# api/server.py import os import time import threading from typing import Dict, List, Optional, Any, Tuple from fastapi import FastAPI, UploadFile, File, Form, Request from fastapi.responses import FileResponse, JSONResponse from fastapi.staticfiles import StaticFiles from fastapi.middleware.cors import CORSMiddleware from pydantic import BaseModel from api.config import DEFAULT_COURSE_TOPICS, DEFAULT_MODEL from api.rag_engine import ( build_rag_chunks_from_file, retrieve_relevant_chunks, retrieve_relevant_chunks_vector, add_file_to_course_index, ensure_course_dirs, ) from api.course_loader import load_all_courses from api.clare_core import ( detect_language, chat_with_clare, update_weaknesses_from_message, update_cognitive_state_from_message, render_session_status, export_conversation, summarize_conversation, ) # ✅ NEW: course directory + workspace schema routes from api.routes_directory import router as directory_router # ✅ LangSmith (optional) try: from langsmith import Client except Exception: Client = None # ---------------------------- # Paths / Constants # ---------------------------- API_DIR = os.path.dirname(__file__) COURSES_DIR = os.path.abspath(os.path.join(API_DIR, "..", "data", "courses")) MODULE10_PATH = os.path.join(API_DIR, "module10_responsible_ai.pdf") MODULE10_DOC_TYPE = "Literature Review / Paper" 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") LS_DATASET_NAME = os.getenv("LS_DATASET_NAME", "clare_user_events").strip() LS_PROJECT = os.getenv("LANGSMITH_PROJECT", os.getenv("LANGCHAIN_PROJECT", "")).strip() EXPERIMENT_ID = os.getenv("CLARE_EXPERIMENT_ID", "RESP_AI_W10").strip() # ---------------------------- # Health / Warmup (cold start mitigation) # ---------------------------- APP_START_TS = time.time() WARMUP_DONE = False WARMUP_ERROR: Optional[str] = None WARMUP_STARTED = False CLARE_ENABLE_WARMUP = os.getenv("CLARE_ENABLE_WARMUP", "1").strip() == "1" CLARE_WARMUP_BLOCK_READY = os.getenv("CLARE_WARMUP_BLOCK_READY", "0").strip() == "1" # Dataset logging (create_example) CLARE_ENABLE_LANGSMITH_LOG = os.getenv("CLARE_ENABLE_LANGSMITH_LOG", "0").strip() == "1" CLARE_LANGSMITH_ASYNC = os.getenv("CLARE_LANGSMITH_ASYNC", "1").strip() == "1" # Feedback logging (create_feedback -> attach to run_id) CLARE_ENABLE_LANGSMITH_FEEDBACK = os.getenv("CLARE_ENABLE_LANGSMITH_FEEDBACK", "1").strip() == "1" # ---------------------------- # App # ---------------------------- app = FastAPI(title="Clare API") app.add_middleware( CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) # ✅ include directory/workspace APIs BEFORE SPA fallback app.include_router(directory_router) # ---------------------------- # Static hosting (Vite build) # ---------------------------- if os.path.isdir(WEB_ASSETS): app.mount("/assets", StaticFiles(directory=WEB_ASSETS), name="assets") if os.path.isdir(WEB_DIST): app.mount("/static", StaticFiles(directory=WEB_DIST), name="static") @app.get("/") def index(): if os.path.exists(WEB_INDEX): return FileResponse(WEB_INDEX) return JSONResponse( {"detail": "web/build not found. Build frontend first (web/build/index.html)."}, status_code=500, ) # ---------------------------- # In-memory session store (MVP) # ---------------------------- SESSIONS: Dict[str, Dict[str, Any]] = {} # Course chunks cache (loaded at startup) COURSE_CHUNKS_CACHE: Dict[str, List[Dict[str, Any]]] = {} def _preload_module10_chunks() -> List[Dict[str, Any]]: if os.path.exists(MODULE10_PATH): try: return build_rag_chunks_from_file(MODULE10_PATH, MODULE10_DOC_TYPE) or [] except Exception as e: print(f"[preload] module10 parse failed: {repr(e)}") return [] return [] MODULE10_CHUNKS_CACHE = _preload_module10_chunks() def _get_session(user_id: str) -> Dict[str, Any]: if user_id not in SESSIONS: