Build on Windows 11, add some feture

[hero]

I've added some features based on v0.196.6:

1. On Windows, Zed's MCP Server fails to start, displaying the following error message:

Error: Cannot find module 'C:\C:\Users\xxx\AppData\Local\Zed\extensions\work\mcp-server-context7\node_modules\@upstash\context7-mcp\dist\index.js'

Root Cause: The C:\ in the path is repeated. This is because the argument contains a malformed path: "/C:\path\to\file". The argument begins with / but is followed by the Windows absolute path C:\. Node.js Interpretation: Node.js interprets /C:\ as a path starting from the root directory, which becomes C:\C:\ on Windows.

Fix it:

// 1. Windows path fix
// crates\context_server\src\transport\stdio_transport.rs, Line 28~45 add the following:
// Fix Windows path issues: remove leading "/" from arguments that look like Windows paths
        #[cfg(target_os = "windows")]
        let fixed_args: Vec<String> = binary.args.iter().map(|arg| {
            if arg.starts_with("/") && arg.len() > 3 {
                let chars: Vec<char> = arg.chars().collect();
                if chars.len() > 2 && chars[1].is_ascii_alphabetic() && chars[2] == ':' {
                    // This looks like "/C:\path\to\file" - remove the leading "/"
                    arg[1..].to_string()
                } else {
                    arg.clone()
                }
            } else {
                arg.clone()
            }
        }).collect();

        #[cfg(not(target_os = "windows"))]
        let fixed_args = &binary.args;
        
        command
            .args(&fixed_args)

// 2. MCP server list fix
// crates\agent_ui\src\agent_configuration.rs, Line 497
// Before
let context_server_ids = self.context_server_store.read(cx).configured_server_ids();

// After
let context_server_ids = self.context_server_store.read(cx).all_server_ids();

2. The Ollama base URL can be customized.

Not all Ollama are deployed on the local machine. Our laboratory has a dedicated AI large model server.

3. The LMStudio base URL can be customized.

Of course, you can modify the configuration file assets\settings\default.json directly, but I still hope that it can be easily modified on the UI interface.

4. The Google AI API URL is now customizable, similar to the OpenAI one.

5. Google AI environment variables are compatible with the Gemini CLI tool.

PS: The Zed team is incredibly diligent. I just finished modifying the previous release, and I just noticed it's already v0.196.7. I'm not planning to modify it again, but I hope the official team can merge it in if you find these features useful.

Here are my steps:

1. Clone the source code and checkout v0.196.6:

git clone https://github.com/zed-industries/zed.git
git checkout v0.196.6 -b dev

2. Case-sensitive bulk replacement:

• GOOGLE_AI_API_KEY --> GEMINI_API_KEY
• google_ai_api_key --> gemini_api_key

This is done to ensure consistency with the environment variables required by the Gemini CLI.

3. I mainly modified these files:

mkdir -p patches
git diff v0.196.6 -- crates\language_models\src\provider\google.rs > .\patches\google_rs.patch
git diff v0.196.6 -- crates\language_models\src\provider\ollama.rs > .\patches\ollama_rs.patch
git diff v0.196.6 -- crates\language_models\src\provider\lmstudio.rs > .\patches\lmstudio_rs.patch
git diff v0.196.6 -- crates\context_server\src\transport\stdio_transport.rs > .\patches\stdio_transport_rs.patch
git diff v0.196.6 -- crates\agent_ui\src\agent_configuration.rs > .\patches\agent_configuration_rs.patch
# This is the complete Git diff patch file after my modification
git diff v0.196.6 -- > .\patches\my_full_changes.patch

patches_on_v0.196.6.zip

4. you can test it (If your environment is not set up well, refer to the next section first.):

git apply  .\patches\my_full_changes.patch
cargo run
# or, build in PowerShell
$env:RELEASE_CHANNEL="stable"; cargo build --release

5. some tips:

• To install Visual Studio 2022, you need to meet these conditions.
• To install rustup on Windows 11:

# Install Rustup with [scoop](https://scoop.sh/)
scoop install rustup
rustup show

# Update rustup itself
rustup self update

# Update all toolchains
rustup update

# Verify fixes
rustup check

# Reinstall the toolchain required to compile the project
rustup toolchain install 1.88-x86_64-pc-windows-msvc

# Verify installation
rustup run 1.88-x86_64-pc-windows-msvc cargo --version

# Run the following commands to diagnose the problem
# 1. Check the default toolchain's cargo
cargo --version

# 2. Check the 1.88 toolchain's cargo
rustup run 1.88-x86_64-pc-windows-msvc cargo --version

# 3. Check the toolchain installation path
rustup show --verbose

# Run PowerShell as administrator and execute the following commands
# Enable Windows long path support
New-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem" -Name "LongPathsEnabled" -Value 1 -PropertyType DWORD -Force

# Verify the setting
Get-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem" -Name "LongPathsEnabled"

# Enable long path support for Git
git config --global core.longpaths true

# Verify the setting
git config --global --get core.longpaths

• Install the Spectre Mitigation Library (to protect against the Spectre CPU vulnerability)

1. Open the Visual Studio Installer

2. Modify the Visual Studio 2022 installation:

• Click the "Modify" button next to the installed Visual Studio 2022 installation.
• Search for "Spectre" in the "Individual Components" tab.
• Ensure the following components are selected:
• MSVC v143 - VS 2022 C++ x64/x86 Spectre Mitigation Library (Latest)

3. After installation is complete, you can verify that the Spectre library is installed correctly:

# Find the Spectre library
Get-ChildItem "C:\Program Files\Microsoft Visual Studio\2022\*\VC\Tools\MSVC\*\lib\*spectre*" -Recurse -ErrorAction SilentlyContinue

That's all, hope it helps