139 lines
4.3 KiB
Rust
139 lines
4.3 KiB
Rust
use crate::api::{self, state::SharedState};
|
|
|
|
use axum::{
|
|
Json,
|
|
extract::{Path, State},
|
|
};
|
|
|
|
#[utoipa::path(
|
|
get,
|
|
path = "/models",
|
|
tag = "models",
|
|
responses(
|
|
(
|
|
status = 200,
|
|
description = "List of locally available Ollama models",
|
|
body = api::types::ApiModelsResponse,
|
|
content_type = "application/json",
|
|
),
|
|
(
|
|
status = 500,
|
|
description = "Internal server error (Ollama or network failure)",
|
|
body = api::errors::ErrorResponse,
|
|
example = json!({ "error": "connection refused" })
|
|
)
|
|
)
|
|
)]
|
|
// #[axum::debug_handler]
|
|
pub async fn list_models(
|
|
State(state): State<SharedState>,
|
|
) -> Result<Json<api::types::ApiModelsResponse>, api::errors::ApiError> {
|
|
let models = state.chat_service.list_models().await?;
|
|
|
|
Ok(Json(models.into()))
|
|
}
|
|
|
|
#[utoipa::path(
|
|
post,
|
|
path = "/models/{model}/load",
|
|
tag = "models",
|
|
params(
|
|
("model" = String, Path, description = "Name of the model to load into memory (e.g. 'llama3')")
|
|
),
|
|
request_body(
|
|
content = api::types::ApiLoadModelRequest,
|
|
description = "Load model request",
|
|
content_type = "application/json",
|
|
example = json!({ "keep_alive": "10m" })
|
|
),
|
|
responses(
|
|
(
|
|
status = 200,
|
|
description = "Model successfully loaded into memory",
|
|
body = api::types::ApiLoadModelResponse,
|
|
content_type = "application/json",
|
|
),
|
|
(
|
|
status = 400,
|
|
description = "Invalid or missing keep_alive format",
|
|
body = api::errors::ErrorResponse,
|
|
examples(
|
|
("Missing" = (value = json!({ "error": "keep alive is required and cannot be empty" }))),
|
|
("Invalid" = (value = json!({ "error": "invalid keep_alive '10x' — use 30s / 10m / 2h, a plain integer, or -1" })))
|
|
)
|
|
),
|
|
(
|
|
status = 404,
|
|
description = "Model not found locally",
|
|
body = api::errors::ErrorResponse,
|
|
example = json!({ "error": "model 'llama3' not found — run `ollama pull llama3`" })
|
|
),
|
|
(
|
|
status = 500,
|
|
description = "Internal server error (Ollama or network failure)",
|
|
body = api::errors::ErrorResponse,
|
|
example = json!({ "error": "connection refused" })
|
|
)
|
|
)
|
|
)]
|
|
pub async fn load_model(
|
|
State(state): State<SharedState>,
|
|
Path(model): Path<String>,
|
|
Json(body): Json<api::types::ApiLoadModelRequest>,
|
|
) -> Result<Json<api::types::ApiLoadModelResponse>, api::errors::ApiError> {
|
|
let response = state
|
|
.chat_service
|
|
.load_model(crate::core::llm::models::LoadModelRequest {
|
|
model,
|
|
keep_alive: body.keep_alive.clone(),
|
|
})
|
|
.await?;
|
|
|
|
Ok(Json(api::types::ApiLoadModelResponse {
|
|
model: response.model,
|
|
keep_alive: body.keep_alive,
|
|
status: "loaded".to_string(),
|
|
}))
|
|
}
|
|
|
|
// #[utoipa::path(
|
|
// delete,
|
|
// path = "/models/{model}/load",
|
|
// tag = "models",
|
|
// params(
|
|
// ("model" = String, Path, description = "Name of the model to unload from memory (e.g. 'llama3')")
|
|
// ),
|
|
// responses(
|
|
// (
|
|
// status = 200,
|
|
// description = "Model successfully unloaded from memory",
|
|
// body = api::types::UnloadModelResponse,
|
|
// content_type = "application/json",
|
|
// ),
|
|
// (
|
|
// status = 404,
|
|
// description = "Model not found locally",
|
|
// body = api::errors::ErrorResponse,
|
|
// example = json!({ "error": "model 'llama3' not found — run `ollama pull llama3`" })
|
|
// ),
|
|
// (
|
|
// status = 500,
|
|
// description = "Internal server error (Ollama or network failure)",
|
|
// body = api::errors::ErrorResponse,
|
|
// example = json!({ "error": "connection refused" })
|
|
// )
|
|
// )
|
|
// )]
|
|
// pub async fn unload_model(
|
|
// State(state): State<AppState>,
|
|
// Path(model): Path<String>,
|
|
// ) -> Result<Json<api::types::UnloadModelResponse>, (axum::http::StatusCode, String)> {
|
|
// let response = state
|
|
// .ollama
|
|
// .unload_model(&model)
|
|
// .await
|
|
// .map_err(into_http_response)?;
|
|
|
|
// Ok(Json(response))
|
|
// }
|