Backends

Idep supports four AI backends. Switch by changing one line in your config. No restart required (coming in v0.2.0-beta — restart required for now).

Ollama (local, no API key)

Best for: privacy, offline use, no ongoing cost.

[ai]
backend  = "ollama"
model    = "codellama:13b"
endpoint = "http://localhost:11434"

Install Ollama from ollama.com, then pull a model:

ollama pull codellama:13b

Recommended models: codellama:13b, deepseek-coder:6.7b, starcoder2:15b.

endpoint defaults to http://localhost:11434 if omitted.

Best for: highest quality completions, Claude models.

[ai]
backend = "anthropic"
model   = "claude-haiku-4-5-20251001"

[ai.auth]
api_key = "sk-ant-..."

Or set the environment variable:

export IDEP_API_KEY="sk-ant-..."

Best for: open models, flexible hosting.

[ai]
backend = "huggingface"
model   = "bigcode/starcoder2-15b"

[ai.auth]
api_key = "hf_..."

Best for: Groq, Together AI, LM Studio, or any OpenAI-compatible endpoint.

[ai]
backend  = "openai"
model    = "gpt-4o-mini"
endpoint = "https://api.groq.com/openai/v1"

[ai.auth]
api_key = "gsk_..."

The endpoint field accepts any OpenAI-compatible base URL. Omit endpoint to use OpenAI directly (https://api.openai.com/v1).

Backend	API key required	Cloud dependency	Best for
Ollama	No	None	Local, private, offline
Anthropic	Yes	Moderate	Quality completions
HuggingFace	Yes	Moderate	Open models
OpenAI-compat	Yes	Moderate–High	Flexibility, Groq speed