AI Inference

Package: com.mimik.mimoe.inference

OpenAI-compatible chat completions, streaming, and embeddings.

ChatCompletion

@Serializable
data class ChatCompletion(
    val id: String,
    val objectType: String = "chat.completion",
    val created: Long = 0,
    val model: String = "",
    val choices: List<ChatCompletion.Choice> = emptyList(),
    val usage: ChatCompletion.Usage? = null
)

Response from a chat completion request.

Constructors

constructor(
    id: String,
    objectType: String = "chat.completion",
    created: Long = 0,
    model: String = "",
    choices: List<ChatCompletion.Choice> = emptyList(),
    usage: ChatCompletion.Usage? = null
)

Types

Name	Summary
Choice	@Serializable data class Choice(val index: Int = 0, val message: ChatMessage, val finishReason: String? = null)
Usage	@Serializable data class Usage(val promptTokens: Int = 0, val completionTokens: Int = 0, val totalTokens: Int = 0)

Properties

Name	Summary
choices	val choices: List<ChatCompletion.Choice>
created	val created: Long = 0
id	val id: String
model	val model: String
objectType	@SerialName(value = "object") val objectType: String
usage	val usage: ChatCompletion.Usage? = null

---

ChatCompletionChunk

@Serializable
data class ChatCompletionChunk(
    val id: String,
    val objectType: String = "chat.completion.chunk",
    val created: Long = 0,
    val model: String = "",
    val choices: List<ChatCompletionChunk.ChunkChoice> = emptyList()
)

A streaming chunk from a chat completion.

Constructors

constructor(
    id: String,
    objectType: String = "chat.completion.chunk",
    created: Long = 0,
    model: String = "",
    choices: List<ChatCompletionChunk.ChunkChoice> = emptyList()
)

Types

Name	Summary
ChunkChoice	@Serializable data class ChunkChoice(val index: Int = 0, val delta: ChatCompletionChunk.Delta, val finishReason: String? = null)
Delta	@Serializable data class Delta(val role: String? = null, val content: String? = null)

Properties

Name	Summary
choices	val choices: List<ChatCompletionChunk.ChunkChoice>
created	val created: Long = 0
id	val id: String
model	val model: String
objectType	@SerialName(value = "object") val objectType: String

---

ChatMessage

@Serializable
data class ChatMessage(val role: String, val content: String)

A message in a chat conversation.

Constructors

constructor(role: String, content: String)

Properties

Name	Summary
content	val content: String Message content.
role	val role: String Role of the message sender (system, user, assistant).

---

EmbeddingData

@Serializable
data class EmbeddingData(
    val objectType: String = "embedding",
    val embedding: List<Float>,
    val index: Int = 0
)

Constructors

constructor(
    objectType: String = "embedding",
    embedding: List<Float>,
    index: Int = 0
)

Properties

Name	Summary
embedding	val embedding: List<Float>
index	val index: Int = 0
objectType	@SerialName(value = "object") val objectType: String

---

EmbeddingResponse

@Serializable
data class EmbeddingResponse(
    val objectType: String = "list",
    val data: List<EmbeddingData> = emptyList(),
    val model: String = "",
    val usage: EmbeddingUsage? = null
)

Response from an embeddings request.

Constructors

constructor(
    objectType: String = "list",
    data: List<EmbeddingData> = emptyList(),
    model: String = "",
    usage: EmbeddingUsage? = null
)

Properties

Name	Summary
data	val data: List<EmbeddingData>
model	val model: String
objectType	@SerialName(value = "object") val objectType: String
usage	val usage: EmbeddingUsage? = null

---

EmbeddingUsage

@Serializable
data class EmbeddingUsage(
    val promptTokens: Int = 0,
    val totalTokens: Int = 0
)

Constructors

constructor(promptTokens: Int = 0, totalTokens: Int = 0)

Properties

Name	Summary
promptTokens	@SerialName(value = "prompt_tokens") val promptTokens: Int = 0
totalTokens	@SerialName(value = "total_tokens") val totalTokens: Int = 0

---

InferenceClient

class InferenceClient

Inference sub-client.

Provides OpenAI-compatible chat completions, streaming, embeddings, and model cache management. Base path: /mimik-ai/openai/v1 Requires the ai-foundation addon.

Access via client.inference.

Functions

chatCompletion

suspend fun chatCompletion(model: String, messages: List<ChatMessage>, temperature: Double? = null, topP: Double? = null, maxTokens: Int? = null): Result<ChatCompletion>

Generate a chat completion.

chatCompletionStream

fun chatCompletionStream(model: String, messages: List<ChatMessage>, temperature: Double? = null, topP: Double? = null, maxTokens: Int? = null): Flow<ChatCompletionChunk>

Stream a chat completion via SSE.

createEmbeddings

suspend fun createEmbeddings(model: String, input: List<String>): Result<EmbeddingResponse>

Generate text embeddings.

listModels

suspend fun listModels(): Result<List<LoadedModel>>

List models currently loaded in the runtime cache.

loadModel

suspend fun loadModel(model: String, chatTemplateHint: String? = null, contextSize: Int? = null, gpuLayerSize: Int? = null): Result<LoadedModel>

Load a model from the Model Registry into runtime cache.

unloadModel

suspend fun unloadModel(modelId: String): Result<Unit>

Unload a model from runtime cache.

