eidechse
Streaming chat with LLMs — straight from your terminal.
curl -fsSL https://eidechse.nnsn.pro/install.sh | bash
Live demo
streaming · syntax highlighting · /copy, /save
eidechse new -m glm -s "You are a Go expert."
Here is a minimal Go program:
package main
import "fmt"
func main() {
fmt.Println("hello, world")
}
What it does
-
Streaming + highlighted code
Responses stream live; code blocks are syntax-highlighted with monokai by default.
-
Many models, one config
Each model is an alias; switch with /model — base URL and key follow automatically.
-
Local-first history
Every session is its own file in ~/.ai-eidechse-cli/. Nothing leaves your machine but the request to your server.
-
REPL commands
/copy, /save (Markdown), /title, /switch, /new — chat without leaving the prompt.
-
OpenAI · Anthropic · llama.cpp
One chat over different API styles; configurable temperature, max-tokens, timeout and code theme.
-
One wheel for Mac/Linux
Ships as a single py3-none-any wheel. No repo on your machine, no source tree.
Install
-
1. Install
Run the command. Inspect the script first if you prefer.
curl -fsSL https://eidechse.nnsn.pro/install.sh | bashscript: https://eidechse.nnsn.pro/install.sh
-
2. Create a config
Add providers (base URL + key) and models (aliases) to ~/.ai-eidechse-cli/config.json.
full config guide → SETUP.md{ "providers": { "zai": { "base_url": "https://api.z.ai/api/paas/v4", "api_key": "..." }, "anthropic": { "base_url": "https://api.anthropic.com/v1", "api_key": "sk-ant-...", "api": "anthropic" } }, "models": { "glm": { "provider": "zai", "model": "glm-4.6" }, "claude": { "provider": "anthropic", "model": "claude-3-5-sonnet-20240620" } }, "default_model": "glm" } -
3. Run
Launch eidechse, or start a model directly with eidechse new -m <alias>.
eidechse
Updating = the same install command (latest is always served).