Alma Client Configuration Guide

Introduction

Alma is a powerful and beautifully designed AI client tool that integrates code writing, terminal operations, Git management, browser, and many other features, providing developers with a complete AI-assisted programming experience.

Official Website: https://alma.now/

Features

✅ Code Writing: Intelligent code generation and completion
✅ Terminal Operations: Integrated terminal for direct command execution
✅ Git Management: Visual Git operations
✅ Browser Integration: Built-in browser support
✅ Multi-Model Support: Supports Claude, Gemini, and other models
✅ Custom Providers: Connect to custom API providers

Installation & Configuration

Install Alma

Visit the Alma official website to download and install the version for your operating system.

Set Up Chinese Interface

For a better experience, we recommend setting the interface to Chinese:

Open the Alma application
Go to Settings → General → Language
Select "Chinese"

Add 元渊智能 Custom Provider

Click Select Provider → Add Custom Provider
Fill in the following information:
- Provider Name: 元渊智能 (or custom name)
- Base URL: https://api.ai.soraliststudio.com/v1
- API Key: Paste your API Key from the 元渊智能 Console

Claude models: Use API Key from the C-prefixed group
Gemini models: Use API Key from the gemini group
元渊智能's Codex does not currently support third-party client calls

After adding, click the Fetch button on the 元渊智能 provider page to automatically retrieve all available models.
Select the model you need to start using.

Configure Multiple Groups (Optional)

元渊智能 supports accessing different model collections through different API Keys (different groups).

In the provider interface, you can:

Enter API Keys for different groups
Retrieve model lists for each group
Switch between them as needed

To use Gemini 3's image generation feature, you need to select a default channel API Key.

Create Chat & Usage

Important Settings Reminder

Strongly Recommended

When creating a new chat, select "Select All" for the context scope instead of "Auto"!

Testing shows that selecting "Auto" causes severe hallucination issues, potentially leading to:

Referencing non-existent files
Generating incorrect code logic
Misunderstanding project structure

Selecting "Select All" ensures the model gets complete context information for more accurate responses.

Usage Tips

Code Writing: Describe your requirements and Alma will generate the corresponding code. Supports multi-file operations and project-level modifications.
Terminal Operations: Integrated terminal for direct command execution with result viewing.
Git Management: Visual Git status viewing with AI-assisted commit message writing.
File Browsing: Browse project file structure, quickly locate and edit files.

Best Practices

1. Choose the Right Model

Claude Sonnet: Suitable for daily coding tasks, fast
Claude Opus: Suitable for complex architecture design and difficult problems
Gemini: Suitable for tasks requiring multimodal capabilities (e.g., image understanding)

2. Use Context Wisely

For small projects, "Select All" files
For large projects, select relevant modules and files
Regularly clean irrelevant context to avoid token waste

3. Leverage Alma's Integrated Features

Use terminal features to verify AI-generated code
Use Git integration to manage code versions
Use browser features to view documentation and references

FAQ

How to switch between different API providers?

In the model selector at the top of the chat interface, you can switch between different providers and models at any time.

Why are some models unavailable?

Check if the corresponding API Key has permission to access that model
Confirm account balance is sufficient
Check the 元渊智能 Console token group settings (Claude uses C-prefixed group, Gemini uses gemini group)

What's the difference between Alma and Claude Code?

Feature	Alma	Claude Code
Interface	Graphical UI	Command Line
Use Case	Full-featured dev environment	Terminal quick operations
Integrations	Terminal, Git, Browser, etc.	Pure CLI tool
Learning Curve	Low	High

Alma Client Configuration Guide

On this page