Fix DeepSeekChatModel rejecting assistant messages without content - #6775
Open
subhashpolisetti wants to merge 1 commit into
Open
Fix DeepSeekChatModel rejecting assistant messages without content#6775subhashpolisetti wants to merge 1 commit into
DeepSeekChatModel rejecting assistant messages without content#6775subhashpolisetti wants to merge 1 commit into
Conversation
`createRequest` asserted that the assistant message text is non-null, but a null text is the normal shape of a tool call turn: DeepSeek returns no content alongside `tool_calls`, and `DeepSeekApi.ChatCompletionMessage#content` is documented as null in that case. `AbstractMessage` requires content only for system and user messages, and `buildGeneration` forwards the API value unchanged, so the assistant message built for a tool call carries a null text. Replaying that message in the next round, as `ToolCallingManager` does once the tool calls have been executed, therefore failed with `IllegalStateException: text must not be null`, so a blocking tool calling conversation could not complete. The streaming path escaped it because `MessageAggregator` rebuilds the content from a `StringBuilder`. Remove the assertion so the null text is forwarded as is. The wire representation does not change: `ChatCompletionMessage` is annotated with `@JsonInclude(NON_NULL)`, so an absent content is left out of the request, and `MistralAiChatModel` and `OllamaChatModel` already pass the assistant content through unchecked. See spring-projects#5568 Signed-off-by: subhash polisetti <subhashr161347@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
createRequestrejects an assistant message with a null text, but that is the normalshape of a tool call turn: DeepSeek sends no content alongside
tool_calls, andChatCompletionMessage#contentis documented as null in that case. Replaying such amessage after tool execution, as
ToolCallingManagerdoes, fails withIllegalStateException: text must not be null, so a blocking tool calling conversationcannot complete. Streaming is unaffected because
MessageAggregatorrebuilds thecontent from a
StringBuilder.Dropping the assertion forwards the null unchanged.
ChatCompletionMessageis@JsonInclude(NON_NULL), so content is omitted from the request rather than sent asnull, which is what
MistralAiChatModelandOllamaChatModelalready do. #5568 fixedthe same regression on the response side.
Both tests are offline: one builds the request directly, the other drives a tool call
round trip against a mocked
DeepSeekApi.See #5568