use futures::Stream;
use std::collections::HashMap;
use std::pin::Pin;
use std::sync::Arc;
use tracing::debug;
use super::config::PredibaseConfig;
use super::model_info::{get_available_models, get_model_info};
use crate::core::providers::base::{GlobalPoolManager, HttpErrorMapper, HttpMethod, header};
use crate::core::providers::unified_provider::ProviderError;
use crate::core::traits::error_mapper::trait_def::ErrorMapper;
use crate::core::traits::{
provider::ProviderConfig as _, provider::llm_provider::trait_definition::LLMProvider,
};
use crate::core::types::{
chat::ChatRequest,
context::RequestContext,
embedding::EmbeddingRequest,
health::HealthStatus,
model::ModelInfo,
model::ProviderCapability,
responses::{ChatChunk, ChatResponse, EmbeddingResponse},
};
const PROVIDER_NAME: &str = "predibase";
const PREDIBASE_CAPABILITIES: &[ProviderCapability] = &[
ProviderCapability::ChatCompletion,
ProviderCapability::ChatCompletionStream,
];
#[derive(Debug, Clone)]
pub struct PredibaseProvider {
config: PredibaseConfig,
pool_manager: Arc<GlobalPoolManager>,
models: Vec<ModelInfo>,
}
impl PredibaseProvider {
pub async fn new(config: PredibaseConfig) -> Result<Self, ProviderError> {
config
.validate()
.map_err(|e| ProviderError::configuration(PROVIDER_NAME, e))?;
let pool_manager = Arc::new(GlobalPoolManager::new().map_err(|e| {
ProviderError::configuration(
PROVIDER_NAME,
format!("Failed to create pool manager: {}", e),
)
})?);
let models = get_available_models()
.iter()
.filter_map(|id| get_model_info(id))
.map(|info| ModelInfo {
id: info.model_id.to_string(),
name: info.display_name.to_string(),
provider: PROVIDER_NAME.to_string(),
max_context_length: info.max_context_length,
max_output_length: Some(info.max_output_length),
supports_streaming: true,
supports_tools: info.supports_tools,
supports_multimodal: info.supports_multimodal,
input_cost_per_1k_tokens: Some(info.input_cost_per_million / 1000.0),
output_cost_per_1k_tokens: Some(info.output_cost_per_million / 1000.0),
currency: "USD".to_string(),
capabilities: vec![
ProviderCapability::ChatCompletion,
ProviderCapability::ChatCompletionStream,
],
created_at: None,
updated_at: None,
metadata: HashMap::new(),
})
.collect();
Ok(Self {
config,
pool_manager,
models,
})
}
pub async fn with_api_key(api_key: impl Into<String>) -> Result<Self, ProviderError> {
let config = PredibaseConfig {
api_key: Some(api_key.into()),
..Default::default()
};
Self::new(config).await
}
async fn execute_request(
&self,
endpoint: &str,
body: serde_json::Value,
) -> Result<serde_json::Value, ProviderError> {
let url = format!("{}{}", self.config.get_api_base(), endpoint);
let mut headers = Vec::with_capacity(2);
if let Some(api_key) = &self.config.get_api_key() {
headers.push(header("Authorization", format!("Bearer {}", api_key)));
}
headers.push(header("Content-Type", "application/json".to_string()));
let response = self
.pool_manager
.execute_request(&url, HttpMethod::POST, headers, Some(body))
.await
.map_err(|e| ProviderError::network(PROVIDER_NAME, e.to_string()))?;
let status = response.status();
let response_bytes = response
.bytes()
.await
.map_err(|e| ProviderError::network(PROVIDER_NAME, e.to_string()))?;
if !status.is_success() {
let error_text = String::from_utf8_lossy(&response_bytes);
return Err(Self::map_http_error(status.as_u16(), &error_text));
}
serde_json::from_slice(&response_bytes).map_err(|e| {
ProviderError::api_error(
PROVIDER_NAME,
500,
format!("Failed to parse response: {}", e),
)
})
}
