feat: Reasoningのコンテキスト管理の対応

2026-05-04 21:31:44 +09:00
parent e31cbcb150
commit 594671edc3
13 changed files with 1009 additions and 35 deletions
@@ -0,0 +1,212 @@
+//! Reasoning history round-trip 統合テスト
+//!
+//! Worker のストリーム → history append → 次リクエスト送出までの
+//! ライフサイクルで `Item::Reasoning` が脱落せず保持されることを確認する。
+//!
+//! 検証点:
+//! - Anthropic 由来の thinking + signature が `Item::Reasoning::signature` として
+//!   history に残る
+//! - OpenAI Responses 由来の reasoning text + summary + encrypted_content が
+//!   `Item::Reasoning` の各フィールドに展開される
+//! - 直前の reasoning は次の outgoing request の `request.items` の先頭付近に
+//!   含まれる（assistant メッセージの先頭、Anthropic 仕様）
+
+mod common;
+
+use common::MockLlmClient;
+use llm_worker::Item;
+use llm_worker::Worker;
+use llm_worker::llm_client::event::{
+    Event, ReasoningItemEvent, ResponseStatus, StatusEvent,
+};
+
+/// Anthropic 風: thinking ブロック → text → 終了 のシーケンス。
+/// Worker history に Reasoning(signature 付き) → assistant_message が並ぶ。
+#[tokio::test]
+async fn anthropic_thinking_round_trips_signature_into_history() {
+    let events = vec![
+        Event::ReasoningItem(ReasoningItemEvent {
+            id: None,
+            text: "let me think...".into(),
+            summary: Vec::new(),
+            encrypted_content: None,
+            signature: Some("SIG-OPUS".into()),
+        }),
+        Event::text_block_start(0),
+        Event::text_delta(0, "Here's the answer"),
+        Event::text_block_stop(0, None),
+        Event::Status(StatusEvent {
+            status: ResponseStatus::Completed,
+        }),
+    ];
+    let client = MockLlmClient::new(events);
+    let worker = Worker::new(client);
+    let out = worker.run("question?").await.expect("run ok");
+    let worker = out.worker;
+
+    let history = worker.history();
+    // user / reasoning / assistant_message
+    assert_eq!(history.len(), 3, "history: {history:?}");
+
+    assert!(matches!(history[0], Item::Message { .. }));
+    match &history[1] {
+        Item::Reasoning {
+            text, signature, ..
+        } => {
+            assert_eq!(text, "let me think...");
+            assert_eq!(signature.as_deref(), Some("SIG-OPUS"));
+        }
+        other => panic!("expected Reasoning, got {other:?}"),
+    }
+    assert_eq!(history[2].as_text(), Some("Here's the answer"));
+}
+
+/// OpenAI Responses 風: encrypted_content + summary を持った reasoning が
+/// `Item::Reasoning` のフィールドに展開されること。
+#[tokio::test]
+async fn openai_reasoning_round_trips_encrypted_and_summary() {
+    let events = vec![
+        Event::ReasoningItem(ReasoningItemEvent {
+            id: Some("r1".into()),
+            text: "inner reasoning".into(),
+            summary: vec!["sum-A".into(), "sum-B".into()],
+            encrypted_content: Some("ENC-OPAQUE".into()),
+            signature: None,
+        }),
+        Event::text_block_start(0),
+        Event::text_delta(0, "answer"),
+        Event::text_block_stop(0, None),
+        Event::Status(StatusEvent {
+            status: ResponseStatus::Completed,
+        }),
+    ];
+    let client = MockLlmClient::new(events);
+    let worker = Worker::new(client);
+    let out = worker.run("q").await.expect("run ok");
+    let worker = out.worker;
+
+    let history = worker.history();
+    match &history[1] {
+        Item::Reasoning {
+            text,
+            summary,
+            encrypted_content,
+            signature,
+            id,
+            ..
