Quickstart (Z.ai cloud)¶

The fastest way to a working session: GLM from the cloud, no hardware needed.

1. Set your key¶

export ZAI_API_KEY=your-key      # add to ~/.zshrc
# or: hummin auth login zai

2. Start a session¶

cd your-project
hummin                            # interactive TUI

GLM-5.3 is the default suggestion. Three GLM models ship in the zai provider catalog:

Model	When to use it
GLM-5.3	Flagship quality, 1M-token context
GLM-5.3-Flash	Faster, cheaper daily driver
GLM-5.3-highspeed	Latency-critical back-and-forth

Pick with /model. The picker's "set as default" action persists your choice.

3. Oneshot and scripted use¶

hummin -p "summarize this repo"          # print mode: answer, then exit
hummin --mode json                        # JSON event stream, for scripts

4. Reasoning control¶

GLM reasoning maps to hummin's thinking level:

/thinking high     # deeper reasoning, slower
/thinking off      # direct answers only

5. Watch what it is doing¶

/context - token breakdown with cache-hit ratio
/cost - spend per model
/status - model, memory, todos, TUI state at a glance

When the cloud is not the answer¶

If you want models on your own hardware (privacy, zero cost per token, or just because it is fun), the Local Models section walks the whole path: download, serve, connect, use.