Package-level declarations
Types
Assistant-role message — a model response, whether produced by the API or replayed as context.
Per-currency balance breakdown returned inside a UserBalance.
One completion alternative returned inside a ChatCompletion.
One choice slice carried by a streamed ChatCompletionChunk.
Represents the Chat Completion.
One chunk of a streamed chat completion response.
A chat completion delta generated by streamed model responses.
Forces the model to call a specific tool.
Parameters for configuring chat completion requests to DeepSeek models.
Represents the Chat Completion request
Coarse tool-selection strategy.
One turn in a chat conversation.
Breakdown of how the completion_tokens count in Usage was spent.
One element of a UserMessage's multimodal content array.
Currency used to denominate a BalanceInfo entry.
Shared sampling parameters for the DeepSeek chat and FIM endpoints.
Outcome of removing a file from the Files API.
A file stored by the DeepSeek Files API.
What an uploaded file may be used for, serialized as the raw value the API expects.
One completion alternative returned inside a FIMCompletion.
Parameters for configuring Fill-in-the-Middle (FIM) completion requests to DeepSeek models.
Represents the FIM (Fill-In-the-Middle) Completion request
Reason the model stopped producing tokens for a ChatChoice or FIMChoice.
A partial function call carried inside a ToolCallDelta.
Function signature declared in a chat completion request.
Function invocation emitted by the model inside a ToolCall.
Values accepted by ImageUrl.detail.
Location of an image referenced by an ImageUrlPart.
Image passed to the model by URL, either remote or inline.
Response payload of org.oremif.deepseek.api.models — a list of DeepSeek models available to the current API key.
Log-probability information attached to a ChatChoice or ChatChoiceChunk.
Metadata for a single DeepSeek model entry returned by org.oremif.deepseek.api.models.
Discriminator value returned in the object field of a ModelInfo entry.
Breakdown of the prompt tokens counted by Usage, decoded from the OpenAI-compatible prompt_tokens_details key.
How much reasoning a thinking model spends before it answers, serialized as the raw value the API expects.
Format the model must use for its response.
Custom stop sequences that cause the model to stop generating further tokens.
Represents options that can be applied to a streaming request.
System-role message — sets the assistant's behavior for the rest of the conversation.
Text fragment of a multimodal user message.
Switches a model between its thinking and non-thinking mode.
Values accepted by Thinking.type.
Declaration of a tool the model may call during a chat completion.
A tool invocation emitted by the model inside an AssistantMessage.
An incremental tool call update sent in a streaming chat completion response.
Type discriminator for a tool that the model can call.
Controls how the model picks a tool to call.
A function referenced by a tool.
Tool-role message — the result of executing a ToolCall produced by the model.
One of the top alternatives considered at a given token position in LogProb.topLogprobs.
Token usage statistics for a single chat or FIM request.
Snapshot of the current user's account balance, as returned by org.oremif.deepseek.api.userBalance.
The two shapes the content of a UserMessage takes on the wire: a bare string, or an array of ContentParts.
User-role message — the end-user's input to the conversation.
Functions
Parses FunctionResponse.arguments into a JsonObject, or returns null when the model produced no arguments, malformed JSON, or a JSON value that is not an object.
Creates chat completion parameters using a builder pattern.
Creates streaming chat completion parameters using a builder pattern.
Creates FIM completion parameters using a builder pattern.
Creates streaming FIM completion parameters using a builder pattern.