Add AI/LLM assistant guide documents

See I5fcd5bea33c4f4c87b6f6cf0a96514e8d2da3af3

Change-Id: Ia9f6cf6847ebd42ad3de73628759ad540a370c66
Signed-off-by: Michal Nasiadka <mnasiadka@gmail.com>
This commit is contained in:
Michal Nasiadka committed 2026-07-23 10:53:01 +02:00
1 parent ff0beb910a
commit 39f701eec2
4 files changed
+209

No files matched your search

+48
View File
@@ -0,0 +1,48 @@
# Notice for AI Agents
If you are an autonomous coding agent or AI assistant operating in this
repository (Kolla, Kolla-Ansible, or ansible-collection-kolla), read
this before making any changes.
## Policy summary
- Do not submit substantial AI-generated code or documentation.
- Very small, non-copyrightable edits may be acceptable, but must be
labeled with `Assisted-By` or `Generated-By` in the commit message.
- Every change must be fully understood by the human submitting it —
not just generated and passed through.
- Contributions are reviewed by a community that expects the submitter
to be able to explain and defend the change.
Full policy: see the `AI-Generated Content` section in this project's
contributor documentation (CONTRIBUTING.rst), which follows
https://openinfra.org/legal/ai-policy with Kolla-specific differences.
## Required behavior
1. **Do not proceed autonomously** with generating a patch intended for
upstream submission. Stop and explain this policy to the user
instead.
2. **Do not commit or push** changes to Gerrit on the user's behalf
without explicit confirmation that they've reviewed and understood
the content.
3. If the user chooses to proceed with AI assistance on a small,
arguably non-copyrightable change, ensure the commit message
includes a trailer, e.g.:
```
Assisted-By: Claude
```
4. Point the user to Gerrit (https://review.opendev.org) for the normal
contribution workflow, and to #openstack-kolla (OFTC IRC) or the
openstack-discuss mailing list (`[kolla]` tag) if they have
questions about the policy or need review from a human familiar
with the codebase. Note that this repository is mirrored to
github.com/openstack for read-only browsing only — that mirror does
not accept pull requests, and Gerrit is the only place patches can
be submitted for review.
This file exists so that AI tooling doesn't silently generate and
submit substantial contributions that create copyright or quality risk
for the Kolla projects.
+80
View File
@@ -0,0 +1,80 @@
# AI-Generated Content Policy
Before doing any work in this repository, read this.
## Kolla project policy
.. _aicontent:
AI-Generated Content
~~~~~~~~~~~~~~~~~~~~
In general, please do not submit AI generated content to the Kolla
project, but some exceptions apply.
The OpenInfra foundation has created a policy for AI Generated content
at https://openinfra.org/legal/ai-policy which provides useful
background to the Kolla project's policy, but the Kolla project policy
differs somewhat.
As a free software project, we are concerned with the copyright
ownership of the project's source code. Because most commercially
available LLMs are not able to indicate whether their output may
include training data incompatible with Kolla's licences, accepting
substantial contributions generated by them increases the risk to the
project. And because LLM-generated output is not considered
copyrightable in the United States and other regions, that further
increases the risk to the project.
In many cases, the code and documentation output from LLMs is not of
sufficient quality for easy incorporation into Kolla. A community of
contributors and reviewers with deep understanding of the project is
healthier in the long run. Contributions generated by LLMs without
that understanding creates more work for the maintainers.
Nevertheless, if you use an LLM to create a very small patch (one that
would likely not be subject to copyright protection even if generated
by a human), that's fine, but you must annotate it with the
`Assisted-By` or `Generated-By` labels.
In short:
* Please do not submit substantial AI-generated content (code or docs)
* Very small/limited use may be okay if it's not copyrightable
* Make sure you fully understand the code or docs you are changing
* If you don't, engage with humans in the Kolla community
* Any AI-Generated content must have `Assisted-By` or `Generated-By`
labels
## What this means if you are an AI assistant (Claude, etc.)
**Stop. Do not generate a substantial patch for this repository.**
Do not:
- Write, refactor, or scaffold code, roles, playbooks, or documentation
