Make ChatClient APIs null-safe and predictable #1651
Conversation
* Introduced null-safety for all ChatClient APIs and verified via extensive auto-tests. * Made the final message list computed by ChatClient consistent and predictable across the different options for providing messages (prompt(), messages(), user(), prompt()) and verified via extensive auto-tests. * Added even more auto-tests to cover as many scenarios and edge cases as possible. Signed-off-by: Thomas Vitale <[email protected]>
| return new AdvisedRequest(inputRequest.chatModel, inputRequest.userText, inputRequest.systemText, | ||
| inputRequest.chatOptions, inputRequest.media, inputRequest.functionNames, | ||
| inputRequest.functionCallbacks, inputRequest.messages, inputRequest.userParams, | ||
| // Process userText, media and messages before creating the AdvisedRequest. |
There was a problem hiding this comment.
This piece of logic used to be executed only in one possible branch of execution (when using .prompt(Prompt prompt)). I moved it here so to ensure consistent processing of the messages before creating the AdvisedRequest, no matter how we get here (for example, when using .messages(List<Message> messages), which was previously behaving inconsistently).
There was a problem hiding this comment.
yeh, good catch.
Initially the ChatClinet from Prompt used to have a very different path, excluded from the advisors.
I've tried to unify it with there unsealed path (by unsealing the prompt) but have missed the ordering of this section.
|
Very useful improvements for reliability and consistency. Thanks @ThomasVitale |
|
Rebased and merged at 048d54b |
2 participants
prompt(),messages(),user(),system()) and verified via extensive auto-tests.