Update api/server.py

api/server.py

@@ -1,7 +1,7 @@
 # api/server.py
 import os
 import time
-from typing import Dict, Any, Optional, List
+from typing import Dict, List, Optional
 
 from fastapi import FastAPI, UploadFile, File, Form, Request
 from fastapi.responses import FileResponse, JSONResponse
@@ -22,57 +22,11 @@ from api.clare_core import (
     summarize_conversation,
 )
 
-# ----------------------------
-# LangSmith (Dataset logging)
-# ----------------------------
-# 你在 HF Space 里需要配置：
-#   LANGSMITH_API_KEY=...
-# 可选：
-#   LANGSMITH_DATASET_NAME=clare_user_events
-#   LANGSMITH_PROJECT=...
+# ✅ LangSmith (same idea as your Gradio app.py)
 try:
-    from langsmith import Client as LangSmithClient  # type: ignore
+    from langsmith import Client
 except Exception:
-    LangSmithClient = None  # type: ignore
-
-LS_DATASET_NAME = os.getenv("LANGSMITH_DATASET_NAME", "clare_user_events").strip()
-LS_PROJECT = os.getenv("LANGSMITH_PROJECT", "").strip()
-
-_ls_client = None
-if LangSmithClient is not None and os.getenv("LANGSMITH_API_KEY"):
-    try:
-        _ls_client = LangSmithClient()
-    except Exception as e:
-        print(f"[langsmith] init failed: {repr(e)}")
-        _ls_client = None
-
-
-def log_event_to_langsmith(
-    *,
-    inputs: Dict[str, Any],
-    outputs: Dict[str, Any],
-    metadata: Dict[str, Any],
-) -> None:
-    """
-    Write a single event as an Example row into LangSmith Dataset.
-    This mirrors your old Gradio pattern (dataset作为事件日志).
-    """
-    if _ls_client is None:
-        return
-    try:
-        # project 不是必须；dataset 足够你做过滤与分析
-        if LS_PROJECT:
-            metadata = {**metadata, "langsmith_project": LS_PROJECT}
-
-        _ls_client.create_example(
-            inputs=inputs,
-            outputs=outputs,
-            metadata=metadata,
-            dataset_name=LS_DATASET_NAME,
-        )
-    except Exception as e:
-        print(f"[langsmith] create_example failed: {repr(e)}")
-
+    Client = None
 
 # ----------------------------
 # Paths / Constants
@@ -82,11 +36,14 @@ API_DIR = os.path.dirname(__file__)
 MODULE10_PATH = os.path.join(API_DIR, "module10_responsible_ai.pdf")
 MODULE10_DOC_TYPE = "Literature Review / Paper"
 
-# Vite build output in your repo is "web/build"
 WEB_DIST = os.path.abspath(os.path.join(API_DIR, "..", "web", "build"))
 WEB_INDEX = os.path.join(WEB_DIST, "index.html")
 WEB_ASSETS = os.path.join(WEB_DIST, "assets")
 
+# ✅ LangSmith dataset name (match what you used before)
+LS_DATASET_NAME = os.getenv("LS_DATASET_NAME", "clare_user_events").strip()
+LS_PROJECT = os.getenv("LANGSMITH_PROJECT", os.getenv("LANGCHAIN_PROJECT", "")).strip()  # optional
+
 # ----------------------------
 # App
 # ----------------------------
@@ -154,6 +111,52 @@ def _get_session(user_id: str) -> Dict:
     return SESSIONS[user_id]
 
 
+# ----------------------------
+# LangSmith helpers
+# ----------------------------
+_ls_client = None
+if Client is not None:
+    try:
+        _ls_client = Client()
+    except Exception as e:
+        print("[langsmith] init failed:", repr(e))
+        _ls_client = None
+
+
+def _log_event_to_langsmith(data: Dict):
+    """
+    Create an Example in LangSmith Dataset (clare_user_events).
+    Mimic your previous Gradio log_event behavior.
+
+    Inputs/Outputs show up as "Inputs" / "Reference Outputs".
+    Everything else goes into metadata columns.
+    """
+    if _ls_client is None:
+        return
+
+    try:
+        inputs = {
+            "question": data.get("question", ""),
+            "student_id": data.get("student_id", ""),
+            "student_name": data.get("student_name", ""),
+        }
+        outputs = {"answer": data.get("answer", "")}
+        metadata = {k: v for k, v in data.items() if k not in ("question", "answer")}
+
+        # helpful for filtering in UI
+        if LS_PROJECT:
+            metadata.setdefault("langsmith_project", LS_PROJECT)
+
+        _ls_client.create_example(
+            inputs=inputs,
+            outputs=outputs,
+            metadata=metadata,
+            dataset_name=LS_DATASET_NAME,
+        )
+    except Exception as e:
+        print("[langsmith] log failed:", repr(e))
+
+
 # ----------------------------
 # Schemas
 # ----------------------------
@@ -182,19 +185,22 @@ class SummaryReq(BaseModel):
 
 
 class FeedbackReq(BaseModel):
-    # FE 会发的最小字段
     user_id: str
     rating: str  # "helpful" | "not_helpful"
-    assistant_message_id: str
+    assistant_message_id: Optional[str] = None
+
+    # what the user is rating
     assistant_text: str
+    user_text: Optional[str] = ""
+
+    # optional free-text comment
+    comment: Optional[str] = ""
 
