Hare/yoi
1
1
Fork 0
Code Actions Packages Releases Activity

refactor: fold provider crate into engine

Browse Source
This commit is contained in:
Hare
2026-07-15 00:58:29 +09:00
parent 1505c5e8a3
commit 839fc7b40c
33 changed files with 263 additions and 279 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/llm-engine/src/llm_client/auth.rs
-8
View File
@@ -40,12 +40,4 @@ pub enum AuthRequirement {
pub trait AuthProvider: Send + Sync + std::fmt::Debug {
/// 1 リクエスト分の認証ヘッダを返す。refresh が必要なら内部で行う。
async fn headers(&self) -> Result<Vec<(HeaderName, HeaderValue)>, ClientError>;
/// Conversation header / request compression が必要な backend profile かどうか。
///
/// transport は呼び出し側の具象型を知らないため、この hook だけで
/// 追加の wire behavior を切り替える。
fn is_codex_backend(&self) -> bool {
false
}
}
crates/llm-engine/src/llm_client/retry.rs
+4 -4
View File
@@ -1,15 +1,15 @@
//! LLM response stream を開く前の transient error 向けリトライポリシー。
//!
//! Engine が `LlmClient::stream` の open error に対して `is_retryable` を見て
//! retry / backoff / TUI event / cancellation をまとめて管理する。
//! `LlmClient::stream` の open error に対して `is_retryable` を見て
//! retry / backoff / cancellation をまとめて管理する。
//! SSE 読み出し開始後の失敗は対象外。
use std::time::Duration;
/// 指数バックオフ + ジッター + 累積タイムアウトを表すポリシー。
///
/// `Default` は llm-engine 全体の固定値を返す。manifest 経由の上書きが
/// 必要になったら拡張する（現状は不要 → `tickets/llm-engine-transient-retry.md`）。
/// `Default` は llm-engine 全体の固定値を返す。呼び出し側からの上書きが
/// 必要になったら拡張する。
#[derive(Debug, Clone)]
pub struct RetryPolicy {
/// 指数の基準値。`base * 2^attempt` を `cap` で頭打ちにした上限から
crates/llm-engine/src/llm_client/scheme/anthropic/capability.rs
+2 -2
View File
@@ -1,7 +1,7 @@
//! Anthropic scheme の wire-level 既定 capability。
//!
//! モデル ID 固有のテーブル(`claude-*` など)は高レベル構築層
//! (`provider::capability`)の責務。ここでは未知モデルでも「この wire で
//! モデル ID 固有のテーブル(`claude-*` など)は client construction layer
//! の責務。ここでは未知モデルでも「この wire で
//! 安全に送れる最小共通項」を返すだけに留める。
use crate::llm_client::capability::{
crates/llm-engine/src/llm_client/scheme/gemini/capability.rs
+1 -1
View File
@@ -1,7 +1,7 @@
//! Gemini scheme の wire-level 既定 capability。
//!
//! モデル ID 固有のテーブル(`gemini-*` バージョン別の reasoning 有無)は
//! 高レベル構築層(`provider::capability`)の責務。ここでは wire の
//! client construction layer の責務。ここでは wire の
//! 保守的 default のみ。
use crate::llm_client::capability::{
crates/llm-engine/src/llm_client/scheme/openai_chat/capability.rs
+2 -2
View File
@@ -1,7 +1,7 @@
//! OpenAI Chat Completions scheme の wire-level 既定 capability。
//!
//! モデル ID 固有のテーブル(`gpt-5` 系など)は高レベル構築層
//! (`provider::capability`)の責務。ここでは wire の保守的 default のみ。
//! モデル ID 固有のテーブル(`gpt-5` 系など)は client construction layer
//! の責務。ここでは wire の保守的 default のみ。
use crate::llm_client::capability::{
CacheStrategy, ModelCapability, StructuredOutput, ToolCallingSupport,
crates/llm-engine/src/llm_client/scheme/openai_responses/scheme_impl.rs
+3 -3
View File
@@ -65,7 +65,7 @@ impl Scheme for OpenAIResponsesScheme {
if !self.send_max_output_tokens && config.max_tokens.is_some() {
warnings.push(ConfigWarning::unsupported(
"max_tokens",
"OpenAI Responses (ChatGPT backend)",
"OpenAI Responses compatible backend",
));
}
// Same for `temperature` / `top_p` on compatible backends that
@@ -74,13 +74,13 @@ impl Scheme for OpenAIResponsesScheme {
if config.temperature.is_some() {
warnings.push(ConfigWarning::unsupported(
"temperature",
"OpenAI Responses (ChatGPT backend)",
"OpenAI Responses compatible backend",
));
}
if config.top_p.is_some() {
warnings.push(ConfigWarning::unsupported(
"top_p",
"OpenAI Responses (ChatGPT backend)",
"OpenAI Responses compatible backend",
));
}
}
crates/llm-engine/src/llm_client/transport.rs
+115 -43
View File
@@ -1,8 +1,8 @@
//! `HttpTransport<S: Scheme>`: すべての LLM wire scheme を共通の 1 本の
//! HTTP クライアントで扱う。
//!
