Supported Providers
orxhestra supports all major LLM providers via LangChain integrations.| Provider | Name | Package | Install |
|---|---|---|---|
| OpenAI | openai | langchain-openai | pip install orxhestra[openai] |
| Azure OpenAI | azure_openai | langchain-openai | pip install orxhestra[openai] |
| Azure AI | azure_ai | langchain-azure-ai | pip install langchain-azure-ai |
| Anthropic | anthropic | langchain-anthropic | pip install orxhestra[anthropic] |
| Anthropic Bedrock | anthropic_bedrock | langchain-aws | pip install langchain-aws |
| Google GenAI | google | langchain-google-genai | pip install orxhestra[google] |
| Google Vertex AI | google_vertexai | langchain-google-vertexai | pip install langchain-google-vertexai |
| AWS Bedrock | bedrock_converse | langchain-aws | pip install langchain-aws |
| Mistral | mistralai | langchain-mistralai | pip install langchain-mistralai |
| Cohere | cohere | langchain-cohere | pip install langchain-cohere |
| Fireworks | fireworks | langchain-fireworks | pip install langchain-fireworks |
| Together | together | langchain-together | pip install langchain-together |
| Groq | groq | langchain-groq | pip install langchain-groq |
| NVIDIA NIM | nvidia | langchain-nvidia-ai-endpoints | pip install langchain-nvidia-ai-endpoints |
| HuggingFace | huggingface | langchain-huggingface | pip install langchain-huggingface |
| DeepSeek | deepseek | langchain-deepseek | pip install langchain-deepseek |
| Ollama | ollama | langchain-ollama | pip install langchain-ollama |
| xAI (Grok) | xai | langchain-xai | pip install langchain-xai |
| IBM watsonx | ibm | langchain-ibm | pip install langchain-ibm |
| Perplexity | perplexity | langchain-perplexity | pip install langchain-perplexity |
| OpenRouter | openrouter | langchain-openrouter | pip install langchain-openrouter |
| Upstage | upstage | langchain-upstage | pip install langchain-upstage |
CLI Auto-Detection
The CLI automatically detects the provider from the model name:orx --model gpt-5.4 # → openai
orx --model claude-sonnet-4-6 # → anthropic
orx --model gemini-2.0-flash # → google
orx --model mistral-large # → mistralai
orx --model deepseek-chat # → deepseek
orx --model grok-2 # → xai
orx --model command-r-plus # → cohere
YAML Configuration
defaults:
model:
provider: anthropic
name: claude-sonnet-4-6
temperature: 0.7
# Or use named models
models:
fast:
provider: groq
name: llama-3.3-70b-versatile
smart:
provider: anthropic
name: claude-opus-4-7
agents:
researcher:
model: fast
writer:
model: smart
Azure OpenAI
Azure OpenAI requires explicit provider since model names overlap with OpenAI:defaults:
model:
provider: azure_openai
name: gpt-5.4
azure_endpoint: https://my-resource.openai.azure.com/
api_version: "2024-12-01-preview"
azure_deployment: gpt-5.4
Custom Providers
Register a custom provider or use a dotted import path:from orxhestra.composer.builders.models import register
register("my_provider", MyCustomChatModel)
defaults:
model:
provider: my_module.MyCustomChatModel
name: my-model
Environment Variables
| Provider | Env Var |
|---|---|
| OpenAI | OPENAI_API_KEY |
| Azure OpenAI | AZURE_OPENAI_API_KEY |
| Anthropic | ANTHROPIC_API_KEY |
GOOGLE_API_KEY | |
| Mistral | MISTRAL_API_KEY |
| Groq | GROQ_API_KEY |
| DeepSeek | DEEPSEEK_API_KEY |
| xAI | XAI_API_KEY |
| Cohere | COHERE_API_KEY |
| Together | TOGETHER_API_KEY |
| Fireworks | FIREWORKS_API_KEY |
| NVIDIA | NVIDIA_API_KEY |
| Perplexity | PPLX_API_KEY |
| OpenRouter | OPENROUTER_API_KEY |
| Upstage | UPSTAGE_API_KEY |