본문으로 건너뛰기

Environment Configuration

FluxLoop automatically detects and uses your project's Python environment to ensure experiments run with the correct dependencies.

Quick Setup​

After installing the FluxLoop VSCode extension:

  1. Create or activate a virtual environment in your project:

    python -m venv .venv
    source .venv/bin/activate

  2. Install FluxLoop packages:

    pip install fluxloop-cli fluxloop fluxloop-mcp

  3. Verify detection in VSCode:

    • Open Command Palette (Cmd+Shift+P)
    • Run FluxLoop: Show Environment Info
    • Confirm paths point to your .venv/bin/ directory

  4. Set target source root (if needed):

    • Expand your project in the Projects view
    • Click Target Source Root…
    • Select the directory containing .venv

FluxLoop will now use this environment for all commands.

Automated Setup Script​

For new projects or onboarding, use the setup script:

cd your-project
bash packages/cli/scripts/setup_fluxloop_env.sh

This automatically:

  • Creates .venv if missing
  • Installs FluxLoop packages
  • Generates .vscode/settings.json with environment defaults
  • Runs fluxloop doctor to verify setup

With custom Python:

bash packages/cli/scripts/setup_fluxloop_env.sh --python python3.11

With target source root:

bash packages/cli/scripts/setup_fluxloop_env.sh --target-source-root ./backend

Execution Modes​

FluxLoop supports four execution modes to match your workflow.

Automatically detects project virtual environments (.venv, venv, .conda) and uses them. Falls back to global PATH if not found.

When to use:

  • Standard projects with optional virtual environments
  • Mixed teams (some use venv, others use global installs)

Set via Command Palette:

FluxLoop: Select Environment → Auto (detect project environment)

Workspace Only​

Requires a virtual environment in the project. Commands fail if venv is not detected.

When to use:

  • Strict dependency isolation
  • Teams that always use virtual environments
  • CI/CD pipelines with project-specific dependencies

Set via Command Palette:

FluxLoop: Select Environment → Workspace only

Global PATH​

Always uses globally installed FluxLoop executables from PATH (e.g., installed via pipx).

When to use:

  • Global FluxLoop installation (pipx)
  • Simple projects without virtual environments
  • Quick prototyping

Set via Command Palette:

FluxLoop: Select Environment → Global PATH

Custom Executables​

Manually specify absolute paths to Python and fluxloop-mcp executables.

When to use:

  • Non-standard installation locations
  • Multiple Python versions
  • Advanced conda/pyenv setups

Set via Command Palette:

FluxLoop: Select Environment → Custom executables…

Then enter:

  • Python executable path (e.g., /opt/python3.11/bin/python)
  • fluxloop-mcp path (optional, e.g., /custom/bin/fluxloop-mcp)

Checking Environment Status​

Show Environment Info​

Quick modal summary of the active environment:

FluxLoop: Show Environment Info

Shows:

  • Source root
  • Environment type (venv, conda, global, etc.)
  • Python path
  • fluxloop path
  • fluxloop-mcp path
  • Any detection notes

Run Doctor​

Comprehensive diagnostics including MCP index and project config:

From Command Palette:

FluxLoop: Run Doctor

From Integration View:

  1. Open Integration → System Status
  2. Click Run Doctor

From Terminal:

fluxloop doctor

Doctor output includes:

  • Python version and executable
  • Virtual environment status
  • FluxLoop CLI version and path
  • FluxLoop MCP availability
  • MCP index location and size
  • Project configuration state

Use --json for machine-readable output:

fluxloop doctor --json

FluxLoop Output Channel​

Detailed environment logs appear in the Output channel:

  1. View → Output (or Shift+Cmd+U)
  2. Select FluxLoop from dropdown
  3. Review environment detection logs

Every time the environment is refreshed, you'll see:

[FluxLoop Env] ----------------------------------------
[FluxLoop Env] Source root: /path/to/project
[FluxLoop Env] Environment type: venv
[FluxLoop Env] Python: /path/.venv/bin/python
[FluxLoop Env] fluxloop: /path/.venv/bin/fluxloop
[FluxLoop Env] fluxloop-mcp: /path/.venv/bin/fluxloop-mcp
[FluxLoop Env] Notes:
[FluxLoop Env]  • Python detected but FluxLoop packages were not found in this environment.
[FluxLoop Env] ----------------------------------------

Common Scenarios​

Using uv​

  1. Create uv project:

    uv venv
    source .venv/bin/activate
    uv pip install fluxloop-cli fluxloop fluxloop-mcp

  2. FluxLoop automatically detects .venv/

  3. (Optional) Set execution wrapper if you want to use uv run:

    FluxLoop: Configure Execution Wrapper → uv run

Using Conda​

  1. Create and activate conda environment:

    conda create -n myproject python=3.11
    conda activate myproject
    pip install fluxloop-cli fluxloop fluxloop-mcp

  2. If conda env is in a non-standard location:

    FluxLoop: Select Environment → Custom executables…

    Enter conda Python path (e.g., /opt/conda/envs/myproject/bin/python)

Multiple Projects​

FluxLoop caches environment per project. When switching:

  1. Environment is re-detected automatically
  2. Each project can have different:
    • Execution mode
    • Target source root
    • Custom paths

Configure per-project in each project's .vscode/settings.json.

Team Setup​

For consistent team environments, commit to your repo:

.vscode/settings.json:

{
  "fluxloop.executionMode": "workspace",
  "fluxloop.targetSourceRoot": "./src"
}

README.md or setup script:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
fluxloop doctor

Or use the automated script:

bash packages/cli/scripts/setup_fluxloop_env.sh --target-source-root ./src

Settings Reference​

fluxloop.executionMode​

Controls environment selection strategy.

  • Type: string
  • Values: auto | workspace | global | custom
  • Default: auto
  • Scope: Workspace or folder

fluxloop.targetSourceRoot​

Path to FluxLoop project source root for environment detection.

  • Type: string
  • Default: "" (uses project root or source_root from project.yaml)
  • Scope: Workspace or folder
  • Supports: Relative paths (from project) or absolute paths

fluxloop.pythonPath​

Custom Python executable path (overrides detection).

  • Type: string
  • Default: "" (auto-detect)
  • Scope: Workspace or folder
  • Example: /Users/me/project/.venv/bin/python

fluxloop.mcpCommandPath​

Custom fluxloop-mcp executable path.

  • Type: string
  • Default: "" (auto-detect)
  • Scope: Workspace or folder
  • Example: /Users/me/project/.venv/bin/fluxloop-mcp

fluxloop.executionWrapper​

Command prefix for CLI invocations (e.g., uv run).

  • Type: string
  • Default: ""
  • Scope: Workspace
  • Example: uv run, pipx run

Next Steps​