use std::time::Duration;
use reqwest::Client;
use reqwest::header::{AUTHORIZATION, CONTENT_TYPE, HeaderMap, HeaderValue};
use serde::{Deserialize, Serialize};
use super::error_classification::{
classify_http_error, map_backend_error, network_error, parse_error,
};
use super::format_contract::finalize_chat_response;
use crate::secret::{EnvSecretProvider, SecretProvider, SecretString};
use converge_core::backend::{BackendError, BackendResult};
use converge_provider_api::{
BoxFuture, ChatBackend, ChatRequest, ChatResponse, ChatRole, FinishReason as ChatFinishReason,
LlmError as ChatLlmError, ResponseFormat, TokenUsage as ChatTokenUsage,
ToolCall as ChatToolCall,
};
pub struct OpenAiBackend {
api_key: SecretString,
model: String,
base_url: String,
client: Client,
temperature: f32,
max_retries: usize,
}
impl OpenAiBackend {
#[must_use]
pub fn new(api_key: impl Into<String>) -> Self {
Self {
api_key: SecretString::new(api_key),
model: "gpt-4o".to_string(),
base_url: "https://api.openai.com".to_string(),
client: Client::new(),
temperature: 0.0,
max_retries: 3,
}
}
pub fn from_env() -> BackendResult<Self> {
Self::from_secret_provider(&EnvSecretProvider)
}
pub fn from_secret_provider(secrets: &dyn SecretProvider) -> BackendResult<Self> {
let api_key =
secrets
.get_secret("OPENAI_API_KEY")
.map_err(|e| BackendError::Unavailable {
message: format!("OPENAI_API_KEY: {e}"),
})?;
Ok(Self {
api_key,
model: "gpt-4o".to_string(),
base_url: "https://api.openai.com".to_string(),
client: Client::new(),
temperature: 0.0,
max_retries: 3,
})
}
#[must_use]
pub fn with_model(mut self, model: impl Into<String>) -> Self {
self.model = model.into();
self
}
#[must_use]
pub fn with_base_url(mut self, url: impl Into<String>) -> Self {
self.base_url = url.into();
self
}
#[must_use]
pub fn with_temperature(mut self, temp: f32) -> Self {
self.temperature = temp;
self
}
#[must_use]
pub fn with_max_retries(mut self, retries: usize) -> Self {
self.max_retries = retries;
self
}
fn build_headers(&self) -> BackendResult<HeaderMap> {
let mut headers = HeaderMap::new();
headers.insert(CONTENT_TYPE, HeaderValue::from_static("application/json"));
let auth_value = format!("Bearer {}", self.api_key.expose());
headers.insert(
AUTHORIZATION,
HeaderValue::from_str(&auth_value).map_err(|e| BackendError::InvalidRequest {
message: format!("Invalid API key: {e}"),
})?,
);
Ok(headers)
}
fn build_request(&self, req: &ChatRequest) -> OpenAiRequest {
let model = req.model.clone().unwrap_or_else(|| self.model.clone());
let temperature = req.temperature.unwrap_or(self.temperature);
let max_tokens = req.max_tokens.map(|t| t as usize).unwrap_or(4096);
let mut messages = Vec::new();
let system_content = if let Some(instruction) = req.response_format.system_instruction() {
let base = req.system.clone().unwrap_or_default();
Some(format!("{base}\n\n{instruction}"))
} else {
req.system.clone()
};
if let Some(system) = &system_content {
messages.push(OpenAiMessage {
role: "system".to_string(),
content: Some(system.clone()),
tool_calls: None,
tool_call_id: None,
});
}
for msg in &req.messages {
let role = match msg.role {
ChatRole::System => "system",
ChatRole::User => "user",
ChatRole::Assistant => "assistant",
ChatRole::Tool => "tool",
};
let tool_calls = if msg.tool_calls.is_empty() {
None
} else {
Some(
msg.tool_calls
.iter()
.map(|tool_call| OpenAiResponseToolCall {
id: tool_call.id.clone(),
function: OpenAiResponseFunction {
name: tool_call.name.clone(),
arguments: tool_call.arguments.clone(),
},
})
.collect(),
)
};
let content = if msg.content.is_empty() && tool_calls.is_some() {
None
} else {