---

LoadedModel

@Serializable
data class LoadedModel(
    val id: String,
    val objectType: String = "model",
    val created: Long = 0,
    val ownedBy: String = "",
    val info: ModelInfoDetail? = null,
    val metrics: ModelMetrics? = null
)

A model loaded in the inference runtime cache.

Returned by GET /mimik-ai/openai/v1/models and POST /mimik-ai/openai/v1/models.

Constructors

constructor(
    id: String,
    objectType: String = "model",
    created: Long = 0,
    ownedBy: String = "",
    info: ModelInfoDetail? = null,
    metrics: ModelMetrics? = null
)

Properties

Name	Summary
created	val created: Long = 0 Creation timestamp (Unix epoch seconds).
id	val id: String Model identifier.
info	val info: ModelInfoDetail? = null Model metadata.
metrics	val metrics: ModelMetrics? = null Runtime performance metrics.
objectType	@SerialName(value = "object") val objectType: String Object type (always "model").
ownedBy	@SerialName(value = "owned_by") val ownedBy: String Model owner.

---

ModelInfoDetail

@Serializable
data class ModelInfoDetail(
    val kind: String = "",
    val chatTemplateHint: String = "",
    val nGpuLayers: Int = 0,
    val maxContext: Int = 0,
    val nVocab: Int = 0,
    val nCtxTrain: Int = 0,
    val nEmbd: Int = 0,
    val nParams: Long = 0,
    val modelSize: Long = 0
)

Detailed model metadata.

Constructors

constructor(
    kind: String = "",
    chatTemplateHint: String = "",
    nGpuLayers: Int = 0,
    maxContext: Int = 0,
    nVocab: Int = 0,
    nCtxTrain: Int = 0,
    nEmbd: Int = 0,
    nParams: Long = 0,
    modelSize: Long = 0
)

Properties

Name	Summary
chatTemplateHint	@SerialName(value = "chat_template_hint") val chatTemplateHint: String Applied chat template.
kind	val kind: String Model kind: "llm", "vlm", or "embed".
maxContext	@SerialName(value = "max_context") val maxContext: Int = 0 Maximum context size.
modelSize	@SerialName(value = "model_size") val modelSize: Long = 0 Model file size in bytes.
nCtxTrain	@SerialName(value = "n_ctx_train") val nCtxTrain: Int = 0 Training context length.
nEmbd	@SerialName(value = "n_embd") val nEmbd: Int = 0 Embedding dimension.
nGpuLayers	@SerialName(value = "n_gpu_layers") val nGpuLayers: Int = 0 Number of GPU-offloaded layers.
nParams	@SerialName(value = "n_params") val nParams: Long = 0 Total parameter count.
nVocab	@SerialName(value = "n_vocab") val nVocab: Int = 0 Vocabulary size.

---

ModelMetrics

@Serializable
data class ModelMetrics(
    val inferenceCount: Int = 0,
    val lastUsed: Long = 0,
    val loadedAt: Long = 0,
    val tokensPerSecond: Double = 0.0,
    val avgTokensPerSecond: Double = 0.0,
    val lastLatencyMs: Double = 0.0,
    val avgLatencyMs: Double = 0.0
)

Runtime performance metrics for a loaded model.

Constructors

constructor(
    inferenceCount: Int = 0,
    lastUsed: Long = 0,
    loadedAt: Long = 0,
    tokensPerSecond: Double = 0.0,
    avgTokensPerSecond: Double = 0.0,
    lastLatencyMs: Double = 0.0,
    avgLatencyMs: Double = 0.0
)

Properties

Name	Summary
avgLatencyMs	@SerialName(value = "avg_latency_ms") val avgLatencyMs: Double = 0.0 Average latency (ms). Embed only.
avgTokensPerSecond	@SerialName(value = "avg_tokens_per_second") val avgTokensPerSecond: Double = 0.0 Average throughput (tokens/sec). LLM/VLM only.
inferenceCount	@SerialName(value = "inference_count") val inferenceCount: Int = 0 Total inference calls.
lastLatencyMs	@SerialName(value = "last_latency_ms") val lastLatencyMs: Double = 0.0 Most recent latency (ms). Embed only.
lastUsed	@SerialName(value = "last_used") val lastUsed: Long = 0 Most recent usage (Unix epoch seconds).
loadedAt	@SerialName(value = "loaded_at") val loadedAt: Long = 0 When the model was loaded (Unix epoch seconds).
tokensPerSecond	@SerialName(value = "tokens_per_second") val tokensPerSecond: Double = 0.0 Most recent throughput (tokens/sec). LLM/VLM only.

---