:AILogin lists detected local backends first; Enter uses the running local node
with NO login (cloud login optional, fails with a clean one-line message). Adds
Ollama (:11434) + LM Studio (:1234) probes alongside the Hanzo engine (:36900);
:AIStatus lists all detected backends + active route. 59 python tests pass.
g:hanzo_route=auto probes the local Hanzo engine /health and routes :AI to its
OpenAI-compatible API directly (coherent, no auth) when running, else falls back
to the cloud gateway/account creds. Adds g:hanzo_local_url/g:hanzo_local_model;
:AIStatus shows the active route. Does not route through `dev` (agent harness
garbles small local models).
:AILogin/:AILogout/:AIStatus delegate OAuth to the installed `dev` CLI
(dev login [--chatgpt] [--device-code] [--with-api-key]) and the provider
resolves creds in dev's order (env -> ~/.codex/auth.json -> ~/.hanzo/auth.json),
sending x-api-key for Anthropic and Bearer for OpenAI/Hanzo. Not vendor-locked;
:HanzoLogin kept as alias.
Add a buffer that invokes a neural completion for the entire buffer
contents.
* Tech: Split neural prompt and run and add line options
* Config: Add initialisation settings for neural buffer
* Config: Add completion trigger plug mapping for neural buffer
* Fix: prompt ui escape keys
* Test: Add filetype in test vimrc for neural buffer tests
Add a NeuralStop command and associated plug mapping to stop Neural.
Unless configured otherwise, Neural will now set <C-c> to stopping
Neural by default in normal mode, unless another mapping is already set
for that key combination.
hanzo-dev