Code-Interpreter Sandbox Support & Base Image Clarification for v2 Templates

[Joao-Tiago-Almeida]

Hello, We’re migrating our templates to v2 and rely on the notebook/Jupyter-style “code-interpreter” sandbox. We have a few questions:

• Do you plan to publish a dedicated “code-interpreter v2” image (or register it via AWS/GCP container registry) rather than relying on the legacy template/image? We’d like to understand if the notebook/Jupyter-style sandbox will evolve into a non-legacy version, and when that might be expected. (Troubleshooting docs reference)

• What is the official base image we should use under v2 templates for code-interpreter? Is e2bdev/code-interpreter:latest still recommended?

• If we build a custom sandbox (not using the official code-interpreter image/template), which notebook features (cells, context, plotting) might be lost?

• What is the long-term support plan for the notebook/Jupyter-style sandbox? Some docs are marked “legacy (pre v1.0)”. Is there an upcoming deprecation or migration path?

Thanks in advance for your guidance.

---

João Almeida | Founding Engineer @ DOJO AI

[linear[bot]]

ENG-3309

[antonioalegria]

Also interested in knowing this. The E2B docs barely mention the Code Interpreter use case and templates now.

[hewliyang]

fwiw i don't think this image is "legacy". you can continue to use it as before:

template = (
    Template()
    .from_image("e2bdev/code-interpreter:latest")
    .run_cmd("pip install {{ }}")
    .set_start_cmd(
        ".jupyter/start-up.sh", wait_for_url("http://localhost:49999/health")
    )

key thing is just adding an explicit start command, otherwise the underlying RPC server doesnt run

[mishushakov]

Hey there, I believe my colleague already answered you on Discord, but here's my take:

Do you plan to publish a dedicated “code-interpreter v2” image (or register it via AWS/GCP container registry) rather than relying on the legacy template/image?

There's no legacy image, the code-interpreter-v1 image is the latest one available.

What is the official base image we should use under v2 templates for code-interpreter? Is e2bdev/code-interpreter:latest still recommended?

If you're using the legacy version of the build system - nothing changes for you (keep using e2bdev/code-interpreter:latest), if you're using the new version, you should use .fromTemplate("code-interpreter-v1") / .from_template("code-interpreter-v1") in JavaScript and Python respectively.

If we build a custom sandbox (not using the official code-interpreter image/template), which notebook features (cells, context, plotting) might be lost?

None, we are supporting all of the said features still.

What is the long-term support plan for the notebook/Jupyter-style sandbox? Some docs are marked “legacy (pre v1.0)”. Is there an upcoming deprecation or migration path?

We are actively maintaining the Code Interpreter Template as well as the SDKs and there are no plans for deprecation. The "legacy" docs are for our build system, there's no deprecation plan and the migration path can be found under V2 Migration Guide (for custom templates).

I am closing the issue, but happy to continue the conversation, in case you have any more questions. Thanks :)