Document pip constraints CI enforcement for all dependencies

This work is being implemented in the main (instructlab/instructlab)
repository. The same approach should be applied across all the repos.

Signed-off-by: Ihar Hrachyshka <[email protected]>

Author: booxter

.spellcheck-en-custom.txt

@@ -120,6 +120,7 @@ Langgraph
 leaderboard
 lifecycle
 lignment
+linters
 LLM
 LLMs
 llms

docs/ci/dependency-constraints.md

@@ -0,0 +1,65 @@
+# Dependency Constraints in InstructLab CI
+
+This document describes how Python dependencies should be managed in
+InstructLab CI.
+
+## Goals
+
+1. Ensure that the CI environment is consistent and reproducible.
+2. Ensure that new dependency releases do not break the CI environment.
+3. Ensure that we adopt new dependency releases in the CI environment in a
+   timely manner.
+
+## Approach
+
+### Pin all versions with constraints files
+
+Each repository should contain a [pip
+constraints](https://pip.pypa.io/en/stable/user_guide/#constraints-files) file
+that lists the pinned versions of all dependencies used in the CI environment.
+
+In case a repository supports multiple platforms (`linux` vs `darwin`) or
+accelerators (`cuda`, `cpu`, `hpu`), multiple constraints files may be
+generated.
+
+The constraints files should be used in all CI jobs that install Python
+dependencies (using `pip` or otherwise), including linters, unit tests,
+integration tests, and functional tests.
+
+### Update constraints files regularly
+
+Each repository should define a new `tox` target called `constraints` that will
+re-generate all the constraints files in the project.
+
+Updates to these files should be generated automatically by the CI system and
+should not be modified manually (subject to rare exceptions). A periodic
+CI job should be added to ensure this happens on a regular basis (at least once
+a week).
+
+The job should run `tox -e constraints` and post the result as a PR for review.
+The PR will be validated by all the relevant CI jobs. Project core team is
+expected to review these PRs in a timely manner (within 3 business days). The
+team should make sure that all the relevant CI jobs are passing before merging
+the PR.
+
+### Uncap all dependencies
+
+At this point, no new dependency releases should affect the CI environment.
+Projects should then uncap all their dependencies in requirements files, as per
+[dependency management policy](../dependency-management.md).
+
+Note: In rare situations, a cap may be justified. For example, when we know for
+sure that a new release of the upstream dependency will break the project, and
+when we don't have capacity to deliver a fix in a timely manner. These
+situations should be rare and a mitigation plan should be in place to uncap the
+dependency.
+
+Specifically,
+
+- A tracking issue should be reported in the issue tracking system capturing
+  any known details about the issue.
+- A new **temporary** constraint should be added to `constraints-dev.txt.in`
+  file. The temporary constraint should refer to the tracking issue in a
+  comment above it.
+- The tracking issue should be assigned a high priority and considered a
+  blocker for an upcoming release.

docs/dependency-management.md

@@ -11,7 +11,7 @@ This document describes the policies for adding and updating build and runtime d
 4. Only apply "caps" to dependencies (using `<`) when that dependency has established a pattern of producing new releases with breaking changes.
 5. Pin versions in CI, with frequent automated updates.
 
-Best practices for python dependencies call for using ranges in package requirements and pinning versions only in CI jobs.
+Best practices for python dependencies call for using ranges in package requirements and [pinning versions only in CI jobs](./ci/dependency-constraints.md).
 
 Using pinned versions in a list of constraints used for tests allows us to know and advertise exactly what versions have been tested in CI.
 That information is useful for users and re-packagers to understand which versions of dependencies are compatible with more specificity than the ranges provide.