Tools and frameworks

Radium inside the frameworks your application already runs on

Tools and frameworks

Radium exposes an OpenAI-compatible Chat Completions API. Any framework, SDK, or library that can point at a custom base URL and model name works with Radium. This section covers the most common integrations.

Frameworks

LangChain

Python chains and agents via ChatOpenAI

LangGraph

State-machine agents with graph nodes

LlamaIndex

RAG and query engines via OpenAILike

Vercel AI SDK

Next.js and Node.js streaming

OpenAI Python SDK

Direct SDK with custom base URL

PydanticAI

Type-safe agents withOpenAIChatModel

CrewAI

Multi-agent crews with LiteLLM routing

DSPy

Program optimization with signatures

AG2

AutoGen successor with OpenAIConfig

Agno

Lightweight agents with OpenAILike

What changes in every integration

Three values change when you point any framework at Radium:

1. Base URL — https://api.radium.cloud/v1 (OpenAI format)
2. API key — your Radium API key from deploy.radium.cloud
3. Model string — hal-1.0, clarke-1.0, or tycho-1.0

Everything else — prompts, tools, chains, agents, streaming logic — stays the same.

Compatibility matrix

Framework	Transport	Streaming	Tool calling	Verified
LangChain	HTTP	Yes	Yes	Aug 2026
LangGraph	HTTP	Yes	Yes	Aug 2026
LlamaIndex	HTTP	Yes	Yes	Aug 2026
Vercel AI SDK	HTTP	Yes	Yes	Aug 2026
OpenAI Python SDK	HTTP	Yes	Yes	Aug 2026
PydanticAI	HTTP	Yes	Yes	Aug 2026
CrewAI	HTTP	Yes	Yes	Aug 2026
DSPy	HTTP	Yes	Yes	Aug 2026
AG2	HTTP	Yes	Yes	Aug 2026
Agno	HTTP	Yes	Yes	Aug 2026

All integrations were verified with Python 3.12.7 or Node.js 22+ against hal-1.0, clarke-1.0, and tycho-1.0.

Common patterns

Environment variables

Every guide uses the same two environment variables:

export RADIUM_API_KEY="your-radium-api-key"
export RADIUM_MODEL="hal-1.0"

Virtual environment

Python guides create a virtual environment before installing packages:

mkdir radium-project && cd radium-project
python3 -m venv .venv
source .venv/bin/activate  # macOS/Linux
# .venv\Scripts\Activate.ps1  # Windows

Model selection

Switch models by changing the environment variable:

export RADIUM_MODEL="clarke-1.0"  # or tycho-1.0

Troubleshooting all integrations

Symptom	Cause	Fix
401 or 403	Missing or wrong API key	Export RADIUM_API_KEY in the active terminal
404	Wrong base URL or model ID	Use https://api.radium.cloud/v1 and exact model names
Empty output	max_tokens too low	Use 512 or higher
Import errors	Wrong virtual environment	Activate .venv and reinstall packages
Calls going to OpenAI	Missing custom provider	Pass the Radium base_url explicitly