Hi Jason, I would go with option 1 and only mention the tool when it performs a significant part of the work.
That said, I don’t feel strongly about it. Since ASF recommendations will likely change again soon, I don't think we need to overthink this. Best, Stamatis On Sat, Jun 13, 2026 at 1:07 AM Jason Fehr <[email protected]> wrote: > It's not a requirement to document what AI tooling was used, but it is > recommended by Apache: > > https://www.apache.org/legal/generative-tooling.html#:~:text=When%20providing%20contributions%20authored%20using%20generative%20AI%20tooling%2C%20a%20recommended%20practice%20is%20for%20contributors%20to%20indicate%20the%20tooling%20used%20to%20create%20the%20contribution > . > > On Fri, Jun 12, 2026 at 2:23 PM Edward Capriolo <[email protected]> > wrote: > > > I think it is somewhat of a mute point. My ide uses a "model" to do code > > completions. Generate toString() . > > My ide if I type some java will say " hey that is throwing an exception, > I > > will add try catch for you!. I can barely belt out a for loop before it > > auto replaces it .map().reduce().collect() . > > > > All this "assisted by :" does is contributes to am ipo of open ai vendor > by > > giving them bragging rights. NetBeans never got any credit for auto > > completing jbutton.onclick for me :) > > > > > > On Fri, Jun 12, 2026, 5:02 PM Michael Smith <[email protected]> > > wrote: > > > > > I agree with option 3. > > > > > > On Fri, Jun 12, 2026 at 10:03 AM Jason Fehr <[email protected]> > wrote: > > > > > > > We have added a requirement to include a commit message trailer > > > > "Assisted-by: <model> (<tool>)" when using generative tooling (see > > > > > > > > > > > > > > https://cwiki.apache.org/confluence/display/IMPALA/Contributing+to+Impala#:~:text=If%20using%20generative%20tooling > > > > ). > > > > Next, we need to consider the situation where multiple generative > tools > > > are > > > > used on the same commit. For example, multiple models may be used to > > > > approach different problems on a single commit or may be used to get > > > > different perspectives on the same problem. Also, multiple tools > could > > > be > > > > used for the same reason. > > > > > > > > How should we document that multiple models/tools were used? Here > are > > > the > > > > options I can see: > > > > 1. Add an "Assisted-by" trailer only for the primary model/agent that > > was > > > > used. > > > > Pros: Simple > > > > Cons: Loses information/context about the commit, may not always > > > have a > > > > primary model/tool that was used. > > > > 2. Use a single "Assisted-by" trailer listing all models/agents. For > > > > example: "Assisted-by: GPT-5.3-Codex, Sonnet 4.6 (Copilot, Claude > > Code)". > > > > Pros: Includes all generative tooling information in a single > line. > > > > Cons: Loses some information/context about the commit since it's > > not > > > > clear which model was used with which agent. > > > > 3. Use a single "Assisted-by" trailer per agent. Each agent will > have > > > its > > > > own "Assisted-by" trailer in the commit message with multiple models > > > > potentially listed. > > > > For example: > > > > Assisted-by: GPT-5.3-Codex, Sonnet 4.6 (Copilot) > > > > Assisted-by: Sonnet 4.6 (Claude Code) > > > > Pros: Does not hide information. > > > > Cons: Potentially more difficult to parse. > > > > 4. Use a single "Assisted-by" trailer per model/agent. > > > > For example: > > > > Assisted-by: GPT-5.3-Codex (Copilot) > > > > Assisted-by: Sonnet 4.6 (Copilot) > > > > Assisted-by: Sonnet 4.6 (Claude Code) > > > > Pros: Does not hide information. > > > > Cons: Most verbose. > > > > > > > > My vote is for option 3. It is the most concise without losing > > > > information. > > > > > > > > Thanks, > > > > Jason > > > > > > > > > >
