diff --git a/research_api/db.py b/research_api/db.py
index f9a04362e..a72ac0014 100644
--- a/research_api/db.py
+++ b/research_api/db.py
@@ -1,18 +1,17 @@
 """Research Workbench database session management."""
-from pathlib import Path
 from sqlalchemy import create_engine
 from sqlalchemy.orm import sessionmaker
-from sqlalchemy.pool import StaticPool
+from sqlalchemy.pool import NullPool
 
-_COMFYUI_ROOT = Path(__file__).parent.parent
-_DB_PATH = _COMFYUI_ROOT / "research_workbench.db"
+from research_app.config import get_research_paths
 
+_DB_PATH = get_research_paths().db_path
 DATABASE_URL = f"sqlite:///{_DB_PATH}"
 
 engine = create_engine(
     DATABASE_URL,
     connect_args={"check_same_thread": False},
-    poolclass=StaticPool,
+    poolclass=NullPool,
 )
 session_maker = sessionmaker(bind=engine)
 
diff --git a/research_api/models.py b/research_api/models.py
index 33f56c03d..c9d84534d 100644
--- a/research_api/models.py
+++ b/research_api/models.py
@@ -1,11 +1,11 @@
 """Research Workbench SQLAlchemy models."""
 import uuid
 from datetime import datetime
-from sqlalchemy import Column, String, Float, Integer, DateTime, Boolean, ForeignKey, Text, JSON
-from sqlalchemy.orm import relationship
-import app.database.models as models
 
-Base = models.Base
+from sqlalchemy import Boolean, Column, DateTime, Float, ForeignKey, Index, Integer, JSON, String, Text
+from sqlalchemy.orm import relationship
+
+from research_api.base import Base
 
 
 def new_id():
@@ -14,12 +14,19 @@ def new_id():
 
 class Project(Base):
     __tablename__ = "projects"
+    __table_args__ = (
+        Index("ix_projects_updated_at", "updated_at"),
+        Index("ix_projects_status", "status"),
+    )
 
     id = Column(String, primary_key=True, default=new_id)
     title = Column(String, nullable=False)
     goal = Column(String, nullable=True)
     current_direction = Column(String, nullable=True)
     status = Column(String, default="active")  # active, paused, completed
+    stage = Column(String, default="文献调研")  # 文献调研, 实验设计, 实验实施, 论文写作, 投稿, 完成
+    urgency = Column(String, default="normal")  # urgent, normal, low
+    expected_completion = Column(DateTime, nullable=True)  # 预期完成时间
     last_active_at = Column(DateTime, nullable=True)
     created_at = Column(DateTime, default=datetime.utcnow)
     updated_at = Column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
@@ -47,8 +54,14 @@ class Intent(Base):
 
 class PaperAsset(Base):
     __tablename__ = "paper_assets"
+    __table_args__ = (
+        Index("ix_paper_assets_library_status", "library_status"),
+        Index("ix_paper_assets_updated_at", "updated_at"),
+        Index("ix_paper_assets_project_id", "project_id"),
+    )
 
     id = Column(String, primary_key=True, default=new_id)
+    project_id = Column(String, ForeignKey("projects.id"), nullable=True)
     title = Column(String, nullable=False)
     authors_text = Column(String, nullable=True)
     journal_or_source = Column(String, nullable=True)
@@ -64,12 +77,20 @@ class PaperAsset(Base):
     read_status = Column(String, default="unread")  # unread, quick-reviewed, skimmed, deeply-read
     library_status = Column(String, default="pending")  # pending, library
     style_candidate = Column(Boolean, default=False)
+    asset_type = Column(String, default="paper")  # paper, figure, table, code, dataset
+    tags = Column(JSON, nullable=True)  # List of tags stored as JSON
+    local_path = Column(String, nullable=True)  # Path to local MinerU folder or asset file
+    content_text = Column(Text, nullable=True)  # Extracted text/markdown content for papers
     created_at = Column(DateTime, default=datetime.utcnow)
     updated_at = Column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
 
 
 class ClaimAsset(Base):
     __tablename__ = "claim_assets"
+    __table_args__ = (
+        Index("ix_claim_assets_project_id", "project_id"),
+        Index("ix_claim_assets_support_level", "support_level"),
+    )
 