SESSIONS[user_id] = { "user_id": user_id, "name": "", "history": [], # List[Tuple[str, str]] "weaknesses": [], "cognitive_state": {"confusion": 0, "mastery": 0}, "course_outline": DEFAULT_COURSE_TOPICS, "rag_chunks": list(MODULE10_CHUNKS_CACHE), # legacy fallback "model_name": DEFAULT_MODEL, "uploaded_files": [], # NEW: profile init (MVP in-memory) "profile_bio": "", "init_answers": {}, "init_dismiss_until": 0, } SESSIONS[user_id].setdefault("uploaded_files", []) SESSIONS[user_id].setdefault("profile_bio", "") SESSIONS[user_id].setdefault("init_answers", {}) SESSIONS[user_id].setdefault("init_dismiss_until", 0) return SESSIONS[user_id] # NEW: deterministic “what files are loaded” hint for the LLM def _build_upload_hint(sess: Dict[str, Any], course_id: str) -> str: files = sess.get("uploaded_files") or [] if not files: return ( f"Course selected: {course_id}\n" "Recently uploaded files in this session: NONE.\n" "If the student asks about a specific uploaded file but none exist, ask them to upload." ) lines = [f"Course selected: {course_id}", "Recently uploaded files in this session:"] for f in files[-5:]: fn = (f.get("filename") or "").strip() dt = (f.get("doc_type") or "").strip() chunks = f.get("added_chunks") lines.append(f"- Uploaded: {fn} (doc_type={dt}, added_chunks={chunks})") lines.append( "If the student says 'the uploaded file', interpret it as the MOST RECENT uploaded file unless they specify otherwise." ) return "\n".join(lines) def _should_force_rag(message: str) -> bool: m = (message or "").lower() if not m: return False triggers = [ "summarize", "summary", "read", "analyze", "explain", "the uploaded file", "uploaded", "file", "document", "pdf", "slides", "ppt", "syllabus", "lecture", "reference", "references", "citation", "cite", "总结", "概括", "阅读", "读一下", "解析", "分析", "这份文件", "上传", "文档", "课件", "讲义", "引用", "参考", "出处", ] return any(t in m for t in triggers) def _extract_filename_hint(message: str) -> Optional[str]: m = (message or "").strip() if not m: return None for token in m.replace("“", '"').replace("”", '"').split(): if any(token.lower().endswith(ext) for ext in [".pdf", ".ppt", ".pptx", ".doc", ".docx"]): return os.path.basename(token.strip('"').strip("'").strip()) return None def _resolve_rag_scope(sess: Dict[str, Any], msg: str) -> Tuple[Optional[List[str]], Optional[List[str]]]: files = sess.get("uploaded_files") or [] msg_l = (msg or "").lower() hinted = _extract_filename_hint(msg) if hinted: known = {os.path.basename(f.get("filename", "")) for f in files if f.get("filename")} if hinted in known: return ([hinted], None) uploaded_intent = any(t in msg_l for t in [ "uploaded file", "uploaded files", "the uploaded file", "this file", "this document", "上传的文件", "这份文件", "这个文件", "文档", "课件", "讲义" ]) if uploaded_intent and files: last = files[-1] fn = os.path.basename(last.get("filename", "")).strip() or None dt = (last.get("doc_type") or "").strip() or None allowed_files = [fn] if fn else None allowed_doc_types = [dt] if dt else None return (allowed_files, allowed_doc_types) return (None, None) # ---------------------------- # Warmup # ---------------------------- def _do_warmup_once(): global WARMUP_DONE, WARMUP_ERROR, WARMUP_STARTED if WARMUP_STARTED: return WARMUP_STARTED = True try: from api.config import client client.models.list() _ = MODULE10_CHUNKS_CACHE WARMUP_DONE = True WARMUP_ERROR = None except Exception as e: WARMUP_DONE = False WARMUP_ERROR = repr(e) def _start_warmup_background(): if not CLARE_ENABLE_WARMUP: return threading.Thread(target=_do_warmup_once, daemon=True).start() @app.on_event("startup") def _on_startup(): global COURSE_CHUNKS_CACHE COURSE_CHUNKS_CACHE = load_all_courses(COURSES_DIR) or {} print("[startup] load_all_courses