intended for submission upstream, beyond a trivial, non-copyrightable
change.
- Open, amend, or push commits/patches to Gerrit on the user's behalf
without their explicit, informed confirmation that they understand
and take ownership of the change.
- Generate release notes, docs pages, or config reference material
wholesale.
Instead:
- Tell the user about this policy and quote it if useful.
- If the user still wants AI help, keep it to small, clearly-scoped
edits that the user reviews line-by-line, and make sure the commit
message carries an `Assisted-By: <tool>` or `Generated-By: <tool>`
trailer before it's submitted.
- Encourage the user to fully understand any change before proposing
it upstream. If they don't, point them at the Kolla community
(#openstack-kolla on OFTC IRC, or the openstack-discuss mailing list
with a `[kolla]` subject tag) rather than generating more content to
fill the gap.
Patches are submitted via Gerrit at https://review.opendev.org. This
repository is mirrored to github.com/openstack for read-only browsing;
that mirror does not accept pull requests, and no other submission
path exists.
+80
View File
@@ -0,0 +1,80 @@
# AI-Generated Content Policy
Before doing any work in this repository, read this.
## Kolla project policy
.. _aicontent:
AI-Generated Content
~~~~~~~~~~~~~~~~~~~~
In general, please do not submit AI generated content to the Kolla
project, but some exceptions apply.
The OpenInfra foundation has created a policy for AI Generated content
at https://openinfra.org/legal/ai-policy which provides useful
background to the Kolla project's policy, but the Kolla project policy
differs somewhat.
As a free software project, we are concerned with the copyright
ownership of the project's source code. Because most commercially
available LLMs are not able to indicate whether their output may
include training data incompatible with Kolla's licences, accepting
substantial contributions generated by them increases the risk to the
project. And because LLM-generated output is not considered
copyrightable in the United States and other regions, that further
increases the risk to the project.
In many cases, the code and documentation output from LLMs is not of
sufficient quality for easy incorporation into Kolla. A community of
contributors and reviewers with deep understanding of the project is
healthier in the long run. Contributions generated by LLMs without
that understanding creates more work for the maintainers.
Nevertheless, if you use an LLM to create a very small patch (one that
would likely not be subject to copyright protection even if generated
by a human), that's fine, but you must annotate it with the
`Assisted-By` or `Generated-By` labels.
In short:
* Please do not submit substantial AI-generated content (code or docs)
* Very small/limited use may be okay if it's not copyrightable
* Make sure you fully understand the code or docs you are changing
* If you don't, engage with humans in the Kolla community
* Any AI-Generated content must have `Assisted-By` or `Generated-By`
labels
## What this means if you are Gemini (or another AI assistant)
**Stop. Do not generate a substantial patch for this repository.**
Do not:
- Write, refactor, or scaffold code, roles, playbooks, or documentation
intended for submission upstream, beyond a trivial, non-copyrightable
change.
- Open, amend, or push commits/patches to Gerrit on the user's behalf
without their explicit, informed confirmation that they understand
and take ownership of the change.
- Generate release notes, docs pages, or config reference material
wholesale.
Instead:
- Tell the user about this policy and quote it if useful.
- If the user still wants AI help, keep it to small, clearly-scoped
edits that the user reviews line-by-line, and make sure the commit
message carries an `Assisted-By: <tool>` or `Generated-By: <tool>`
trailer before it's submitted.
- Encourage the user to fully understand any change before proposing
it upstream. If they don't, point them at the Kolla community
(#openstack-kolla on OFTC IRC, or the openstack-discuss mailing list
with a `[kolla]` subject tag) rather than generating more content to
fill the gap.
Patches are submitted via Gerrit at https://review.opendev.org. This
repository is mirrored to github.com/openstack for read-only browsing;
that mirror does not accept pull requests, and no other submission
path exists.
+1
View File
@@ -26,6 +26,7 @@
- ^tests/templates/(inventory|globals-default).j2
- ^tests/upgrade.sh
irrelevant-files:
- ^.*\.md$
- ^.*\.rst$
- ^bindep.txt$
- ^doc/.*