     id = Column(String, primary_key=True, default=new_id)
     project_id = Column(String, ForeignKey("projects.id"), nullable=False)
@@ -107,14 +128,20 @@ class Source(Base):
 
 class FeedItem(Base):
     __tablename__ = "feed_items"
+    __table_args__ = (
+        Index("ix_feed_items_status", "status"),
+        Index("ix_feed_items_source_id", "source_id"),
+    )
 
     id = Column(String, primary_key=True, default=new_id)
     source_id = Column(String, ForeignKey("sources.id"), nullable=True)
     external_key = Column(String, nullable=True)
     title = Column(String, nullable=False)
+    title_zh = Column(String, nullable=True)  # 中文标题
     authors_text = Column(String, nullable=True)
     published_at = Column(String, nullable=True)
     abstract = Column(Text, nullable=True)
+    abstract_zh = Column(Text, nullable=True)  # 中文摘要
     source_url = Column(String, nullable=True)
     pdf_url = Column(String, nullable=True)
     doi = Column(String, nullable=True)
diff --git a/research_api/routes/_db_helpers.py b/research_api/routes/_db_helpers.py
index 051c93c19..f9d1ef717 100644
--- a/research_api/routes/_db_helpers.py
+++ b/research_api/routes/_db_helpers.py
@@ -1,9 +1,10 @@
 """Async DB helpers that wrap sync SQLAlchemy with run_in_executor."""
 import asyncio
 from functools import partial
+
+from research_api.base import to_dict
 from research_api.db import create_session
 from research_api.models import Project, Intent, PaperAsset, ClaimAsset, Source, FeedItem, StyleAsset
-from app.database.models import to_dict
 
 
 def _sync_list_projects():
@@ -30,6 +31,28 @@ def _sync_get_project(project_id):
         return to_dict(p) if p else None
 
 
+def _sync_update_project(project_id, data):
+    from datetime import datetime
+    with create_session() as session:
+        from sqlalchemy import select
+        result = session.execute(select(Project).where(Project.id == project_id))
+        project = result.scalar_one_or_none()
+        if not project:
+            return None
+        for key, value in data.items():
+            if hasattr(project, key):
+                # Handle date conversion for DateTime columns
+                if key == "expected_completion" and isinstance(value, str):
+                    try:
+                        value = datetime.fromisoformat(value.replace("Z", "+00:00"))
+                    except (ValueError, AttributeError):
+                        pass  # Keep original value if parsing fails
+                setattr(project, key, value)
+        session.commit()
+        session.refresh(project)
+        return to_dict(project)
+
+
 def _sync_list_intents(project_id):
     with create_session() as session:
         from sqlalchemy import select
@@ -176,19 +199,20 @@ def _sync_update_source(source_id, data):
         return to_dict(source)
 
 
-def _sync_get_today_feed():
+def _sync_get_today_feed(limit: int = 50, offset: int = 0):
     with create_session() as session:
         from sqlalchemy import select
         result = session.execute(
             select(FeedItem)
             .where(FeedItem.status.in_(["discovered", "ranked", "presented"]))
             .order_by(FeedItem.rank_score.desc())
-            .limit(20)
+            .limit(limit)
+            .offset(offset)
         )
         return [to_dict(i) for i in result.scalars().all()]
 
 
-def _sync_list_feed(source_id=None, status=None):
+def _sync_list_feed(source_id=None, status=None, limit: int = 50, offset: int = 0):
     with create_session() as session:
         from sqlalchemy import select
         query = select(FeedItem)
@@ -196,7 +220,7 @@ def _sync_list_feed(source_id=None, status=None):
             query = query.where(FeedItem.source_id == source_id)
         if status:
             query = query.where(FeedItem.status == status)
-        query = query.order_by(FeedItem.rank_score.desc())
+        query = query.order_by(FeedItem.rank_score.desc()).limit(limit).offset(offset)
         result = session.execute(query)
         return [to_dict(i) for i in result.scalars().all()]
 
@@ -226,87 +250,90 @@ def _sync_update_feed_item(item_id, data):
 
