Configuration

oxo-call uses a layered configuration system with sensible defaults, file-based overrides, and environment variable support.

Configuration File

Settings are stored in a TOML file at the platform-specific configuration directory:

Platform	Path
Linux	~/.config/oxo-call/config.toml
macOS	~/Library/Application Support/io.traitome.oxo-call/config.toml
Windows	%APPDATA%\traitome\oxo-call\config.toml

Find your config path:

oxo-call config path

Configuration Keys

Key	Default	Environment Variable	Description
llm.provider	github-copilot	OXO_CALL_LLM_PROVIDER	LLM provider
llm.api_token	(unset)	OXO_CALL_LLM_API_TOKEN	API token
llm.api_base	(auto)	OXO_CALL_LLM_API_BASE	Override API base URL
llm.model	(auto)	OXO_CALL_LLM_MODEL	Model name
llm.max_tokens	2048	OXO_CALL_LLM_MAX_TOKENS	Maximum tokens
llm.temperature	0.0	OXO_CALL_LLM_TEMPERATURE	Temperature (0.0 = deterministic)
llm.context_window	0 (auto-detect)	—	Model context window size in tokens (0 = auto-detect)
llm.prompt_tier	auto	OXO_CALL_LLM_PROMPT_TIER	Prompt compression tier: auto, full, medium, compact
llm.cache_enabled	false	—	Enable LLM response caching (reduces API calls for repeated tasks)
llm.stream	true	—	Enable SSE streaming for LLM responses (reduces perceived latency)
docs.auto_update	true	OXO_CALL_DOCS_AUTO_UPDATE	Auto-refresh docs on first use

Setting Values

# Set a value
oxo-call config set llm.provider openai
oxo-call config set llm.api_token sk-...

# Set prompt compression tier (useful for small models)
oxo-call config set llm.prompt_tier compact    # force compact for ≤3B models
oxo-call config set llm.prompt_tier auto       # auto-detect (default)

# Set context window size
oxo-call config set llm.context_window 4096    # force Medium tier

# Enable LLM response caching (useful for repeated tasks)
oxo-call config set llm.cache_enabled true     # cache LLM responses
oxo-call config set llm.cache_enabled false    # disable cache (default)

# Disable streaming output (useful for CI/batch scripts and benchmarks)
oxo-call config set llm.stream false           # disable streaming
oxo-call config set llm.stream true            # enable streaming (default)

# Get the effective value (includes env overrides)
oxo-call config get llm.provider
oxo-call config get llm.prompt_tier

# Show all configuration
oxo-call config show

# Verify LLM connectivity
oxo-call config verify

Environment Variables

Environment variables override config.toml values. Provider-specific token variables are also supported as fallbacks:

• GitHub: GITHUB_TOKEN, GH_TOKEN
• OpenAI: OPENAI_API_KEY
• Anthropic: ANTHROPIC_API_KEY

LLM Provider Details

GitHub Copilot (Default)

• Default model: gpt-5-mini (lightweight, free tier ⭐)
• API base: https://api.individual.githubcopilot.com
• Authentication: Use oxo-call config login for interactive OAuth login
• Important: Requires GitHub App token (ghu_), not Personal Access Token (ghp_)

# Recommended: Interactive login
oxo-call config login

# Manual setup (if you have a ghu_ token)
oxo-call config set llm.api_token ghu_xxxxxxxxxxxxxxxxxxxx

Note: GitHub Copilot ignores GITHUB_TOKEN and OXO_CALL_LLM_API_TOKEN environment variables because they often contain Personal Access Tokens that don't work with Copilot's token exchange endpoint. Always use oxo-call config login or manually set a ghu_ token.

OpenAI

• Default model: gpt-4.1 (1M context, April 2025)
• API base: https://api.openai.com/v1
• Compatible with Azure OpenAI via llm.api_base override

Anthropic

• Default model: claude-sonnet-4-6-20250514 (1M context)
• API base: https://api.anthropic.com/v1

DeepSeek

• Default model: deepseek-chat (128K context)
• API base: https://api.deepseek.com/v1
• OpenAI-compatible API

MiniMax

• Default model: minimax-chat (1M context)
• API base: https://api.minimax.chat/v1
• OpenAI-compatible API

Ollama

• Default model: llama3.2
• API base: http://localhost:11434/v1
• No API token required (local inference)

Custom

export OXO_CALL_LLM_PROVIDER=cherryin
export OXO_CALL_LLM_API_BASE=https://open.cherryin.cc/v1
export OXO_CALL_LLM_API_TOKEN=yxxxx-xxxxxx
export OXO_CALL_LLM_MODEL=google/gemini-2.5-flash-lite

---

Troubleshooting

Wrong or Missing License

If your license file is missing, expired, or invalid, you will see an error like:

Error: License verification failed — no valid license found.
Checked locations (in order):
  1. --license CLI flag
  2. OXO_CALL_LICENSE environment variable
  3. ~/.config/oxo-call/license.oxo.json

Run `oxo-call license verify` for details.

Fix: Ensure your license.oxo.json is at one of the checked paths. See License Setup.

Failed LLM Connection

If oxo-call config verify fails with a connection error:

✗ LLM provider: openai
✗ Connection: Failed — could not reach https://api.openai.com/v1

Common causes:

• No API token: Run oxo-call config get llm.api_token to check
• Wrong provider: Verify with oxo-call config get llm.provider
• Network issue: Check internet connectivity or proxy settings
• Ollama not running: Start with ollama serve

Config File Not Found

Config file not found at ~/.config/oxo-call/config.toml
Using default values.

This is normal on first use. Set your first value to create the file:

oxo-call config set llm.provider openai

CI / HPC Cluster Considerations

When running oxo-call in non-interactive environments (CI pipelines, SLURM job scripts, HPC clusters):

1. License: Set OXO_CALL_LICENSE to the path of your license file in your job script or CI environment
2. API tokens: Use environment variables instead of config files:

   export OXO_CALL_LLM_PROVIDER=openai
   export OXO_CALL_LLM_API_TOKEN=$OPENAI_API_KEY
   

3. No GITHUB_TOKEN: If your CI environment does not set GITHUB_TOKEN, switch to OpenAI, Anthropic, or Ollama
4. Ollama on clusters: Run Ollama as a service on a shared node, then set llm.api_base to point to it:

   export OXO_CALL_LLM_PROVIDER=ollama
   export OXO_CALL_LLM_API_BASE=http://ollama-node:11434/v1
   

5. SLURM example:

   #!/bin/bash
   #SBATCH --job-name=oxo-call-pipeline
   #SBATCH --cpus-per-task=8
   
   export OXO_CALL_LICENSE=/shared/licenses/license.oxo.json
   export OXO_CALL_LLM_PROVIDER=ollama
   export OXO_CALL_LLM_API_BASE=http://ollama-node:11434/v1
   
   oxo-call run samtools "sort input.bam by coordinate using 8 threads"