Some(msg.content.clone())
};
messages.push(OpenAiMessage {
role: role.to_string(),
content,
tool_calls,
tool_call_id: msg.tool_call_id.clone(),
});
}
let tools: Option<Vec<OpenAiTool>> = if req.tools.is_empty() {
None
} else {
Some(
req.tools
.iter()
.map(|t| OpenAiTool {
r#type: "function".to_string(),
function: OpenAiFunction {
name: t.name.clone(),
description: Some(t.description.clone()),
parameters: Some(t.parameters.clone()),
},
})
.collect(),
)
};
let response_format = match req.response_format {
ResponseFormat::Json => Some(serde_json::json!({"type": "json_object"})),
_ => None,
};
let stop = if req.stop_sequences.is_empty() {
None
} else {
Some(req.stop_sequences.clone())
};
OpenAiRequest {
model,
messages,
temperature: Some(temperature),
max_tokens: Some(max_tokens),
tools,
response_format,
stop,
}
}
async fn chat_async(&self, req: ChatRequest) -> Result<ChatResponse, ChatLlmError> {
let openai_req = self.build_request(&req);
let model = req.model.clone().unwrap_or_else(|| self.model.clone());
let response = self.execute_with_retries(&model, &openai_req).await?;
let choice = response.choices.first();
let content = choice
.and_then(|c| c.message.content.clone())
.unwrap_or_default();
let tool_calls = choice
.and_then(|c| c.message.tool_calls.as_ref())
.map(|calls| {
calls
.iter()
.map(|tc| ChatToolCall {
id: tc.id.clone(),
name: tc.function.name.clone(),
arguments: tc.function.arguments.clone(),
})
.collect()
})
.unwrap_or_default();
let finish_reason = choice.and_then(|c| match c.finish_reason.as_deref() {
Some("stop") => Some(ChatFinishReason::Stop),
Some("length") => Some(ChatFinishReason::Length),
Some("tool_calls") => Some(ChatFinishReason::ToolCalls),
Some("content_filter") => Some(ChatFinishReason::ContentFilter),
_ => None,
});
finalize_chat_response(
req.response_format,
ChatResponse {
content,
tool_calls,
usage: response.usage.map(|u| ChatTokenUsage {
prompt_tokens: u.prompt_tokens,
completion_tokens: u.completion_tokens,
total_tokens: u.total_tokens,
}),
model: Some(response.model),
finish_reason,
metadata: Default::default(),
},
)
}
async fn execute_with_retries(
&self,
model: &str,
request: &OpenAiRequest,
) -> Result<OpenAiResponse, ChatLlmError> {
let url = format!("{}/v1/chat/completions", self.base_url);
let headers = self.build_headers().map_err(map_backend_error)?;
let mut last_error = None;
for attempt in 0..=self.max_retries {
if attempt > 0 {
tokio::time::sleep(Duration::from_millis(100 * 2_u64.pow(attempt as u32))).await;
}
let result = self
.client
.post(&url)
.headers(headers.clone())
.json(request)
.send()
.await;
match result {
Ok(response) => {
let status = response.status();
if status.is_success() {
match response.json::<OpenAiResponse>().await {
Ok(parsed) => return Ok(parsed),
Err(e) => {
last_error = Some(parse_error(e));
}
}
} else if status.as_u16() == 429 || status.as_u16() >= 500 {
let body = response.text().await.unwrap_or_default();
last_error = Some(classify_http_error(status.as_u16(), &body, model));
} else {
let body = response.text().await.unwrap_or_default();
return Err(classify_http_error(status.as_u16(), &body, model));
}
}
Err(e) => {
last_error = Some(network_error(e));
}
}
}
Err(last_error.unwrap_or_else(|| ChatLlmError::ProviderError {
message: "unknown error".to_string(),
code: None,
}))
}
}
impl ChatBackend for OpenAiBackend {
type ChatFut<'a>
= BoxFuture<'a, Result<ChatResponse, ChatLlmError>>
where
Self: 'a;
fn chat(&self, req: ChatRequest) -> Self::ChatFut<'_> {
Box::pin(async move { self.chat_async(req).await })
}
}
#[derive(Debug, Serialize)]
struct OpenAiRequest {
model: String,