 
 # Async wrappers using run_in_executor
-loop = asyncio.get_event_loop
 
 
 async def asyncio_get_projects():
-    return await loop().run_in_executor(None, _sync_list_projects)
+    return await asyncio.get_running_loop().run_in_executor(None, _sync_list_projects)
 
 
 async def asyncio_create_project(data):
-    return await loop().run_in_executor(None, partial(_sync_create_project, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_project, data))
 
 
 async def asyncio_get_project(project_id):
-    return await loop().run_in_executor(None, partial(_sync_get_project, project_id))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_get_project, project_id))
+
+
+async def asyncio_update_project(project_id, data):
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_project, project_id, data))
 
 
 async def asyncio_list_intents(project_id):
-    return await loop().run_in_executor(None, partial(_sync_list_intents, project_id))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_list_intents, project_id))
 
 
 async def asyncio_create_intent(data):
-    return await loop().run_in_executor(None, partial(_sync_create_intent, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_intent, data))
 
 
 async def asyncio_update_intent(intent_id, data):
-    return await loop().run_in_executor(None, partial(_sync_update_intent, intent_id, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_intent, intent_id, data))
 
 
 async def asyncio_list_papers(library_status=None, read_status=None):
-    return await loop().run_in_executor(None, partial(_sync_list_papers, library_status, read_status))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_list_papers, library_status, read_status))
 
 
 async def asyncio_create_paper(data):
-    return await loop().run_in_executor(None, partial(_sync_create_paper, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_paper, data))
 
 
 async def asyncio_get_paper(paper_id):
-    return await loop().run_in_executor(None, partial(_sync_get_paper, paper_id))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_get_paper, paper_id))
 
 
 async def asyncio_update_paper(paper_id, data):
-    return await loop().run_in_executor(None, partial(_sync_update_paper, paper_id, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_paper, paper_id, data))
 
 
 async def asyncio_list_claims(project_id=None, support_level=None):
-    return await loop().run_in_executor(None, partial(_sync_list_claims, project_id, support_level))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_list_claims, project_id, support_level))
 
 
 async def asyncio_create_claim(data):
-    return await loop().run_in_executor(None, partial(_sync_create_claim, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_claim, data))
 
 
 async def asyncio_update_claim(claim_id, data):
-    return await loop().run_in_executor(None, partial(_sync_update_claim, claim_id, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_claim, claim_id, data))
 
 
 async def asyncio_list_sources():
-    return await loop().run_in_executor(None, _sync_list_sources)
+    return await asyncio.get_running_loop().run_in_executor(None, _sync_list_sources)
 
 
 async def asyncio_create_source(data):
-    return await loop().run_in_executor(None, partial(_sync_create_source, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_source, data))
 
 
 async def asyncio_update_source(source_id, data):
-    return await loop().run_in_executor(None, partial(_sync_update_source, source_id, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_source, source_id, data))
 
 
-async def asyncio_get_today_feed():
-    return await loop().run_in_executor(None, _sync_get_today_feed)
+async def asyncio_get_today_feed(limit: int = 50, offset: int = 0):
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_get_today_feed, limit, offset))
 
 
-async def asyncio_list_feed(source_id=None, status=None):
-    return await loop().run_in_executor(None, partial(_sync_list_feed, source_id, status))
+async def asyncio_list_feed(source_id=None, status=None, limit: int = 50, offset: int = 0):
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_list_feed, source_id, status, limit, offset))
 
 
 async def asyncio_create_feed_item(data):
-    return await loop().run_in_executor(None, partial(_sync_create_feed_item, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_feed_item, data))
 
 
 async def asyncio_update_feed_item(item_id, data):
-    return await loop().run_in_executor(None, partial(_sync_update_feed_item, item_id, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_feed_item, item_id, data))
 
 
 # StyleAsset helpers
@@ -362,20 +389,29 @@ def _sync_delete_style(style_id):
 
 
 async def asyncio_list_styles():
-    return await loop().run_in_executor(None, _sync_list_styles)
+    return await asyncio.get_running_loop().run_in_executor(None, _sync_list_styles)
 
 
 async def asyncio_create_style(data):
-    return await loop().run_in_executor(None, partial(_sync_create_style, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_create_style, data))
 
 
 async def asyncio_get_style(style_id):
-    return await loop().run_in_executor(None, partial(_sync_get_style, style_id))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_get_style, style_id))
 
