FAQ

Updated at:

What is ANOLISA?

Alibaba Cloud's operating system is specially built for AI Agents. It is built on Alinux4 and is fully compatible with all capabilities of Alinux4. It can be understood as an enhanced version of Alinux4's Agent.

Official website product introduction document:Product overview

What enhancements does ANOLISA have for Agent?

ANOLISA's current enhancements to Agent mainly come from 5 parts:

  1. Built-in Skills improve Agent work efficiency, cover system-related scenarios, and will later spread to common scenarios;

  2. Copilot Shell (cosh) replaces the traditional command line and reduces the difficulty of interacting with Agent through natural language interaction;

  3. Agent security enhancement at the system layer: including Skills signature anti-poisoning, sandbox isolation and system call control, security baseline reinforcement, etc.

  4. AgentSight is observable: it realizes fine-grained data collection and correlation analysis of the entire Agent operation link without intruding business logic.

  5. Token-Less, a Token optimization toolkit for LLM: significantly reduces the Token consumption of the context window and improves operating efficiency through three core strategies: pattern compression, response compression, and command rewriting.

What scenarios are suitable for ANOLISA? Are there any best practices?

All businesses used to run Agent are suitable for Agentic OS. For other businesses, please choose Alibaba Cloud Linux 3/4 basic version.

Deploy OpenClaw/Claude Code in one sentence

Are there any fees for ANOLISA?

The mirror itself is free. However, resource consumption during use is billed normally, including ECS ​​instances (vCPU, memory, storage, public network bandwidth, snapshots) and large model calling fees.

How to create an ANOLISA instance?

On the ECS creation page, select the system image as Alibaba Cloud Linux 4 LTS 64-bit Agentic EditionNeed to bind public network IP, recommended memory ≥ 2GB, only supports x86 architecture.

Quick Start:Quick start

User Manual:How to use Alibaba Cloud Linux 4 Agentic Edition (ANOLISA)

How are large model calls billed?

Depends on the authentication method you choose. There is a certain free quota for using Alibaba Cloud authentication; if you use your own API Key, you will be billed according to the pricing of the corresponding service provider.

  • API Key: Supports Bailian/OpenAI compatible endpoints and depends on the pricing of each model provider.

  • Alibaba Cloud Certification: Use Alibaba Cloud authentication (ECS role or AK/SK), free (only supports Qianwen series text models). Speed ​​and success rate are not guaranteed, only trial use is supported.

Does it only support configuring large models of Alibaba Cloud? If all three modes are configured, which one will take effect?

No, all OpenAI compatible model manufacturers support it. The three modes are subject to the last configured mode. There will be instructions for using the model when starting and ending the conversation.

How to troubleshoot authentication failure?

Check whether the API Key is correct (note the spaces before and after) and whether the network can access the API endpoint. OK /bash Use it after switching to bash co --debug View detailed errors.

What is cosh?

ANOLISA's default command line entry replaces the traditional bash and can interact with the OS in natural language.

How to switch from cosh to traditional bash?

Enter /bash in cosh to switch.

How to switch to cosh in bash?

Execute the exit command in bash or Ctrl+D to switch back to cosh.

What is the difference between installing OpenClaw in cosh mode and one-click deployment installation?

The installation of OpenClaw is triggered by natural language under COSH, and parameters are automatically obtained based on user input. If verification fails, it can be automatically located and resolved, and whether to add DingTalk/custom endpoint and other switches are decided according to the context. It can achieve "understanding requirements + decision-making routing + fault diagnosis", allowing users to install without having to remember parameters.

What is the principle of Token saving? For which scenarios does optimization have the greatest effect?

Tokenless uses five complementary strategies to reduce redundancy before content enters the LLM context window:

  1. Schema compression(~57%): Streamline Function Calling tool definition, remove markdown syntax in description, delete examples and title Fields, truncated function and parameter descriptions that are too long. The effect is most significant on Agent scenarios that contain a large number of tool definitions (dozens of Function Calling schemas).

  2. response compression(26–78%): Apply 7 rules to API returns and tool execution results - remove low-value fields such as debug/trace/stack/logs (R3), remove nulls and empty values ​​(R4/R5), truncate very long strings exceeding 4096 characters (R1), truncate arrays exceeding 32 elements (R2), truncate nesting depths exceeding 8 levels (R6). The input types with the best optimization results: API responses containing a lot of debugging information (such as Kubernetes API, cloud resource query results), web_fetch results containing redundant fields (measured ~78%), and deeply nested JSON structures.

  3. TOON encoding(15–40%): Losslessly encode JSON into a token-oriented compact format, eliminating JSON syntax overhead such as quotes, commas, and curly braces. For data with regular structure and long field names (such as containing descriptionresource_id Objects with keys of equal length) have the best optimization effect.

  4. Command rewriting(60–90%): The integrated RTK engine intelligently filters the output of 70+ CLI commands to eliminate interference information such as progress bars, ANSI escape codes, and redundant logs. Input types that optimize best: build tool output (cargo/npm/go/pytest), package manager output (dnf/yum/apt), and lengthy results of file listing commands (ls -la, find).

  5. Tool readiness check: Verify binaries, configurations, permissions and network dependencies before calling the tool, automatically fix missing items and mark environment class failures as "Skip retry" to prevent LLM from wasting tokens by repeatedly retrying commands that must fail.

Overall, Tokenless has the greatest optimization effect on the following input and output:Agent configuration with extensive tool definitions(Schema compression),API response with debug/redundant fields(response compression),Structured JSON data(TOON encoding),CI/CD build and package management command output(RTK rewrite).

Can AgentSecCore/ws-ckpt be enabled only in the Agentic version?

It can be enabled, and the corresponding components can be installed in the non-Agentic version through the anolisa cli command, or the source code can be compiled and installed directly.

Why does it not take effect after I modified the source code of the Agent plug-in?

Please check whether you modified the effective path of the corresponding Agent plug-in source code. The ANOLISA component plug-in installation path is in the /usr/share/anolisa directory. If you directly modify the plug-in code in the installation path, you need to uninstall and reinstall it in the corresponding Agent.

Do multiple user logins share the same configuration? Is the configuration independent after creating a new user?

Yes, the current system multi-user login does not share configurations and has independent model configurations and API Keys.

Does it support Chinese language setting?

Supported, the language can be set to Chinese in Copilot Shell.

Can built-in skills be used by Agents such as OpenClaw?

Yes, after installing OpenClaw, the system's built-in skills will be automatically loaded into its skill library to achieve compatibility.

Is ANOLISA open source?

The ANOLISA code has been open sourced, and the open source address is:https://github.com/alibaba/anolisa

How to get technical support?

Technical support DingTalk group: 90400034325