Dataspell 2024.3 Conda auto-detection picks latest env instead of pinned spec — how to lock interpreter per project

Answer

Yes. In Dataspell 2024.3 you can bind a project to a specific Conda interpreter by using the Existing environment option under Settings → Project: Python Interpreter → Add → Conda Environment. Once set, the IDE will stop auto‑selecting the most recently modified environment.

Likely explanation

Dataspell’s auto‑detect routine scans known Conda environment locations and, when it finds multiple matches, chooses the environment whose directory was most recently modified. It does not consult environment.yml or lock files to pick a particular version unless an interpreter is explicitly configured.

Confirmed facts

• When the “Make available to all projects” toggle is turned off, Dataspell stores the interpreter choice in the project‑specific .idea/misc.xml file (look for <option name="projectSdkPath">).
• Disabling that toggle prevents the interpreter from being offered as a global interpreter, but Conda’s shared package cache and pip’s download cache remain outside IDE control.
• If the interpreter path recorded in .idea/misc.xml stays valid, reopening the project will reuse the same interpreter and no new environment will be auto‑selected.

Steps to lock the interpreter

1. Create the environment from the pinned spec (do this before launching Dataspell):
   conda env create -f environment.yml -n myproj_env
2. Locate the Python executable inside the new environment, e.g.
   $CONDA_ROOT/envs/myproj_env/bin/python (or Scripts\python.exe on Windows).
3. Open Dataspell, go to Settings → Project: Python Interpreter.
4. Click the gear icon, choose Add… → Conda Environment → Existing environment.
5. Browse to the Python executable you noted in step 2 and select it.
6. Make sure Make available to all projects is off.
7. Click OK to apply.

Verification

1. Check the IDE status bar or Settings → Project: Python Interpreter; it should show the full path to the env’s Python.
2. Open .idea/misc.xml and confirm the projectSdkPath matches that path.
3. Close the project, reopen it, and verify the interpreter remains unchanged and no new Conda environment appears in the interpreter list.
4. In a fresh clone, repeat the interpreter assignment, then run conda list or pip freeze from the IDE’s terminal and compare the output to the original spec to ensure identical package sets.

Missing diagnostic detail

Do all machines share the same Conda root path (so the interpreter path stays valid after cloning), or does the env location differ per clone? If the path varies, you will need to repeat the interpreter selection on each machine or adopt a strategy that uses a relative or variable‑based path.