 
 async def asyncio_update_style(style_id, data):
-    return await loop().run_in_executor(None, partial(_sync_update_style, style_id, data))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_update_style, style_id, data))
 
 
 async def asyncio_delete_style(style_id):
-    return await loop().run_in_executor(None, partial(_sync_delete_style, style_id))
+    return await asyncio.get_running_loop().run_in_executor(None, partial(_sync_delete_style, style_id))
+
+
+async def asyncio_run_feed_discovery(categories, keywords, limit_per_keyword):
+    """Run feed discovery using academic APIs (Semantic Scholar, CrossRef)."""
+    def _sync_run():
+        from custom_nodes.research.feed_discovery import run_discovery
+        return run_discovery(categories, keywords, limit_per_keyword)
+
+    return await asyncio.get_running_loop().run_in_executor(None, _sync_run)
diff --git a/research_api/routes/research_routes.py b/research_api/routes/research_routes.py
index 1d64c77c3..cc1924fc5 100644
--- a/research_api/routes/research_routes.py
+++ b/research_api/routes/research_routes.py
@@ -1,9 +1,19 @@
 """Research API routes using aiohttp."""
+import os
 from aiohttp import web
+
+from research_app.config import resolve_external_asset_path
+from research_app.integrations import (
+    ensure_default_jobs,
+    get_default_keywords,
+    get_feed_categories,
+    get_scheduler,
+)
 from research_api.routes._db_helpers import (
     asyncio_get_projects,
     asyncio_create_project,
     asyncio_get_project,
+    asyncio_update_project,
     asyncio_list_intents,
     asyncio_create_intent,
     asyncio_update_intent,
@@ -26,13 +36,19 @@ from research_api.routes._db_helpers import (
     asyncio_get_style,
     asyncio_update_style,
     asyncio_delete_style,
+    asyncio_run_feed_discovery,
 )
 
+# In-memory store for latest evidence graph (per session)
+_latest_evidence_graph = {}
+
 
 class ResearchRoutes:
     def __init__(self):
         self.routes: web.RouteTableDef = web.RouteTableDef()
         self._app: web.Application = None
+        self.scheduler = get_scheduler()
+        self._jobs_registered = False
 
     def get_app(self) -> web.Application:
         if self._app is None:
@@ -61,6 +77,15 @@ class ResearchRoutes:
                 return web.json_response({"error": "Not found"}, status=404)
             return web.json_response(project)
 
+        @self.routes.patch("/research/projects/{project_id}")
+        async def update_project(request):
+            project_id = request.match_info["project_id"]
+            data = await request.json()
+            project = await asyncio_update_project(project_id, data)
+            if not project:
+                return web.json_response({"error": "Not found"}, status=404)
+            return web.json_response(project)
+
         @self.routes.get("/research/projects/{project_id}/intents")
         async def list_intents(request):
             project_id = request.match_info["project_id"]
@@ -115,6 +140,92 @@ class ResearchRoutes:
                 return web.json_response({"error": "Not found"}, status=404)
             return web.json_response(paper)
 