done. courses=", sorted(list(COURSE_CHUNKS_CACHE.keys()))) _start_warmup_background() # ---------------------------- # LangSmith helpers # ---------------------------- _ls_client = None if (Client is not None) and CLARE_ENABLE_LANGSMITH_LOG: try: _ls_client = Client() except Exception as e: print("[langsmith] init failed:", repr(e)) _ls_client = None def _log_event_to_langsmith(data: Dict[str, Any]): if _ls_client is None: return def _do(): 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")} 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)) if CLARE_LANGSMITH_ASYNC: threading.Thread(target=_do, daemon=True).start() else: _do() def _write_feedback_to_langsmith_run( run_id: str, rating: str, comment: str = "", tags: Optional[List[str]] = None, metadata: Optional[Dict[str, Any]] = None, ) -> bool: if not CLARE_ENABLE_LANGSMITH_FEEDBACK: return False if Client is None: return False rid = (run_id or "").strip() if not rid: return False try: ls = Client() score = 1 if rating == "helpful" else 0 meta = metadata or {} if tags is not None: meta["tags"] = tags if LS_PROJECT: meta.setdefault("langsmith_project", LS_PROJECT) ls.create_feedback( run_id=rid, key="ui_rating", score=score, comment=comment or "", metadata=meta, ) return True except Exception as e: print("[langsmith] create_feedback failed:", repr(e)) return False # ---------------------------- # Health endpoints # ---------------------------- @app.get("/health") def health(): return { "ok": True, "uptime_s": round(time.time() - APP_START_TS, 3), "warmup_enabled": CLARE_ENABLE_WARMUP, "warmup_started": bool(WARMUP_STARTED), "warmup_done": bool(WARMUP_DONE), "warmup_error": WARMUP_ERROR, "ready": bool(WARMUP_DONE) if CLARE_WARMUP_BLOCK_READY else True, "langsmith_enabled": bool(CLARE_ENABLE_LANGSMITH_LOG), "langsmith_async": bool(CLARE_LANGSMITH_ASYNC), "langsmith_feedback_enabled": bool(CLARE_ENABLE_LANGSMITH_FEEDBACK), "ts": int(time.time()), } @app.get("/ready") def ready(): if not CLARE_ENABLE_WARMUP or not CLARE_WARMUP_BLOCK_READY: return {"ready": True} if WARMUP_DONE: return {"ready": True} return JSONResponse({"ready": False, "error": WARMUP_ERROR}, status_code=503) # ---------------------------- # Quiz (Micro-Quiz) Instruction # ---------------------------- MICRO_QUIZ_INSTRUCTION = ( "We are running a short micro-quiz session based ONLY on **Module 10 – " "Responsible AI (Alto, 2024, Chapter 12)** and the pre-loaded materials.\n\n" "Step 1 – Before asking any content question:\n" "• First ask me which quiz style I prefer right now:\n" " - (1) Multiple-choice questions\n" " - (2) Short-answer / open-ended questions\n" "• Ask me explicitly: \"Which quiz style do you prefer now: 1) Multiple-choice or 2) Short-answer? " "Please reply with 1 or 2.\"\n" "• Do NOT start a content question until I have answered 1 or 2.\n\n" "Step 2 – After I choose the style:\n" "• If I choose 1 (multiple-choice):\n" " - Ask ONE multiple-choice question at a time, based on Module 10 concepts " "(Responsible AI definition, risk types, mitigation layers, EU AI Act, etc.).\n" " - Provide 3–4 options (A, B, C, D) and make only one option clearly correct.\n" "• If I choose 2 (short-answer):\n" " - Ask ONE short-answer question at a time, also based on Module 10 concepts.\n" " - Do NOT show the answer when you ask the question.\n\n" "Step 3 – For each answer I give:\n" "• Grade my answer (correct / partially correct / incorrect).\n" "• Give a brief explanation and the correct answer.\n" "• Then ask if I want another question of the SAME style.\n" "• Continue this pattern until I explicitly say to stop.