messages: Vec<OpenAiMessage>,
#[serde(skip_serializing_if = "Option::is_none")]
temperature: Option<f32>,
#[serde(skip_serializing_if = "Option::is_none")]
max_tokens: Option<usize>,
#[serde(skip_serializing_if = "Option::is_none")]
tools: Option<Vec<OpenAiTool>>,
#[serde(skip_serializing_if = "Option::is_none")]
response_format: Option<serde_json::Value>,
#[serde(skip_serializing_if = "Option::is_none")]
stop: Option<Vec<String>>,
}
#[derive(Debug, Serialize)]
struct OpenAiMessage {
role: String,
#[serde(skip_serializing_if = "Option::is_none")]
content: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
tool_calls: Option<Vec<OpenAiResponseToolCall>>,
#[serde(skip_serializing_if = "Option::is_none")]
tool_call_id: Option<String>,
}
#[derive(Debug, Serialize)]
struct OpenAiTool {
r#type: String,
function: OpenAiFunction,
}
#[derive(Debug, Serialize)]
struct OpenAiFunction {
name: String,
#[serde(skip_serializing_if = "Option::is_none")]
description: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
parameters: Option<serde_json::Value>,
}
#[derive(Debug, Deserialize)]
struct OpenAiResponse {
model: String,
choices: Vec<OpenAiChoice>,
usage: Option<OpenAiUsage>,
}
#[derive(Debug, Deserialize)]
struct OpenAiChoice {
message: OpenAiResponseMessage,
finish_reason: Option<String>,
}
#[derive(Debug, Deserialize)]
struct OpenAiResponseMessage {
content: Option<String>,
tool_calls: Option<Vec<OpenAiResponseToolCall>>,
}
#[derive(Debug, Serialize, Deserialize)]
struct OpenAiResponseToolCall {
id: String,
function: OpenAiResponseFunction,
}
#[derive(Debug, Serialize, Deserialize)]
struct OpenAiResponseFunction {
name: String,
arguments: String,
}
#[derive(Debug, Deserialize)]
struct OpenAiUsage {
prompt_tokens: u32,
completion_tokens: u32,
total_tokens: u32,
}
#[cfg(test)]
mod tests {
use super::*;
use converge_core::traits::{
ChatMessage, ChatRequest, ChatRole, ResponseFormat, ToolDefinition,
};
use wiremock::matchers::{method, path};
use wiremock::{Mock, MockServer, ResponseTemplate};
#[test]
fn test_openai_backend_creation() {
let backend = OpenAiBackend::new("test-key")
.with_model("gpt-4o-mini")
.with_temperature(0.5);
assert_eq!(backend.model, "gpt-4o-mini");
assert_eq!(backend.temperature, 0.5);
assert_eq!(backend.api_key.expose(), "test-key");
}
#[test]
fn test_build_request_basic() {
let backend = OpenAiBackend::new("test-key");
let req = ChatRequest {
messages: vec![ChatMessage {
role: ChatRole::User,
content: "Hello".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
}],
system: None,
tools: Vec::new(),
response_format: ResponseFormat::default(),
max_tokens: None,
temperature: None,
stop_sequences: Vec::new(),
model: None,
};
let openai_req = backend.build_request(&req);
assert_eq!(openai_req.model, "gpt-4o");
assert_eq!(openai_req.messages.len(), 1);
assert_eq!(openai_req.messages[0].role, "user");
assert!(openai_req.tools.is_none());
assert!(openai_req.response_format.is_none());
}
#[test]
fn test_build_request_with_system() {
let backend = OpenAiBackend::new("test-key");
let req = ChatRequest {
messages: vec![ChatMessage {
role: ChatRole::User,
content: "Hi".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
}],
system: Some("You are helpful.".to_string()),
tools: Vec::new(),
response_format: ResponseFormat::default(),
max_tokens: None,
temperature: None,
stop_sequences: Vec::new(),
model: None,
};
let openai_req = backend.build_request(&req);
assert_eq!(openai_req.messages.len(), 2);
assert_eq!(openai_req.messages[0].role, "system");
assert_eq!(
openai_req.messages[0].content.as_deref(),
Some("You are helpful.")
