zopyx.llm-moonshot

LLM plugin for Moonshot AI's models

LLM plugin for models hosted by Moonshot AI.

Project Status

This package is maintained as the zopyx.llm-moonshot fork of the original llm-moonshot project.

The fork keeps the plugin usable in current environments and aligns packaging, CI/CD, and release workflow with the zopyx distribution. The post-fork changes and the reasons for them are documented in CHANGELOG.md.

Installation

First, install the LLM command-line utility.

Now install this plugin in the same environment as LLM:

llm install zopyx.llm-moonshot

Configuration

You'll need an API key from Moonshot. Grab one at platform.moonshot.cn.

Set secret key:

llm keys set moonshot

Enter key: <paste key here>

Usage

List what's on the menu:

llm models list

You'll see something like:

[[[cog
import cog
from llm_moonshot._models import DEFAULT_MOONSHOT_MODEL_IDS
for model_id in DEFAULT_MOONSHOT_MODEL_IDS:
    cog.outl(f"Moonshot: moonshot/{model_id}")
]]]
Moonshot: moonshot/kimi-latest
Moonshot: moonshot/moonshot-v1-auto
Moonshot: moonshot/moonshot-v1-128k-vision-preview
Moonshot: moonshot/kimi-k2-0711-preview
Moonshot: moonshot/moonshot-v1-128k
Moonshot: moonshot/moonshot-v1-32k-vision-preview
Moonshot: moonshot/moonshot-v1-8k-vision-preview
Moonshot: moonshot/moonshot-v1-8k
Moonshot: moonshot/kimi-thinking-preview
Moonshot: moonshot/moonshot-v1-32k
Moonshot: moonshot/kimi-k2-thinking
[[[end]]]

Fire up a chat:

llm chat -m moonshot/kimi-k2-0711-preview

Chatting with  moonshot/kimi-k2-0711-preview
Type 'exit' or 'quit' to exit
Type '!multi' to enter multiple lines, then '!end' to finish
> yo moonie
yo! what's up, moonie?
>

Need raw completion?

llm -m moonshot/moonshot-v1-8k "Finish this haiku: Neon city rain"

Neon city rain,
Glistening streets, a symphony,
Echoes of the night.

Reasoning Content Support

This plugin now supports reasoning content for Moonshot's thinking models (models with "thinking" in the name). When using thinking models, you'll see the model's reasoning process displayed in real-time before the final response:

llm chat -m moonshot/kimi-k2-thinking

[Reasoning] (shown in cyan dim)

The user is asking me to solve a complex problem. Let me think through this step by step...
First, I need to understand the core requirements...
Then I'll analyze the available options...

[Response] (shown in bold green)

Here's my well-reasoned answer to your question...

Available Thinking Models

• moonshot/kimi-k2-thinking - Latest reasoning model
• moonshot/kimi-thinking-preview - Preview reasoning model

The reasoning content helps you understand:

• Decision-making process - See how the model analyzes problems
• Multi-step reasoning - Follow complex thought chains
• Error detection - Catch logical gaps or misunderstandings early

Aliases

Save your wrists:

llm aliases set kimi moonshot/kimi-latest

Now:

llm -m kimi "write a haiku about the AI chatbot Sidney is misbehaving"

Troubleshooting

Models don't appear in llm models list

• Make sure you have set a Moonshot API key with llm keys set moonshot.
• The plugin caches the model catalog for one hour. If the API was unreachable, it falls back to a built-in catalog.

Streaming connection dropped

• The plugin automatically retries without streaming when Moonshot closes the connection mid-stream. This is a known upstream behavior and is handled transparently.

401 Unauthorized

• Double-check your API key at https://platform.moonshot.cn.

Development

Clone, sync, build:

git clone https://github.com/zopyx/llm-moonshot.git
cd llm-moonshot
uv sync --extra dev
make check
make dist

To publish using .pypirc with twine:

make upload

See CONTRIBUTING.md for more details.