 ---
 title: "WebSocket Endpoints"
+description: "Expose bidirectional WebSocket endpoints on Cerebrium apps for real-time voice agents, chat, and streaming inference over persistent connections."
 ---
 
 WebSocket endpoints stream responses to the client, enabling real-time, bidirectional communication.

@@ -1,6 +1,6 @@
 ---
 title: Vercel Integration
-description: Integrate Cerebrium with Vercel to build AI applications
+description: Connect Cerebrium to Vercel to call serverless GPU inference from Next.js apps, sync environment variables, and ship AI features end to end.
 ---
 
 The Cerebrium + Vercel integration provides access to Cerebrium-deployed apps via REST endpoints from Vercel projects. Install it from the Vercel AI marketplace.

@@ -1,6 +1,6 @@
 ---
 title: "Migrating from Mystic"
-description: "Deploy a Model from Mystic on Cerebrium"
+description: "Migrate models from Mystic AI to Cerebrium with a step-by-step guide covering configuration, dependencies, and serverless GPU deployment."
 ---
 
 ## Introduction

@@ -1,6 +1,6 @@
 ---
 title: Deepgram
-description: Deploy Deepgram speech-to-text services on Cerebrium
+description: Deploy Deepgram speech-to-text and transcription models on Cerebrium serverless GPUs for low-latency, high-accuracy real-time audio inference.
 ---
 
 Cerebrium's partnership with [Deepgram](https://www.deepgram.com/) enables simple deployment of speech-to-text (STT) services with simplified configuration and independent scaling.

@@ -1,6 +1,6 @@
 ---
 title: Rime
-description: Deploy Rime text-to-speech services on Cerebrium
+description: Deploy Rime text-to-speech models on Cerebrium serverless GPUs to generate lifelike voice audio with low latency for real-time voice agents.
 ---
 
 <Note>

@@ -1,6 +1,6 @@
 ---
 title: TOML Reference
-description: Complete reference for all parameters available in Cerebrium's default `cerebrium.toml` configuration file.
+description: Complete reference for every parameter in the `cerebrium.toml` config file, including deployment, hardware, scaling, dependencies, and runtime options.
 ---
 
 The configuration is organized into the following main sections:

@@ -1,6 +1,6 @@
 ---
-title: "Real-time Voice Agent"
-description: "Deploy a real-time AI voice agent"
+title: "Real-time voice agent with PipeCat, Deepgram, and Llama 3"
+description: "Build a low-latency real-time AI voice agent on Cerebrium using Deepgram speech-to-text, a Llama 3 LLM, and text-to-speech over WebSockets."
 ---
 
 This tutorial creates a real-time voice agent that responds to queries via speech in ~500ms. The implementation supports swapping in any Large Language Model (LLM) or Text-to-Speech (TTS) model, making it ideal for voice-based use cases like customer support bots and receptionists.

@@ -1,6 +1,6 @@
 ---
 title: "Gradio Chat Interface"
-description: "Using FastAPI, Gradio and Cerebrium to deploy an LLM chat interface"
+description: "Deploy a Gradio chat interface on Cerebrium using FastAPI as an ASGI server to expose an interactive LLM chat UI backed by serverless GPUs."
 ---
 
 This tutorial covers creating and deploying a Gradio chat interface connected to a Llama 8B language model using Cerebrium's custom ASGI runtime. The architecture runs the frontend on CPU instances while the model runs separately on GPU instances for optimal resource utilization.

@@ -1,6 +1,6 @@
 ---
 title: "ComfyUI application at Scale"
-description: "Deploy a ComfyUI application"
+description: "Deploy a ComfyUI Stable Diffusion image generation workflow at scale on Cerebrium serverless GPUs with autoscaling, custom nodes, and API access."
 ---
 
 ### Introduction

@@ -1,6 +1,6 @@
 ---
 title: "Deploy a Vision Language Model with SGLang"
-description: "Build an intelligent ad analysis system that evaluates advertisements across multiple dimensions"
+description: "Deploy a vision-language model on Cerebrium with SGLang to analyze ad creatives, extract structured insights, and score them across multiple dimensions."
 ---
 
 This tutorial deploys a Vision Language Model (VLM) using SGLang on Cerebrium. A VLM combines a large language model (LLM) with a vision encoder, enabling it to understand and process both images and text.

@@ -1,6 +1,6 @@
 ---
 title: "Deploy Triton Inference server and TensorRT-LLM"
-description: "Achieve high throughput with Triton Inference Server and the TensorRT-LLM framework"
+description: "Deploy an LLM on Cerebrium with NVIDIA Triton Inference Server and TensorRT-LLM to achieve high throughput and low-latency GPU inference at scale."
 ---
 
 This tutorial deploys Llama 3.2 3B using TensorRT-LLM's PyTorch backend served through Nvidia Triton Inference Server.

@@ -1,6 +1,6 @@
 ---
 title: "Serving GPT-OSS with vLLM"
-description: "Deploy OpenAI's Latest Open Source Model"
+description: "Deploy OpenAI's GPT-OSS open-source LLM on Cerebrium with vLLM for a scalable, OpenAI-compatible chat completions endpoint on serverless GPUs."
 ---
 
 GPT recently released GPT-OSS ([gpt-oss-20b](https://huggingface.co/openai/gpt-oss-20b) and [gpt-oss-120b](https://huggingface.co/openai/gpt-oss-120b)) two state-of-the-art open-weight language models that deliver strong real-world performance at low cost. Available under the flexible Apache 2.0 license, these models outperform similarly sized open models on reasoning tasks, demonstrate strong tool use capabilities, and are optimized for efficient deployment on consumer hardware.

