fix: remove pysqlite3-binary auto-install via subprocess at import time

[rkfshakti]

Summary

Removes the that runs at import time when Chroma detects an old SQLite version in Google Colab.

Problem

Running pip install via subprocess at import time is fragile:

• Fails on non-standard architectures (ppc64, arm64 variants, etc.) where the user needs to specify a custom package index
• Is generally considered bad practice — package dependencies should be managed by the package manager, not at runtime
• The auto-install path was only triggered in Colab, but the same underlying issue (old SQLite) exists everywhere

Fix

• Removed the subprocess.check_call auto-install block
• Removed the now-unused is_in_colab() helper and IN_COLAB constant
• The Colab path now raises the same clear RuntimeError as the non-Colab path, directing users to the troubleshooting guide

Closes #7402

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[github-actions]

Reviewer Checklist

Please leverage this checklist to ensure your code review is thorough before approving

Testing, Bugs, Errors, Logs, Documentation

• Can you think of any use case in which the code does not behave as intended? Have they been tested?
• Can you think of any inputs or external events that could break the code? Is user input validated and safe? Have they been tested?
• If appropriate, are there adequate property based tests?
• If appropriate, are there adequate unit tests?
• Should any logging, debugging, tracing information be added or removed?
• Are error messages user-friendly?
• Have all documentation changes needed been made?
• Have all non-obvious changes been commented?

System Compatibility

• Are there any potential impacts on other parts of the system or backward compatibility?
• Does this change intersect with any items on our roadmap, and if so, is there a plan for fitting them together?

Quality

• Is this code of a unexpectedly high quality (Readability, Modularity, Intuitiveness)

[eeshsaxena]

Strongly agree with the intent. Shelling out to pip install as a side effect of import chromadb is a genuine anti-pattern - it does network I/O and mutates the environment at import time, then swaps sys.modules["sqlite3"] out from under everything else in the process. Replacing it with the same clear RuntimeError + docs link that every other platform already gets is the right call.

However, I think this breaks client startup as written. The PR deletes is_in_colab from chromadb/__init__.py, but it still has a live consumer:

# chromadb/telemetry/product/events.py
class ClientStartEvent(ProductTelemetryEvent):
    def __init__(self) -> None:
        super().__init__()
        # Lazy import to avoid circular imports
        from chromadb import is_in_colab

        self.in_colab = is_in_colab()

and ClientStartEvent is constructed unconditionally on the client-start path:

# chromadb/api/shared_system_client.py
def _submit_client_start_event(self) -> None:
    telemetry_client = self._system.instance(ProductTelemetryClient)
    telemetry_client.capture(ClientStartEvent())

The event object is built before capture() gets to decide whether telemetry is even enabled, so ClientStartEvent() runs and hits ImportError: cannot import name 'is_in_colab' from 'chromadb'.

The nasty part is that the import is lazy (inside __init__), so this does not fail at import time - import chromadb still succeeds, and a smoke-import test in CI would stay green. It only blows up when a client is actually created, which is the very first thing a user does.

Two ways out, depending on what you want:

1. Keep is_in_colab, drop only the hotswap. The google.colab probe is cheap and side-effect-free, and telemetry legitimately wants it. This PR's actual goal is removing the subprocess/pip install/sys.modules surgery, and that's fully achieved without deleting the predicate.
2. If you do want is_in_colab gone from the public namespace, events.py needs to move the probe inline (or import it from an internal module) in the same PR.

Either way it'd be worth adding a test that constructs a client (not just imports the package), since that's the only thing that would have caught this.

Deferring to the maintainers.

[rkfshakti]

Thanks @eeshsaxena for catching that! You're absolutely right — is_in_colab() is still used by chromadb/telemetry/product/events.py for telemetry.

I've pushed a fix commit that restores is_in_colab() but keeps the subprocess/pip install hotswap removed. The change is now scoped to:

• Removing the IN_COLAB constant (no longer needed)
• Removing the subprocess.check_call pip install block
• Keeping is_in_colab() since telemetry still needs it
• The old SQLite version now raises the same RuntimeError for both Colab and non-Colab paths

Could you take another look?

[rkfshakti]

Hi Team, just checking if there's anything else needed on this PR. Thanks!

[rkfshakti]

@rescrv @HammadB @codetheweb — this PR removes the subprocess/pip-install hotswap for pysqlite3-binary at import time (a known anti-pattern) while keeping the is_in_colab predicate that telemetry needs. @eeshsaxena reviewed and agreed with the intent; the fix commit (acc3114) addresses the concern about the broken import. Could a maintainer take a look? Thanks!

[rkfshakti]

Thanks for the review @eeshsaxena — you were right that deleting is_in_colab from __init__.py would break ClientStartEvent. I've inlined the colab check directly in ClientStartEvent.__init__ (simple try/except import google.colab) so the function can be safely removed from __init__.py. Pushed to the branch.

[rkfshakti]

Hi maintainers — just a friendly bump on this one. This removes the subprocess/pip-install hotswap for pysqlite3-binary at import time (a known anti-pattern). @eeshsaxena reviewed and agreed with the approach; the fix commit addresses the colab telemetry concern. Would appreciate a maintainer review when you get a chance. Thanks!

[rkfshakti]

Hi maintainers — just circling back on this one. I'm genuinely curious to hear your thoughts on the approach. The fix removes the subprocess/pip-install hotswap at import time while keeping the colab telemetry predicate intact. Happy to adjust if you'd prefer a different direction. Keen to contribute and learn from your feedback. Thanks!

[rkfshakti]

Friendly ping — this PR has been open for a couple of days with CI passing. Would appreciate a review when time allows.

[rkfshakti]

Hi maintainers — following up again on this fix for #7402. Removes the subprocess/pip-install hotswap for pysqlite3-binary and lets the system sqlite3 be used directly. CI is green. Would appreciate a review when time allows. Thanks!