);
assert_eq!(openai_req.messages[1].role, "user");
}
#[test]
fn test_build_request_with_tools() {
let backend = OpenAiBackend::new("test-key");
let req = ChatRequest {
messages: vec![ChatMessage {
role: ChatRole::User,
content: "What's the weather?".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
}],
system: None,
tools: vec![ToolDefinition {
name: "get_weather".to_string(),
description: "Get current weather".to_string(),
parameters: serde_json::json!({"type": "object", "properties": {"city": {"type": "string"}}}),
}],
response_format: ResponseFormat::default(),
max_tokens: None,
temperature: None,
stop_sequences: Vec::new(),
model: None,
};
let openai_req = backend.build_request(&req);
let tools = openai_req.tools.unwrap();
assert_eq!(tools.len(), 1);
assert_eq!(tools[0].r#type, "function");
assert_eq!(tools[0].function.name, "get_weather");
}
#[test]
fn test_build_request_json_format() {
let backend = OpenAiBackend::new("test-key");
let req = ChatRequest {
messages: vec![ChatMessage {
role: ChatRole::User,
content: "Return JSON".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
}],
system: None,
tools: Vec::new(),
response_format: ResponseFormat::Json,
max_tokens: None,
temperature: None,
stop_sequences: Vec::new(),
model: None,
};
let openai_req = backend.build_request(&req);
assert_eq!(
openai_req.response_format,
Some(serde_json::json!({"type": "json_object"}))
);
}
#[test]
fn test_build_request_with_stop_sequences() {
let backend = OpenAiBackend::new("test-key");
let req = ChatRequest {
messages: vec![ChatMessage {
role: ChatRole::User,
content: "Go".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
}],
system: None,
tools: Vec::new(),
response_format: ResponseFormat::default(),
max_tokens: None,
temperature: None,
stop_sequences: vec!["STOP".to_string()],
model: None,
};
let openai_req = backend.build_request(&req);
assert_eq!(openai_req.stop, Some(vec!["STOP".to_string()]));
}
#[test]
fn test_chat_runtime_multiturn_and_tool_calls() {
let runtime = tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
.unwrap();
let server = runtime.block_on(MockServer::start());
runtime.block_on(async {
Mock::given(method("POST"))
.and(path("/v1/chat/completions"))
.respond_with(ResponseTemplate::new(200).set_body_json(serde_json::json!({
"id": "chatcmpl_test",
"model": "gpt-4o",
"choices": [{
"message": {
"content": "I'll use a tool.",
"tool_calls": [{
"id": "call_1",
"type": "function",
"function": {
"name": "lookup_weather",
"arguments": "{\"city\":\"Paris\"}"
}
}]
},
"finish_reason": "tool_calls"
}],
"usage": {
"prompt_tokens": 12,
"completion_tokens": 4,
"total_tokens": 16
}
})))
.mount(&server)
.await;
});
let backend = OpenAiBackend::new("test-key").with_base_url(server.uri());
let response = runtime
.block_on(backend.chat(ChatRequest {
messages: vec![
ChatMessage {
role: ChatRole::User,
content: "Weather?".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
},
ChatMessage {
role: ChatRole::Assistant,
content: "Let me check.".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
},
],
system: Some("You are helpful.".to_string()),
tools: vec![ToolDefinition {
name: "lookup_weather".to_string(),
description: "Lookup weather".to_string(),
parameters: serde_json::json!({
"type": "object",
"properties": {"city": {"type": "string"}}
}),
}],
response_format: ResponseFormat::Json,
max_tokens: Some(64),
temperature: Some(0.0),
stop_sequences: vec!["DONE".to_string()],
model: None,
}))
.unwrap();
assert_eq!(response.content, "I'll use a tool.");
assert_eq!(response.tool_calls.len(), 1);
assert_eq!(response.tool_calls[0].name, "lookup_weather");
assert_eq!(response.finish_reason, Some(ChatFinishReason::ToolCalls));
let requests = runtime.block_on(server.received_requests()).unwrap();
assert_eq!(requests.len(), 1);
let body: serde_json::Value = serde_json::from_slice(&requests[0].body).unwrap();
assert_eq!(body["messages"][0]["role"], "system");
assert_eq!(body["messages"][1]["role"], "user");
assert_eq!(body["messages"][2]["role"], "assistant");
assert_eq!(body["tools"][0]["function"]["name"], "lookup_weather");
assert_eq!(body["response_format"]["type"], "json_object");
assert_eq!(body["stop"][0], "DONE");
drop(server);
drop(runtime);
}
#[test]
fn test_build_request_with_assistant_tool_call_history() {
let backend = OpenAiBackend::new("test-key");
let req = ChatRequest {
messages: vec![
ChatMessage {
role: ChatRole::User,
content: "Weather in Paris?".to_string(),
tool_calls: Vec::new(),
tool_call_id: None,
},
ChatMessage {
role: ChatRole::Assistant,
content: String::new(),
tool_calls: vec![ChatToolCall {
id: "call_1".to_string(),
name: "lookup_weather".to_string(),
arguments: r#"{"city":"Paris"}"#.to_string(),
}],
tool_call_id: None,
},
ChatMessage {
role: ChatRole::Tool,
content: r#"{"temp_c":18}"#.to_string(),
tool_calls: Vec::new(),
tool_call_id: Some("call_1".to_string()),
},
],
system: None,
tools: Vec::new(),
response_format: ResponseFormat::default(),
max_tokens: None,
temperature: None,
stop_sequences: Vec::new(),
model: None,
};
let request = backend.build_request(&req);
assert_eq!(request.messages[1].role, "assistant");
assert!(request.messages[1].content.is_none());
assert_eq!(
request.messages[1].tool_calls.as_ref().map(Vec::len),
Some(1)
);
assert_eq!(request.messages[2].role, "tool");
assert_eq!(request.messages[2].tool_call_id.as_deref(), Some("call_1"));
}
}