\n\n" "Please start by asking me which quiz style I prefer (1 = multiple-choice, 2 = short-answer). " "Do not ask any content question before I choose." ) # ---------------------------- # Schemas # ---------------------------- class LoginReq(BaseModel): name: str user_id: str class ChatReq(BaseModel): user_id: str message: str learning_mode: str language_preference: str = "Auto" doc_type: str = "Syllabus" course_id: str = "course_ist345" class QuizStartReq(BaseModel): user_id: str language_preference: str = "Auto" doc_type: str = MODULE10_DOC_TYPE learning_mode: str = "quiz" course_id: str = "course_ist345" class ExportReq(BaseModel): user_id: str learning_mode: str class SummaryReq(BaseModel): user_id: str learning_mode: str language_preference: str = "Auto" class FeedbackReq(BaseModel): class Config: extra = "ignore" user_id: str rating: str # "helpful" | "not_helpful" run_id: Optional[str] = None assistant_message_id: Optional[str] = None assistant_text: str user_text: Optional[str] = "" comment: Optional[str] = "" tags: Optional[List[str]] = [] refs: Optional[List[str]] = [] learning_mode: Optional[str] = None doc_type: Optional[str] = None timestamp_ms: Optional[int] = None class ProfileDismissReq(BaseModel): user_id: str days: int = 7 class ProfileInitSubmitReq(BaseModel): user_id: str answers: Dict[str, Any] language_preference: str = "Auto" def _generate_profile_bio_with_clare( sess: Dict[str, Any], answers: Dict[str, Any], language_preference: str = "Auto", ) -> str: student_name = (sess.get("name") or "").strip() prompt = f""" You are Clare, an AI teaching assistant. Task: Generate a concise English Profile Bio for the student using ONLY the initialization answers provided below. Hard constraints: - Output language: English. - Tone: neutral, supportive, non-judgmental. - No medical/psychological diagnosis language. - Do not infer sensitive attributes (race, religion, political views, health status, sexuality, immigration status). - Length: 60–120 words. - Structure (4 short sentences max): 1) background & current context 2) learning goal for this course 3) learning preferences (format + pace) 4) how Clare will support them going forward (practical and concrete) Student name (if available): {student_name} Initialization answers (JSON): {answers} Return ONLY the bio text. Do not add a title. """.strip() resolved_lang = "English" try: bio, _unused_history, _run_id = chat_with_clare( message=prompt, history=[], model_name=sess["model_name"], language_preference=resolved_lang, learning_mode="summary", doc_type="Other Course Document", course_outline=sess["course_outline"], weaknesses=sess["weaknesses"], cognitive_state=sess["cognitive_state"], rag_context="", ) return (bio or "").strip() except Exception as e: print("[profile_bio] generate failed:", repr(e)) return "" # ---------------------------- # API Routes # ---------------------------- @app.post("/api/login") def login(req: LoginReq): user_id = (req.user_id or "").strip() name = (req.name or "").strip() if not user_id or not name: return JSONResponse({"ok": False, "error": "Missing name/user_id"}, status_code=400) sess = _get_session(user_id) sess["name"] = name return {"ok": True, "user": {"name": name, "user_id": user_id}} @app.post("/api/chat") 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) sess = _get_session(user_id) if not msg: return { "reply": "", "session_status_md": render_session_status( req.learning_mode, sess["weaknesses"], sess["cognitive_state"] ), "refs": [], "latency_ms": 0.0, "run_id": None, } t0 = time.time() marks_ms: Dict[str, float] = {"start": 0.0} resolved_lang = detect_language(msg, req.language_preference) marks_ms["language_detect_done"] = (time.time() - t0) * 1000.0 sess["weaknesses"] = update_weaknesses_from_message(msg, sess["weaknesses"]) marks_ms["weakness_update_done"] = (time.time() - t0) * 1000.0 sess["cognitive_state"] = update_cognitive_state_from_message(msg, sess["cognitive_state"]) marks_ms["cognitive_update_done"] = (time.time() - t0) * 