-    # 可选：用于更好的分析
-    user_text: Optional[str] = None
-    comment: Optional[str] = None
-    refs: Optional[List[str]] = None
+    # context for analysis
+    refs: Optional[List[str]] = []
     learning_mode: Optional[str] = None
     doc_type: Optional[str] = None
-    timestamp_ms: Optional[float] = None
+    timestamp_ms: Optional[int] = None
 
 
 # ----------------------------
@@ -216,7 +222,6 @@ def login(req: LoginReq):
 def chat(req: ChatReq):
     user_id = (req.user_id or "").strip()
     msg = (req.message or "").strip()
-
     if not user_id:
         return JSONResponse({"error": "Missing user_id"}, status_code=400)
 
@@ -265,27 +270,24 @@ def chat(req: ChatReq):
         for c in (rag_used_chunks or [])
     ]
 
-    # 可选：把 chat_turn 也写入 dataset（你以前 Gradio 有）
-    try:
-        log_event_to_langsmith(
-            inputs={
-                "question": msg,
-                "student_id": user_id,
-            },
-            outputs={
-                "answer": answer,
-            },
-            metadata={
-                "event_type": "chat_turn",
-                "timestamp": time.time(),
-                "latency_ms": latency_ms,
-                "learning_mode": req.learning_mode,
-                "language": resolved_lang,
-                "doc_type": req.doc_type,
-            },
-        )
-    except Exception:
-        pass
+    # ✅ log chat_turn to LangSmith (uses login name/id; NO hardcoding)
+    _log_event_to_langsmith(
+        {
+            "experiment_id": "RESP_AI_W10",
+            "student_id": user_id,
+            "student_name": sess.get("name", ""),
+            "event_type": "chat_turn",
+            "timestamp": time.time(),
+            "latency_ms": latency_ms,
+            "question": msg,
+            "answer": answer,
+            "model_name": sess["model_name"],
+            "language": resolved_lang,
+            "learning_mode": req.learning_mode,
+            "doc_type": req.doc_type,
+            "refs": refs,
+        }
+    )
 
     return {
         "reply": answer,
@@ -323,7 +325,6 @@ async def upload(
     if doc_type == "Syllabus":
         class _F:
             pass
-
         fo = _F()
         fo.name = tmp_path
         try:
@@ -339,42 +340,58 @@ async def upload(
         new_chunks = []
 
     status_md = f"✅ Loaded base reading + uploaded {doc_type} file."
+
+    # ✅ optional: log upload event
+    _log_event_to_langsmith(
+        {
+            "experiment_id": "RESP_AI_W10",
+            "student_id": user_id,
+            "student_name": sess.get("name", ""),
+            "event_type": "upload",
+            "timestamp": time.time(),
+            "doc_type": doc_type,
+            "filename": safe_name,
+            "added_chunks": len(new_chunks),
+            "question": f"[upload] {safe_name}",
+            "answer": status_md,
+        }
+    )
+
     return {"ok": True, "added_chunks": len(new_chunks), "status_md": status_md}
 
 
 @app.post("/api/feedback")
-def feedback(req: FeedbackReq):
+def api_feedback(req: FeedbackReq):
     user_id = (req.user_id or "").strip()
     if not user_id:
         return JSONResponse({"ok": False, "error": "Missing user_id"}, status_code=400)
 
+    sess = _get_session(user_id)
+    student_name = sess.get("name", "")
+
     rating = (req.rating or "").strip().lower()
     if rating not in ("helpful", "not_helpful"):
-        return JSONResponse({"ok": False, "error": "rating must be helpful|not_helpful"}, status_code=400)
-
-    # 写入 LangSmith dataset（与你 Gradio 时代一致）
-    try:
-        log_event_to_langsmith(
-            inputs={
-                "question": req.user_text or "",   # 允许为空（只对 assistant reply 点赞）
-                "student_id": user_id,
-                "assistant_message_id": req.assistant_message_id,
-            },
-            outputs={
-                "answer": req.assistant_text or "",
-            },
-            metadata={
-                "event_type": "feedback",
-                "rating": rating,
-                "comment": (req.comment or "").strip(),
-                "learning_mode": req.learning_mode or "",
-                "doc_type": req.doc_type or "",
-                "refs": req.refs or [],
-                "timestamp_ms": req.timestamp_ms or (time.time() * 1000.0),
-            },
-        )
-    except Exception as e:
-        print(f"[feedback] log failed: {repr(e)}")
+        return JSONResponse({"ok": False, "error": "Invalid rating"}, status_code=400)
+
+    # ✅ record feedback as its own event row in the SAME dataset
+    _log_event_to_langsmith(
+        {
+            "experiment_id": "RESP_AI_W10",
+            "student_id": user_id,
+            "student_name": student_name,
+            "event_type": "feedback",
+            "timestamp": time.time(),
+            "rating": rating,
+            "assistant_message_id": req.assistant_message_id,
+            "question": (req.user_text or "").strip(),
+            "answer": (req.assistant_text or "").strip(),
+            "comment": (req.comment or "").strip(),
+            "refs": req.refs or [],
+            "learning_mode": req.learning_mode,
+            "doc_type": req.doc_type,
+            "timestamp_ms": req.timestamp_ms,
+        }
+    )
 
     return {"ok": True}