+        @self.routes.get("/research/papers/{paper_id}/content")
+        async def get_paper_content(request):
+            paper_id = request.match_info["paper_id"]
+            paper = await asyncio_get_paper(paper_id)
+            if not paper:
+                return web.json_response({"error": "Not found"}, status=404)
+
+            content = {"type": paper.get("asset_type", "paper")}
+
+            if paper.get("asset_type") == "paper" and paper.get("local_path"):
+                import os
+                local_path = paper.get("local_path")
+
+                # Extract folder name from local_path for image URLs
+                # local_path is like C:/r-assests/papers/FolderName or /c/r-assests/papers/FolderName
+                folder_name = os.path.basename(local_path.rstrip('/\\'))
+                content["folder_name"] = folder_name
+
+                # Read meta.json if exists
+                meta_path = os.path.join(local_path, "meta.json")
+                if os.path.exists(meta_path):
+                    import json
+                    with open(meta_path, 'r', encoding='utf-8') as f:
+                        content["meta"] = json.load(f)
+
+                # Read paper.md if exists
+                md_path = os.path.join(local_path, "paper.md")
+                if os.path.exists(md_path):
+                    with open(md_path, 'r', encoding='utf-8') as f:
+                        content["markdown"] = f.read()
+
+                # List images
+                img_path = os.path.join(local_path, "images")
+                if os.path.exists(img_path):
+                    content["images"] = [f for f in os.listdir(img_path) if f.endswith(('.jpg', '.png', '.gif', '.jpeg'))]
+
+            elif paper.get("asset_type") in ("figure", "table") and paper.get("local_path"):
+                import os
+                content["local_path"] = paper.get("local_path")
+                if os.path.exists(paper.get("local_path")):
+                    content["filename"] = os.path.basename(paper.get("local_path"))
+
+            return web.json_response(content)
+
+        @self.routes.get("/research/assets/{type}/{asset_id}")
+        async def get_asset(request):
+            """Get a specific asset by type and ID."""
+            asset_type = request.match_info["type"]
+            asset_id = request.match_info["asset_id"]
+
+            if asset_type == "paper":
+                paper = await asyncio_get_paper(asset_id)
+                if not paper:
+                    return web.json_response({"error": "Not found"}, status=404)
+                return web.json_response({"type": "paper", **paper})
+            elif asset_type == "claim":
+                # asyncio_get_claim not yet implemented in _db_helpers
+                return web.json_response({"error": "Not implemented"}, status=501)
+            elif asset_type == "style":
+                style = await asyncio_get_style(asset_id)
+                if not style:
+                    return web.json_response({"error": "Not found"}, status=404)
+                return web.json_response({"type": "style", **style})
+            else:
+                return web.json_response({"error": "Unknown asset type"}, status=400)
+
+        @self.routes.get("/research/assets/{path:.*}")
+        async def get_asset_file(request):
+            """Serve asset files from local paths."""
+            file_path = request.match_info["path"]
+            full_path = resolve_external_asset_path(file_path)
+            if full_path is None:
+                return web.json_response({"error": "Forbidden"}, status=403)
+
+            if not full_path.exists():
+                return web.json_response({"error": "Not found"}, status=404)
+
+            if full_path.is_dir():
+                # Return directory listing
+                files = os.listdir(full_path)
+                return web.json_response({"type": "directory", "files": files})
+
+            # Return file
+            from aiohttp.web_fileresponse import FileResponse
+            return FileResponse(full_path)
+
         # Claims
         @self.routes.get("/research/claims/")
         async def list_claims(request):
@@ -162,14 +273,18 @@ class ResearchRoutes:
         # Feed
         @self.routes.get("/research/feed/today")
         async def get_today_feed(request):
-            items = await asyncio_get_today_feed()
+            limit = int(request.query.get("limit", 50))
+            offset = int(request.query.get("offset", 0))
+            items = await asyncio_get_today_feed(limit=limit, offset=offset)
             return web.json_response(items)
 
         @self.routes.get("/research/feed/")
         async def list_feed(request):
             source_id = request.query.get("source_id")
             status = request.query.get("status")
-            items = await asyncio_list_feed(source_id, status)
+            limit = int(request.query.get("limit", 50))
+            offset = int(request.query.get("offset", 0))
+            items = await asyncio_list_feed(source_id, status, limit=limit, offset=offset)
             return web.json_response(items)
 
         @self.routes.post("/research/feed/")
@@ -187,6 +302,16 @@ class ResearchRoutes:
                 return web.json_response({"error": "Not found"}, status=404)
             return web.json_response(item)
 