fn map_http_error(status: u16, body: &str) -> ProviderError {
match status {
400 => ProviderError::invalid_request(PROVIDER_NAME, body.to_string()),
401 => ProviderError::authentication(PROVIDER_NAME, "Invalid API key"),
403 => ProviderError::authentication(PROVIDER_NAME, "Access forbidden"),
404 => ProviderError::model_not_found(PROVIDER_NAME, "Model not found"),
429 => ProviderError::rate_limit(PROVIDER_NAME, None),
500 => HttpErrorMapper::map_status_code(PROVIDER_NAME, 500, "Internal server error"),
503 => ProviderError::provider_unavailable(PROVIDER_NAME, "Service unavailable"),
_ => HttpErrorMapper::map_status_code(PROVIDER_NAME, status, body),
}
}
}
impl LLMProvider for PredibaseProvider {
fn name(&self) -> &'static str {
PROVIDER_NAME
}
fn capabilities(&self) -> &'static [ProviderCapability] {
PREDIBASE_CAPABILITIES
}
fn models(&self) -> &[ModelInfo] {
&self.models
}
fn get_supported_openai_params(&self, _model: &str) -> &'static [&'static str] {
&[
"temperature",
"top_p",
"max_tokens",
"stream",
"stop",
"frequency_penalty",
"presence_penalty",
]
}
async fn map_openai_params(
&self,
params: HashMap<String, serde_json::Value>,
_model: &str,
) -> Result<HashMap<String, serde_json::Value>, ProviderError> {
Ok(params)
}
async fn transform_request(
&self,
request: ChatRequest,
_context: RequestContext,
) -> Result<serde_json::Value, ProviderError> {
serde_json::to_value(&request)
.map_err(|e| ProviderError::invalid_request(PROVIDER_NAME, e.to_string()))
}
async fn transform_response(
&self,
raw_response: &[u8],
_model: &str,
_request_id: &str,
) -> Result<ChatResponse, ProviderError> {
serde_json::from_slice(raw_response).map_err(|e| {
ProviderError::api_error(
PROVIDER_NAME,
500,
format!("Failed to parse response: {}", e),
)
})
}
fn get_error_mapper(&self) -> Box<dyn ErrorMapper<ProviderError>> {
Box::new(crate::core::traits::error_mapper::DefaultErrorMapper)
}
async fn chat_completion(
&self,
request: ChatRequest,
_context: RequestContext,
) -> Result<ChatResponse, ProviderError> {
debug!("Predibase chat request: model={}", request.model);
let request_json = serde_json::to_value(&request)
.map_err(|e| ProviderError::invalid_request(PROVIDER_NAME, e.to_string()))?;
let response = self
.execute_request("/v1/chat/completions", request_json)
.await?;
serde_json::from_value(response).map_err(|e| {
ProviderError::api_error(
PROVIDER_NAME,
500,
format!("Failed to parse chat response: {}", e),
)
})
}
async fn chat_completion_stream(
&self,
_request: ChatRequest,
_context: RequestContext,
) -> Result<Pin<Box<dyn Stream<Item = Result<ChatChunk, ProviderError>> + Send>>, ProviderError>
{
Err(ProviderError::not_implemented(
PROVIDER_NAME,
"Streaming not yet implemented",
))
}
async fn embeddings(
&self,
_request: EmbeddingRequest,
_context: RequestContext,
) -> Result<EmbeddingResponse, ProviderError> {
Err(ProviderError::not_supported(PROVIDER_NAME, "Embeddings"))
}
async fn health_check(&self) -> HealthStatus {
if self.config.get_api_key().is_some() {
HealthStatus::Healthy
} else {
HealthStatus::Unhealthy
}
}
async fn calculate_cost(
&self,
model: &str,
input_tokens: u32,
output_tokens: u32,
) -> Result<f64, ProviderError> {
let model_info = get_model_info(model).ok_or_else(|| {
ProviderError::model_not_found(PROVIDER_NAME, format!("Unknown model: {}", model))
})?;
let input_cost = (input_tokens as f64) * (model_info.input_cost_per_million / 1_000_000.0);
let output_cost =
(output_tokens as f64) * (model_info.output_cost_per_million / 1_000_000.0);
Ok(input_cost + output_cost)
}
}