Module local 

Local OpenAI-compatible provider.

Targets an on-device inference server (Ollama / LM Studio / llama.cpp) that exposes the OpenAI /v1/chat/completions API. Unlike the cloud openai provider, the endpoint is per-request (req.base_url, sourced from the user’s local_ai_base_url setting) and authentication is optional - local servers typically ignore the bearer token.

Gated by allow_local_ai at the executor layer (not allow_cloud_ai): a localhost call is not network egress, so it stays usable even when cloud AI is disabled.

Structs

LocalOpenAiProvider

Functions

local_chat_url

Resolve the chat-completions endpoint from any accepted base form.

local_embeddings_url

Resolve the embeddings endpoint from any accepted base form.

local_models_url

Resolve the model-listing endpoint from any accepted base form.

normalize_local_base

Reduce a user-entered server address to its origin/base, tolerating both a bare http://127.0.0.1:1234 and a fully-qualified http://127.0.0.1:1234/v1/chat/completions. Trailing slashes and the known OpenAI-compat path suffixes are stripped so the result can have a fresh path appended.

parse_model_list