1000.0 course_id = (req.course_id or "course_ist345").strip() # ✅ always define once allowed_files, allowed_doc_types = _resolve_rag_scope(sess, msg) force_rag = _should_force_rag(msg) need_rag = (force_rag or (len(msg) >= 20) or ("?" in msg)) if not need_rag: rag_context_text, rag_used_chunks = "", [] else: # ✅ Prefer course vector index rag_context_text, rag_used_chunks = retrieve_relevant_chunks_vector( query=msg, course_id=course_id, allowed_source_files=allowed_files, allowed_doc_types=allowed_doc_types, ) # ✅ Fallback to in-memory session chunks (legacy) if not rag_context_text and not rag_used_chunks: rag_context_text, rag_used_chunks = retrieve_relevant_chunks( msg, sess["rag_chunks"], allowed_source_files=allowed_files, allowed_doc_types=allowed_doc_types, ) marks_ms["rag_retrieve_done"] = (time.time() - t0) * 1000.0 # ✅ prepend deterministic hint so the model knows course+upload state upload_hint = _build_upload_hint(sess, course_id) if upload_hint: rag_context_text = (upload_hint + "\n\n---\n\n" + (rag_context_text or "")).strip() try: answer, new_history, run_id = chat_with_clare( message=msg, history=sess["history"], model_name=sess["model_name"], language_preference=resolved_lang, learning_mode=req.learning_mode, doc_type=req.doc_type, course_outline=sess["course_outline"], weaknesses=sess["weaknesses"], cognitive_state=sess["cognitive_state"], rag_context=rag_context_text, ) except Exception as e: print(f"[chat] error: {repr(e)}") return JSONResponse({"error": f"chat failed: {repr(e)}"}, status_code=500) marks_ms["llm_done"] = (time.time() - t0) * 1000.0 total_ms = marks_ms["llm_done"] ordered = [ "start", "language_detect_done", "weakness_update_done", "cognitive_update_done", "rag_retrieve_done", "llm_done", ] segments_ms: Dict[str, float] = {} for i in range(1, len(ordered)): a = ordered[i - 1] b = ordered[i] segments_ms[b] = max(0.0, marks_ms.get(b, 0.0) - marks_ms.get(a, 0.0)) latency_breakdown = {"marks_ms": marks_ms, "segments_ms": segments_ms, "total_ms": total_ms} sess["history"] = new_history refs = [ {"source_file": c.get("source_file"), "section": c.get("section")} for c in (rag_used_chunks or []) ] _log_event_to_langsmith( { "experiment_id": EXPERIMENT_ID, "student_id": user_id, "student_name": sess.get("name", ""), "event_type": "chat_turn", "timestamp": time.time(), "latency_ms": total_ms, "latency_breakdown": latency_breakdown, "rag_context_chars": len((rag_context_text or "")), "rag_used_chunks_count": len(rag_used_chunks or []), "history_len": len(sess["history"]), "question": msg, "answer": answer, "model_name": sess["model_name"], "language": resolved_lang, "learning_mode": req.learning_mode, "doc_type": req.doc_type, "refs": refs, "run_id": run_id, "course_id": course_id, } ) return { "reply": answer, "session_status_md": render_session_status( req.learning_mode, sess["weaknesses"], sess["cognitive_state"] ), "refs": refs, "latency_ms": total_ms, "run_id": run_id, } @app.post("/api/quiz/start") def quiz_start(req: QuizStartReq): user_id = (req.user_id or "").strip() if not user_id: return JSONResponse({"error": "Missing user_id"}, status_code=400) sess = _get_session(user_id) quiz_instruction = MICRO_QUIZ_INSTRUCTION t0 = time.time() resolved_lang = detect_language(quiz_instruction, req.language_preference) rag_context_text, rag_used_chunks = retrieve_relevant_chunks( "Module 10 quiz", sess["rag_chunks"] ) course_id = (req.course_id or "course_ist345").strip() upload_hint = _build_upload_hint(sess, course_id) if upload_hint: rag_context_text = (upload_hint + "\n\n---\n\n" + (rag_context_text or "")).strip() try: answer, new_history, run_id = chat_with_clare( message=quiz_instruction, history=sess["history"], model_name=sess["model_name"], language_preference=resolved_lang, learning_mode=req.learning_mode, doc_type=req.doc_type, course_outline=sess["course_outline"], weaknesses=sess["weaknesses"], cognitive_state=sess["cognitive_state"], rag_context=rag_context_text, ) except Exception as e: print(f"[quiz_start] error: {repr(e)}") return JSONResponse({"error": f"quiz_start failed: {repr(e)}"}, status_code=500) total_ms = (time.time() - t0) * 1000.0 sess["history"] = new_history refs = [ {"source_file": c.get("source_file"), "section": c.get("section")} for c in (rag_used_chunks or []) ] _log_event_to_langsmith( { "experiment_id": EXPERIMENT_ID, "student_id": user_id, "student_name": sess.get("name", ""), "event_type": "micro_quiz_start", "timestamp": time.time(), "latency_ms": total_ms, "question": "[micro_quiz_start] " + quiz_instruction[:200], "answer": answer, "model_name": sess["model_name"], "language": resolved_lang, "learning_mode": req.learning_mode, "doc_type": req.doc_type, "refs": refs, "rag_used_chunks_count": len(rag_used_chunks or []), "history_len": len(sess["history"]), "run_id": run_id, "course_id": course_id, } ) return { "reply": answer, "session_status_md": render_session_status( req.learning_mode, sess["weaknesses"], sess["cognitive_state"] ), "refs": refs, "latency_ms": total_ms, "run_id": run_id, } @app.post("/api/upload") async def upload( user_id: str = Form(...), doc_type: str = Form(...), course_id: str = Form("course_ist345"), file: UploadFile = File(...), ): user_id = (user_id or "").strip() doc_type = (doc_type or "").strip() course_id = (course_id or "course_ist345").strip() if not user_id: return JSONResponse({"ok": False, "error": "Missing user_id"}, status_code=400) if not file or not file.filename: return JSONResponse({"ok": False, "error": "Missing file"}, status_code=400) sess = _get_session(user_id) ensure_course_dirs(course_id) safe_name = os.path.basename(file.filename).replace("..", "_") raw_dir = os.path.join("data", "courses", course_id, "raw") os.makedirs(raw_dir, exist_ok=True) dest_path = os.path.join(raw_dir, safe_name) content = await file.read() with open(dest_path, "wb") as f: f.write(content) # ✅ Incremental vector index (course-level) try: result = add_file_to_course_index(course_id, dest_path, doc_type) added_chunks = int(result.get("added_chunks", 0)) total_chunks = int(result.get("total_chunks", 0)) except Exception as e: print(f"[upload] course index error: {repr(e)}") added_chunks, total_chunks = 0, 0 # ✅ also keep session chunks for legacy fallback try: new_chunks = build_rag_chunks_from_file(dest_path, doc_type) or [] sess["rag_chunks"] = (sess["rag_chunks"] or []) + new_chunks except Exception: pass sess["uploaded_files"] = sess.get("uploaded_files") or [] sess["uploaded_files"].append( { "filename": safe_name, "doc_type": doc_type, "added_chunks": added_chunks, "course_id": course_id, "ts": int(time.time()), } ) status_md = f"✅ Uploaded to {course_id}: {safe_name} (added_chunks={added_chunks}, total_chunks={total_chunks})" return { "ok": True, "course_id": course_id, "added_chunks": added_chunks, "total_chunks": total_chunks, "status_md": status_md, } @app.post("/api/feedback") 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": "Invalid rating"}, status_code=400) assistant_text = (req.assistant_text or "").strip() user_text = (req.user_text or "").strip() comment = (req.comment or "").strip() refs = req.refs or [] tags = req.tags or [] timestamp_ms = int(req.timestamp_ms or int(time.time() * 1000)) _log_event_to_langsmith( { "experiment_id": EXPERIMENT_ID, "student_id": user_id, "student_name": student_name, "event_type": "feedback", "timestamp": time.time(), "timestamp_ms": timestamp_ms, "rating": rating, "assistant_message_id": req.assistant_message_id, "run_id": req.run_id, "question": user_text, "answer": assistant_text, "comment": comment, "tags": tags, "refs": refs, "learning_mode": req.learning_mode, "doc_type": req.doc_type, } ) wrote_run_feedback = False if req.run_id: wrote_run_feedback = _write_feedback_to_langsmith_run( run_id=req.run_id, rating=rating, comment=comment, tags=tags, metadata={ "experiment_id": EXPERIMENT_ID, "student_id": user_id, "student_name": student_name, "assistant_message_id": req.assistant_message_id, "learning_mode": req.learning_mode, "doc_type": req.doc_type, "refs": refs, "timestamp_ms": timestamp_ms, }, ) return {"ok": True, "run_feedback_written": wrote_run_feedback} @app.post("/api/export") def api_export(req: ExportReq): user_id = (req.user_id or "").strip() if not user_id: return JSONResponse({"error": "Missing user_id"}, status_code=400) sess = _get_session(user_id) md = export_conversation( sess["history"], sess["course_outline"], req.learning_mode, sess["weaknesses"], sess["cognitive_state"], ) return {"markdown": md} @app.post("/api/summary") def api_summary(req: SummaryReq): user_id = (req.user_id or "").strip() if not user_id: return JSONResponse({"error": "Missing user_id"}, status_code=400) sess = _get_session(user_id) md = summarize_conversation( sess["history"], sess["course_outline"], sess["weaknesses"], sess["cognitive_state"], sess["model_name"], req.language_preference, ) return {"markdown": md} @app.get("/api/memoryline") def memoryline(user_id: str): _ = _get_session((user_id or "").strip()) return {"next_review_label": "T+7", "progress_pct": 0.4} @app.get("/api/profile/status") def profile_status(user_id: str): user_id = (user_id or "").strip() if not user_id: return JSONResponse({"error": "Missing user_id"}, status_code=400) sess = _get_session(user_id) bio = (sess.get("profile_bio") or "").strip() bio_len = len(bio) now = int(time.time()) dismissed_until = int(sess.get("init_dismiss_until") or 0) need_init = (bio_len <= 50) and (now >= dismissed_until) return {"need_init": need_init, "bio_len": bio_len, "dismissed_until": dismissed_until} @app.get("/api/debug/courses") def debug_courses(): course_ids = sorted(list(COURSE_CHUNKS_CACHE.keys())) return { "courses_dir": COURSES_DIR, "course_ids": course_ids, "chunk_counts": {cid: len(COURSE_CHUNKS_CACHE.get(cid) or []) for cid in course_ids}, "cache_loaded": True, } @app.post("/api/profile/dismiss") def profile_dismiss(req: ProfileDismissReq): user_id = (req.user_id or "").strip() if not user_id: return JSONResponse({"error": "Missing user_id"}, status_code=400) sess = _get_session(user_id) days = max(1, min(int(req.days or 7), 30)) sess["init_dismiss_until"] = int(time.time()) + days * 24 * 3600 return {"ok": True, "dismissed_until": sess["init_dismiss_until"]} @app.post("/api/profile/init_submit") def profile_init_submit(req: ProfileInitSubmitReq): user_id = (req.user_id or "").strip() if not user_id: return JSONResponse({"error": "Missing user_id"}, status_code=400) sess = _get_session(user_id) answers = req.answers or {} sess["init_answers"] = answers bio = _generate_profile_bio_with_clare(sess, answers, req.language_preference) if not bio: return JSONResponse({"error": "Failed to generate bio"}, status_code=500) sess["profile_bio"] = bio return {"ok": True, "bio": bio} # ---------------------------- # SPA Fallback # ---------------------------- @app.get("/{full_path:path}") def spa_fallback(full_path: str, request: Request): if ( full_path.startswith("api/") or full_path.startswith("assets/") or full_path.startswith("static/") ): return JSONResponse({"detail": "Not Found"}, status_code=404) if os.path.exists(WEB_INDEX): return FileResponse(WEB_INDEX) return JSONResponse( {"detail": "web/build not found. Build frontend first (web/build/index.html)."}, status_code=500, )