@@ -1,7 +1,7 @@
 ---
 title: "Deploy a High Throughput Server for Embeddings and Reranking"
 sidebarTitle: "High-Throughput Embeddings Server"
-description: "Deploy a a high-throughput, low-latency REST API for serving text-embeddings, reranking models, clip, clap and colpali"
+description: "Deploy a high-throughput, low-latency REST API on Cerebrium for text embeddings, reranking, CLIP, CLAP, and ColPali models on serverless GPUs."
 ---
 
 This tutorial covers deploying a high-throughput, low-latency REST API for serving text-embeddings, reranking models, clip, clap, and colpali using the open-source framework

@@ -1,6 +1,6 @@
 ---
 title: "Langchain Q&A on a YouTube Video"
-description: "To deploy a Q&A application around a YouTube video"
+description: "Build a LangChain question-and-answer application over YouTube video transcripts on Cerebrium, using embeddings, a vector store, and an LLM."
 ---
 
 This tutorial recreates a question-answering bot based on YouTube video content, inspired by [this application](https://colab.research.google.com/drive/1sKSTjt9cPstl_WMZ86JsgEqFG-aSAwkn?usp=sharing) by @m_morzywolek.

@@ -1,6 +1,6 @@
 ---
 title: "Mistral 7B with vLLM"
-description: "Deploy Mistral 7B with vLLM"
+description: "Deploy Mistral 7B on Cerebrium using the vLLM inference engine to serve fast, batched LLM completions from a serverless GPU-backed endpoint."
 ---
 
 <Note>

@@ -1,6 +1,6 @@
 ---
 title: "OpenAI compatible vLLM endpoint"
-description: "Create a OpenAI compatible endpoint using the vLLM framework"
+description: "Deploy an OpenAI-compatible chat completions endpoint on Cerebrium using vLLM so existing OpenAI SDK clients can call your self-hosted LLM."
 ---
 
 This tutorial creates an OpenAI-compatible endpoint that works with any open-source model. Use existing OpenAI code with Cerebrium serverless functions by changing just two lines of code.

@@ -1,7 +1,7 @@
 ---
-title: "Real-time Voice Agent"
+title: "500ms low-latency real-time voice agent"
 sidebarTitle: "500ms Low-latency Voice Agent"
-description: "Deploy a real-time AI voice agent"
+description: "Build a 500ms end-to-end latency real-time voice agent on Cerebrium with PipeCat, Daily, and co-located STT, LLM, and TTS services for fast turns."
 ---
 
 This tutorial creates a real-time voice agent that responds to queries via speech in ~500ms. The implementation supports swapping in any Large Language Model (LLM) or Text-to-Speech (TTS) model, making it ideal for voice-based use cases like customer support bots and receptionists.

@@ -1,6 +1,6 @@
 ---
 title: "Generate Images using SDXL"
-description: "Generate high quality images using SDXL with refiner"
+description: "Generate high-quality images with Stable Diffusion XL and the SDXL refiner on Cerebrium serverless GPUs through a low-latency REST inference API."
 ---
 
 <Note>

@@ -1,6 +1,6 @@
 ---
 title: "Streaming LLM Output"
-description: "Stream outputs live from Falcon 7B using SSE"
+description: "Stream LLM tokens live from Falcon 7B on Cerebrium using server-sent events (SSE) for a low-latency, interactive chat-style inference response."
 ---
 
 <Note>

@@ -1,6 +1,6 @@
 ---
-title: "Transcribe 1 hour podcast"
-description: "Using Distill Whisper to transcribe an audio file"
+title: "Transcribe a 1-hour podcast with Distil-Whisper"
+description: "Transcribe a one-hour podcast on Cerebrium using the Distil-Whisper speech-to-text model to convert long-form audio into text at low cost and latency."
 ---
 
 This tutorial transcribes an hour-long audio file using Distill Whisper — an optimized version of Whisper-large-v2 that's 60% faster while maintaining accuracy within 1% of the original. The endpoint accepts either a base64-encoded string of the audio file or a URL to download the audio file.

@@ -1,6 +1,6 @@
 ---
 title: "Twilio Voice Agent with PipeCat"
-description: "Integrate a real-time AI voice agent with Twilio"
+description: "Connect a real-time AI voice agent on Cerebrium to a Twilio phone number using PipeCat so callers can talk to an LLM-powered voice assistant."
 ---
 
 This tutorial creates a real-time voice agent that responds to phone calls via Twilio. The implementation supports any LLM or Text-to-Speech (TTS) model, making it ideal for voice applications like customer support bots and receptionists.

@@ -1,6 +1,6 @@
 ---
 title: "Hyperparameter Sweep training Llama 3.2 with WandB"
-description: "Run a hyperparameter sweep on Llama 3.2 with WandB"
+description: "Run a Weights & Biases hyperparameter sweep to fine-tune Llama 3.2 on Cerebrium serverless GPUs and track experiments, metrics, and best configs."
 ---
 
 Hyperparameter sweeps systematically test parameter combinations to find the best-performing model for the least compute or training time.