+        @self.routes.post("/research/feed/discover")
+        async def discover_feed(request):
+            """Discover new papers from configured journal sources using academic APIs."""
+            data = await request.json() if request.can_read_body else {}
+            categories = data.get("categories", get_feed_categories())
+            keywords = data.get("keywords", get_default_keywords())
+            limit_per_keyword = data.get("limit_per_keyword", 3)
+            result = await asyncio_run_feed_discovery(categories, keywords, limit_per_keyword)
+            return web.json_response(result)
+
         # Styles
         @self.routes.get("/research/assets/styles/")
         async def list_styles(request):
@@ -223,3 +348,116 @@ class ResearchRoutes:
             if not result:
                 return web.json_response({"error": "Not found"}, status=404)
             return web.json_response({"status": "deleted"})
+
+        # Unified asset listing by type
+        @self.routes.get("/research/assets/")
+        async def list_assets(request):
+            """List assets filtered by type: ?type=paper|claim|style"""
+            asset_type = request.query.get("type")
+
+            if asset_type == "paper" or not asset_type:
+                papers = await asyncio_list_papers(
+                    library_status=request.query.get("library_status"),
+                    read_status=request.query.get("read_status")
+                )
+                return web.json_response([{"type": "paper", **p} for p in papers])
+            elif asset_type == "claim":
+                claims = await asyncio_list_claims(
+                    project_id=request.query.get("project_id"),
+                    support_level=request.query.get("support_level")
+                )
+                return web.json_response([{"type": "claim", **c} for c in claims])
+            elif asset_type == "style":
+                styles = await asyncio_list_styles()
+                return web.json_response([{"type": "style", **s} for s in styles])
+            else:
+                return web.json_response({"error": "Unknown asset type"}, status=400)
+
+        # Evidence Graph API
+        @self.routes.get("/research/api/graph/latest")
+        async def get_latest_graph(request):
+            """Get the latest evidence graph from node execution."""
+            return web.json_response(_latest_evidence_graph)
+
+        @self.routes.post("/research/api/graph/latest")
+        async def set_latest_graph(request):
+            """Set the latest evidence graph from node execution."""
+            global _latest_evidence_graph
+            data = await request.json()
+            _latest_evidence_graph = data
+            return web.json_response({"status": "ok"})
+
+        # Automation routes
+        @self.routes.get("/research/automation/jobs/")
+        async def list_jobs(request):
+            jobs = self.scheduler.list_jobs()
+            return web.json_response(jobs)
+
+        @self.routes.get("/research/automation/jobs/{job_id}")
+        async def get_job(request):
+            job_id = request.match_info["job_id"]
+            job = self.scheduler.get_job(job_id)
+            if not job:
+                return web.json_response({"error": "Job not found"}, status=404)
+            return web.json_response({
+                "id": job.id,
+                "name": job.name,
+                "schedule": {
+                    "interval_hours": job.schedule.interval_hours,
+                    "cron_hour": job.schedule.cron_hour,
+                    "cron_minute": job.schedule.cron_minute,
+                    "enabled": job.schedule.enabled,
+                },
+                "last_run": job.last_run.isoformat() if job.last_run else None,
+                "last_result": str(job.last_result) if job.last_result else None,
+                "last_error": job.error,
+                "run_count": job.run_count,
+            })
+
+        @self.routes.post("/research/automation/jobs/{job_id}/run")
+        async def trigger_job(request):
+            job_id = request.match_info["job_id"]
+            result = self.scheduler.trigger_now(job_id)
+            if result["status"] == "error" and "not found" in result.get("message", ""):
+                return web.json_response(result, status=404)
+            return web.json_response(result)
+
+        @self.routes.patch("/research/automation/jobs/{job_id}")
+        async def update_job(request):
+            job_id = request.match_info["job_id"]
+            job = self.scheduler.get_job(job_id)
+            if not job:
+                return web.json_response({"error": "Job not found"}, status=404)
+            data = await request.json()
+            if "enabled" in data:
+                job.schedule.enabled = data["enabled"]
+            if "interval_hours" in data:
+                job.schedule.interval_hours = float(data["interval_hours"])
+            if "cron_hour" in data:
+                job.schedule.cron_hour = data["cron_hour"]
+            if "cron_minute" in data:
+                job.schedule.cron_minute = data["cron_minute"]
+            return web.json_response({"status": "ok", "job_id": job_id})
+
+        @self.routes.post("/research/automation/scheduler/start")
+        async def start_scheduler(request):
+            self.scheduler.start()
+            return web.json_response({"status": "started"})
+
+        @self.routes.post("/research/automation/scheduler/stop")
+        async def stop_scheduler(request):
+            self.scheduler.stop()
+            return web.json_response({"status": "stopped"})
+
+        @self.routes.get("/research/automation/scheduler/status")
+        async def scheduler_status(request):
+            jobs = self.scheduler.list_jobs()
+            return web.json_response({
+                "running": True,
+                "jobs_count": len(jobs),
+                "jobs": jobs,
+            })
+
+        if not self._jobs_registered:
+            ensure_default_jobs()
+            self._jobs_registered = True