+        } => {
+            assert_eq!(text, "inner reasoning");
+            assert_eq!(summary, &vec!["sum-A".to_string(), "sum-B".to_string()]);
+            assert_eq!(encrypted_content.as_deref(), Some("ENC-OPAQUE"));
+            assert!(signature.is_none());
+            assert_eq!(id.as_deref(), Some("r1"));
+        }
+        other => panic!("expected Reasoning, got {other:?}"),
+    }
+}
+
+/// Reasoning は assistant ターン内で text/tool_call より先に並ぶこと（Anthropic
+/// が thinking を assistant メッセージの先頭に要求するため）。
+#[tokio::test]
+async fn reasoning_precedes_text_in_assistant_burst() {
+    let events = vec![
+        // text/tool_call とは独立に、ReasoningItem が中盤で発火しても、
+        // history append 時には assistant items の先頭に置かれる。
+        Event::text_block_start(0),
+        Event::text_delta(0, "intermediate"),
+        Event::text_block_stop(0, None),
+        Event::ReasoningItem(ReasoningItemEvent {
+            text: "after text".into(),
+            signature: Some("SIG".into()),
+            ..Default::default()
+        }),
+        Event::Status(StatusEvent {
+            status: ResponseStatus::Completed,
+        }),
+    ];
+    let client = MockLlmClient::new(events);
+    let worker = Worker::new(client);
+    let out = worker.run("q").await.expect("run ok");
+    let worker = out.worker;
+
+    let history = worker.history();
+    // user / reasoning(先頭) / assistant_message
+    assert!(matches!(history[1], Item::Reasoning { .. }));
+    assert_eq!(history[2].as_text(), Some("intermediate"));
+}
+
+/// resume シナリオ: history.json 由来の Item::Reasoning(signature) を Worker に
+/// 注入して run しても、次の outgoing request の `Request::items` にそのまま
+/// 載って LLM へ渡る（worker は items を改変しない契約）。
+#[tokio::test]
+async fn injected_reasoning_survives_into_outgoing_request() {
+    use async_trait::async_trait;
+    use futures::Stream;
+    use std::pin::Pin;
+    use std::sync::{Arc, Mutex};
+
+    use llm_worker::llm_client::{ClientError, LlmClient, Request};
+
+    /// Request を 1 度だけキャプチャして空ストリームを返す client。
+    #[derive(Clone)]
+    struct CapturingClient {
+        captured: Arc<Mutex<Option<Request>>>,
+    }
+
+    #[async_trait]
+    impl LlmClient for CapturingClient {
+        fn clone_boxed(&self) -> Box<dyn LlmClient> {
+            Box::new(self.clone())
+        }
+
+        async fn stream(
+            &self,
+            request: Request,
+        ) -> Result<Pin<Box<dyn Stream<Item = Result<Event, ClientError>> + Send>>, ClientError>
+        {
+            *self.captured.lock().unwrap() = Some(request);
+            let stream = futures::stream::iter(vec![Ok(Event::Status(StatusEvent {
+                status: ResponseStatus::Completed,
+            }))]);
+            Ok(Box::pin(stream))
+        }
+    }
+
+    let captured = Arc::new(Mutex::new(None));
+    let client = CapturingClient {
+        captured: captured.clone(),
+    };
+
+    let mut worker = Worker::new(client);
+    // resume: 既存 history を流し込む
+    worker.set_history(vec![
+        Item::user_message("prior question"),
+        Item::reasoning("prior thinking").with_signature("SIG-PRIOR"),
+        Item::assistant_message("prior answer"),
+    ]);
+
+    let _ = worker.run("follow up").await.expect("run ok");
+
+    let req = captured
+        .lock()
+        .unwrap()
+        .take()
+        .expect("client should have received a request");
+    // Reasoning item が outgoing items に保持されていること
+    let mut found = false;
+    for item in &req.items {
+        if let Item::Reasoning {
+            text, signature, ..
+        } = item
+        {
+            assert_eq!(text, "prior thinking");
+            assert_eq!(signature.as_deref(), Some("SIG-PRIOR"));
+            found = true;
+        }
+    }
+    assert!(
+        found,
+        "Reasoning item must survive into outgoing request items: {req:?}",
+        req = req.items,
+    );
+}