//! 旧 `providers/{anthropic,openai,gemini,ollama}.rs` を置き換える。
//! scheme 固有の差分は [`Scheme`] trait 実装に委譲する。
//! scheme 固有の差分は [`Scheme`] trait 実装に委譲し、backend 固有の
//! HTTP policy は [`TransportPolicy`] で明示的に差し込む。
use std::pin::Pin;
use std::sync::Arc;
@@ -63,6 +63,71 @@ impl ResolvedAuth {
}
}
/// Request body encoding policy used by [`HttpTransport`].
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum RequestBodyEncoding {
/// Send the request body as plain JSON.
Json,
/// Send the request body as zstd-compressed JSON with `Content-Encoding: zstd`.
ZstdJson,
}
impl Default for RequestBodyEncoding {
fn default() -> Self {
Self::Json
}
}
/// Conversation header policy used by [`HttpTransport`].
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum ConversationHeaderPolicy {
/// Do not derive any transport headers from [`Request::cache_key`].
None,
/// Send OpenAI-compatible conversation headers from [`Request::cache_key`].
OpenAiCompatible {
/// Send the legacy `session_id` header in addition to `session-id`.
include_legacy_session_id: bool,
/// Send `thread-id` with the same value as `session-id`.
include_thread_id: bool,
/// Send `x-client-request-id` with the same value as `session-id`.
include_client_request_id: bool,
},
}
impl Default for ConversationHeaderPolicy {
fn default() -> Self {
Self::None
}
}
/// Backend-specific HTTP transport policy.
#[derive(Debug, Clone, PartialEq, Eq, Default)]
pub struct TransportPolicy {
/// How to serialize and encode the HTTP request body.
pub request_body_encoding: RequestBodyEncoding,
/// Optional backend-specific conversation headers derived from request metadata.
pub conversation_headers: ConversationHeaderPolicy,
}
impl TransportPolicy {
/// Plain JSON requests with no derived conversation headers.
pub fn standard() -> Self {
Self::default()
}
/// OpenAI-compatible backend profile that uses zstd JSON bodies and
/// conversation headers derived from [`Request::cache_key`].
pub fn openai_compatible_zstd() -> Self {
Self {
request_body_encoding: RequestBodyEncoding::ZstdJson,
conversation_headers: ConversationHeaderPolicy::OpenAiCompatible {
include_legacy_session_id: true,
include_thread_id: true,
include_client_request_id: true,
},
}
}
}
fn header_value_for_diagnostics(headers: &HeaderMap, name: &str) -> Option<String> {
headers
.get(name)
@@ -120,6 +185,7 @@ pub struct HttpTransport<S: Scheme> {
base_url: String,
auth: ResolvedAuth,
capability: ModelCapability,
policy: TransportPolicy,
}
impl<S: Scheme> HttpTransport<S> {
@@ -141,9 +207,16 @@ impl<S: Scheme> HttpTransport<S> {
base_url,
auth,
capability,
policy: TransportPolicy::default(),
}
}
/// Set a backend-specific HTTP transport policy.
pub fn with_transport_policy(mut self, policy: TransportPolicy) -> Self {
self.policy = policy;
self
}
/// カスタム HTTP クライアントを差し込む（テスト等）。
pub fn with_http_client(mut self, client: reqwest::Client) -> Self {
self.http_client = client;
@@ -206,13 +279,6 @@ impl<S: Scheme> HttpTransport<S> {
Ok(headers)
}
fn is_codex_backend(&self) -> bool {
match &self.auth {
ResolvedAuth::Custom(provider) => provider.is_codex_backend(),
_ => false,
}
}
fn apply_stream_headers(
&self,
headers: &mut HeaderMap,
@@ -220,18 +286,26 @@ impl<S: Scheme> HttpTransport<S> {
) -> Result<(), ClientError> {
headers.insert(ACCEPT, HeaderValue::from_static("text/event-stream"));
if self.is_codex_backend()
if let ConversationHeaderPolicy::OpenAiCompatible {
include_legacy_session_id,
include_thread_id,
include_client_request_id,
} = self.policy.conversation_headers
&& let Some(cache_key) = request.cache_key.as_deref()
{
let value = HeaderValue::from_str(cache_key).map_err(|e| {
ClientError::Config(format!("invalid Codex conversation header: {e}"))
ClientError::Config(format!("invalid conversation header value: {e}"))
})?;
// Send both current hyphenated conversation headers and the
// legacy underscore form for compatibility with existing backends.
