INSTALL ↓
xllama / wiki / providers

Models & providers

Use local llama.cpp by default, or switch to any supported OpenAI-compatible HTTP endpoint without changing the agent workflow.

Provider catalog

llamacpp, openai, azure, anthropic, gemini, grok, deepseek, glm, zai, cerebras, ollama, vllm, lmstudio and custom.

xllama setup --provider openai --api-key sk-… --model gpt-4o-mini
xllama setup --provider anthropic --api-key sk-ant-… --model claude-sonnet-4-5
xllama setup --provider ollama --model llama3.1
xllama setup --provider custom --api-base https://host/v1 --api-key … --model m

Switch inside the TUI

/provider <id> [api_base] [api_key] [api_version]
/apikey <key>
/apiversion <version>
/model <id>

Switching provider or model keeps the current history. Use /new when you want a fresh conversation.

Remote-only launch

A local model is optional. If llama-server is unavailable, xllama still opens and lets you configure a remote provider from inside the TUI—useful on Android/Termux and lightweight hosts.

Provider compatibility

Anthropic and Gemini work through their OpenAI-compatible endpoints without a separate protocol. If a strict provider rejects a deprecated or unsupported request parameter, xllama removes the offending field and retries automatically.

xllama setup --provider anthropic --api-key sk-ant-… --model claude-sonnet-4-5
xllama setup --provider gemini --api-key AIza… --model gemini-2.5-flash

Local llama.cpp

  • Accepts a local .gguf path or Hugging Face owner/repo[:quant].
  • Reuses a healthy llama-server or starts one automatically.
  • Downloads a Hugging Face model on first run with a live progress heartbeat.
  • /model reloads a managed server; /ctx changes its context size.

Azure OpenAI

export AZURE_OPENAI_RESOURCE_NAME=my-resource
export AZURE_OPENAI_KEY=…
export AZURE_OPENAI_API_VERSION=2024-12-01-preview
xllama setup --provider azure --model o4-mini