reafctor: all code without stream

This commit is contained in:
2026-06-02 18:30:39 +02:00
parent 974be437af
commit 28352d8bdb
90 changed files with 3581 additions and 2434 deletions
+293
View File
@@ -0,0 +1,293 @@
use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use utoipa::ToSchema;
use uuid::Uuid;
// ------ Models ------
#[derive(Debug, Serialize, ToSchema)]
pub struct ModelInfo {
pub name: String,
pub family: Option<String>,
pub parameter_size: Option<String>,
pub metadata: ModelMetadata,
}
#[derive(Debug, Clone, Default, Serialize, ToSchema)]
pub struct ModelMetadata {
pub extra: HashMap<String, String>,
}
// ------ Endpoint: /models ------
#[derive(Debug, Serialize, ToSchema)]
pub struct ModelsResponse {
pub models: Vec<ModelInfo>,
}
// ------ Endpoint: /models/{model}/load ------
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct LoadModelResponse {
pub model: String,
pub status: String,
pub keep_alive: String,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct LoadModelRequest {
pub keep_alive: String,
}
// ------ Endpoint: /models/{model}/unload ------
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct UnloadModelResponse {
pub model: String,
pub status: String,
}
// ------ Completions ------
#[derive(Debug, Deserialize, Serialize, ToSchema, Default)]
pub struct LLMOptions {
#[serde(default)]
pub stream: bool,
pub temperature: Option<f32>,
pub top_p: Option<f32>,
pub top_k: Option<u32>,
pub repeat_penalty: Option<f32>,
pub seed: Option<i64>,
pub num_ctx: Option<u32>,
pub num_predict: Option<u32>,
pub stop: Option<Vec<String>>,
pub keep_alive: Option<String>,
pub context_depth: Option<u32>,
}
// ------ Endpoint: /completions ------
#[derive(Debug, Deserialize, Serialize, ToSchema)]
pub struct CompletionRequest {
#[serde(flatten)]
pub options: LLMOptions,
pub model: String,
pub prompt: String,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct CompletionResponse {
pub id: Uuid,
pub object: CompletionObject,
pub created: String,
pub model: String,
pub choices: Vec<Choice>,
pub usage: Usage,
}
#[derive(Debug, Serialize, Deserialize, ToSchema, PartialEq, Eq)]
#[serde(rename_all = "snake_case")]
pub enum CompletionObject {
TextCompletion,
}
// ------ Endpoint: /chat/completions ------
#[derive(Debug, Serialize, Deserialize, ToSchema, PartialEq, Eq)]
#[serde(rename_all = "snake_case")]
pub enum FinishReason {
Stop,
Length,
ContentFilter,
ToolCalls,
Error,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct Choice {
pub text: String,
pub index: u32,
pub finish_reason: FinishReason,
}
#[derive(Debug, Serialize, Deserialize, ToSchema, Clone, Copy)]
pub struct Usage {
pub prompt_tokens: u32,
pub completion_tokens: u32,
pub total_tokens: u32,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct CompletionChunk {
pub id: String,
pub object: String,
pub choices: Vec<Choice>,
}
#[derive(Debug, Deserialize, Serialize, ToSchema)]
pub struct ChatRequest {
#[serde(flatten)]
pub base: LLMOptions,
pub model: String,
pub message: Message,
// Non standard Open AI
pub conversation_id: Option<Uuid>,
pub parent_id: Option<Uuid>, // used when branching, regenerate, etc
}
#[derive(Debug, Deserialize, Serialize, ToSchema)]
pub struct Message {
pub role: Role,
pub content: String,
}
#[derive(Debug, Deserialize, Serialize, ToSchema, PartialEq, Eq)]
#[serde(rename_all = "lowercase")]
pub enum Role {
System,
User,
Assistant,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct ChatResponse {
pub id: String,
pub object: String,
pub created: u64,
pub model: String,
pub choices: Vec<ChatChoice>,
pub usage: Usage,
pub conversation_id: Uuid,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct ChatChoice {
pub index: u32,
pub message: Message,
pub finish_reason: FinishReason,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct ChatCompletionChunk {
pub id: String,
pub object: String,
pub choices: Vec<ChatChunkChoice>,
pub usage: Option<Usage>,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct ChatChunkChoice {
pub index: u32,
pub delta: Delta,
pub finish_reason: Option<FinishReason>,
}
#[derive(Debug, Serialize, Deserialize, ToSchema)]
pub struct Delta {
pub content: Option<String>,
pub role: Option<Role>,
}
// #[derive(Debug, Serialize, Deserialize, ToSchema)]
// pub struct StartEventData {
// pub conversation_id: Uuid,
// pub created: u64,
// pub id: Uuid,
// }
// #[derive(Debug, Serialize, Deserialize, ToSchema)]
// pub struct EndEventData {
// pub created: u64,
// pub id: Uuid,
// pub usage: Usage,
// }
// #[derive(Debug, Serialize, Deserialize, ToSchema)]
// #[serde(tag = "type", content = "data")]
// pub enum StreamEvent {
// #[serde(rename = "start")]
// Start(StartEventData),
// #[serde(rename = "end")]
// End(EndEventData),
// #[serde(rename = "delta")]
// Delta(ChatCompletionChunk),
// }
// ------ Api Key ------
#[derive(Debug, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum ApiKeyScope {
User,
Admin,
}
#[derive(serde::Deserialize)]
pub struct CreateApiKeyRequest {
pub name: String,
pub scopes: Vec<ApiKeyScope>,
}
#[derive(serde::Serialize)]
pub struct CreateApiKeyResponse {
pub api_key: String,
}
// ------ Fetch database ------
// --- Shared ---
#[derive(Debug, Deserialize)]
pub struct CursorPage {
pub limit: Option<u32>,
pub before: Option<chrono::DateTime<chrono::Utc>>,
}
// --- Conversation ---
#[derive(Debug, Serialize)]
pub struct ConversationSummary {
pub id: Uuid,
pub title: String,
pub created_at: chrono::DateTime<chrono::Utc>,
pub updated_at: chrono::DateTime<chrono::Utc>,
}
#[derive(Debug, Serialize)]
pub struct ConversationListResponse {
pub conversations: Vec<ConversationSummary>,
pub has_more: bool,
}
// --- Message ---
#[derive(Debug, Serialize)]
pub enum ApiChatRole {
System,
User,
Assistant,
}
#[derive(Debug, Serialize)]
pub struct MessageSummary {
pub id: Uuid,
pub parent_id: Option<Uuid>,
pub role: ApiChatRole,
pub content: String,
pub created_at: chrono::DateTime<chrono::Utc>,
pub tokens: Option<i32>,
}
#[derive(Debug, Serialize)]
pub struct MessageListResponse {
pub messages: Vec<MessageSummary>,
pub has_more: bool,
}