headers.insert(HeaderName::from_static("session-id"), value.clone());
headers.insert(HeaderName::from_static("thread-id"), value.clone());
headers.insert(HeaderName::from_static("session_id"), value.clone());
headers.insert(HeaderName::from_static("x-client-request-id"), value);
if include_thread_id {
headers.insert(HeaderName::from_static("thread-id"), value.clone());
}
if include_legacy_session_id {
headers.insert(HeaderName::from_static("session_id"), value.clone());
}
if include_client_request_id {
headers.insert(HeaderName::from_static("x-client-request-id"), value);
}
}
Ok(())
@@ -242,19 +316,22 @@ impl<S: Scheme> HttpTransport<S> {
body: &serde_json::Value,
headers: &mut HeaderMap,
) -> Result<RequestBody, ClientError> {
if !self.is_codex_backend() {
return Ok(RequestBody::Json(body.clone()));
match self.policy.request_body_encoding {
RequestBodyEncoding::Json => Ok(RequestBody::Json(body.clone())),
RequestBodyEncoding::ZstdJson => {
let raw = serde_json::to_vec(body)?;
let raw_json_bytes = raw.len();
let compressed =
zstd::stream::encode_all(std::io::Cursor::new(raw), 3).map_err(|e| {
ClientError::Config(format!("failed to zstd-compress request: {e}"))
})?;
headers.insert(CONTENT_ENCODING, HeaderValue::from_static("zstd"));
Ok(RequestBody::CompressedJson {
bytes: compressed,
raw_json_bytes,
})
}
}
let raw = serde_json::to_vec(body)?;
let raw_json_bytes = raw.len();
let compressed = zstd::stream::encode_all(std::io::Cursor::new(raw), 3)
.map_err(|e| ClientError::Config(format!("failed to zstd-compress request: {e}")))?;
headers.insert(CONTENT_ENCODING, HeaderValue::from_static("zstd"));
Ok(RequestBody::CompressedJson {
bytes: compressed,
raw_json_bytes,
})
}
}
@@ -383,6 +460,7 @@ impl<S: Scheme + Clone> Clone for HttpTransport<S> {
base_url: self.base_url.clone(),
auth: self.auth.clone(),
capability: self.capability.clone(),
policy: self.policy.clone(),
}
}
}
@@ -446,7 +524,8 @@ impl<S: Scheme + Clone + 'static> LlmClient for HttpTransport<S> {
"path": path,
"auth_kind": auth_kind(&self.auth),
"required_auth": format!("{:?}", self.scheme.required_auth()),
"codex_backend": self.is_codex_backend(),
"request_body_encoding": format!("{:?}", self.policy.request_body_encoding),
"conversation_headers": format!("{:?}", self.policy.conversation_headers),
"cache_key_present": request.cache_key.is_some(),
"stream_open_timeout_ms": DEFAULT_STREAM_OPEN_TIMEOUT.as_millis() as u64,
}),
@@ -677,9 +756,7 @@ mod tests {
use serde_json::json;
#[derive(Debug)]
struct TestAuthProvider {
codex: bool,
}
struct TestAuthProvider;
#[async_trait]
impl AuthProvider for TestAuthProvider {
@@ -695,10 +772,6 @@ mod tests {
),
])
}
fn is_codex_backend(&self) -> bool {
self.codex
}
}
#[derive(Clone)]
@@ -827,10 +900,9 @@ mod tests {
}
#[tokio::test]
async fn codex_backend_adds_conversation_headers_and_zstd_body() {
let transport = transport(ResolvedAuth::Custom(Arc::new(TestAuthProvider {
codex: true,
})));
async fn transport_policy_adds_conversation_headers_and_zstd_body() {
let transport = transport(ResolvedAuth::Custom(Arc::new(TestAuthProvider)))
.with_transport_policy(TransportPolicy::openai_compatible_zstd());
let request = Request::new().user("hello").cache_key("segment-123");
let mut headers = transport.build_headers().await.unwrap();
transport
@@ -869,7 +941,7 @@ mod tests {
raw_json_bytes,
} = encoded
else {
panic!("Codex backend request body must be zstd-compressed");
panic!("transport policy should zstd-compress request body");
};
assert!(raw_json_bytes > 0);
let decoded = zstd::stream::decode_all(std::io::Cursor::new(compressed)).unwrap();
@@ -878,7 +950,7 @@ mod tests {
}
#[tokio::test]
async fn non_codex_request_does_not_get_codex_only_headers_or_compression() {
async fn standard_policy_does_not_get_conversation_headers_or_compression() {
let transport = transport(ResolvedAuth::ApiKey("api-key".to_string()));
let request = Request::new().user("hello").cache_key("segment-123");
let mut headers = transport.build_headers().await.unwrap();
@@ -900,7 +972,7 @@ mod tests {
assert!(headers.get(CONTENT_ENCODING).is_none());
let RequestBody::Json(decoded) = encoded else {
panic!("non-Codex request body must remain normal JSON");
panic!("standard transport policy should keep request body as JSON");
};
assert_eq!(decoded["prompt_cache_key"], "segment-123");
}