Update the release information of DeerFlow

* fix(middleware): handle repeated tool call ids

* add tests

* refactor(middleware): rely on tool result queues

backend/app/gateway/routers/thread_runs.py

@@ -66,6 +66,14 @@ class RunResponse(BaseModel):
     multitask_strategy: str = "reject"
     created_at: str = ""
     updated_at: str = ""
+    total_input_tokens: int = 0
+    total_output_tokens: int = 0
+    total_tokens: int = 0
+    llm_call_count: int = 0
+    lead_agent_tokens: int = 0
+    subagent_tokens: int = 0
+    middleware_tokens: int = 0
+    message_count: int = 0
 
 
 class ThreadTokenUsageModelBreakdown(BaseModel):
@@ -111,6 +119,14 @@ def _record_to_response(record: RunRecord) -> RunResponse:
         multitask_strategy=record.multitask_strategy,
         created_at=record.created_at,
         updated_at=record.updated_at,
+        total_input_tokens=record.total_input_tokens,
+        total_output_tokens=record.total_output_tokens,
+        total_tokens=record.total_tokens,
+        llm_call_count=record.llm_call_count,
+        lead_agent_tokens=record.lead_agent_tokens,
+        subagent_tokens=record.subagent_tokens,
+        middleware_tokens=record.middleware_tokens,
+        message_count=record.message_count,
     )
 
 
@@ -402,8 +418,15 @@ async def list_run_events(
 
 @router.get("/{thread_id}/token-usage", response_model=ThreadTokenUsageResponse)
 @require_permission("threads", "read", owner_check=True)
-async def thread_token_usage(thread_id: str, request: Request) -> ThreadTokenUsageResponse:
+async def thread_token_usage(
+    thread_id: str,
+    request: Request,
+    include_active: bool = Query(default=False, description="Include running run progress snapshots"),
+) -> ThreadTokenUsageResponse:
     """Thread-level token usage aggregation."""
     run_store = get_run_store(request)
-    agg = await run_store.aggregate_tokens_by_thread(thread_id)
+    if include_active:
+        agg = await run_store.aggregate_tokens_by_thread(thread_id, include_active=True)
+    else:
+        agg = await run_store.aggregate_tokens_by_thread(thread_id)
     return ThreadTokenUsageResponse(thread_id=thread_id, **agg)

backend/packages/harness/deerflow/agents/middlewares/dangling_tool_call_middleware.py

@@ -15,6 +15,7 @@ to the end of the message list as before_model + add_messages reducer would do.
 
 import json
 import logging
+from collections import defaultdict, deque
 from collections.abc import Awaitable, Callable
 from typing import override
 
@@ -109,10 +110,10 @@ class DanglingToolCallMiddleware(AgentMiddleware[AgentState]):
         This normalizes model-bound causal order before provider serialization while
         preserving already-valid transcripts unchanged.
         """
-        tool_messages_by_id: dict[str, ToolMessage] = {}
+        tool_messages_by_id: dict[str, deque[ToolMessage]] = defaultdict(deque)
         for msg in messages:
             if isinstance(msg, ToolMessage):
-                tool_messages_by_id.setdefault(msg.tool_call_id, msg)
+                tool_messages_by_id[msg.tool_call_id].append(msg)
 
         tool_call_ids: set[str] = set()
         for msg in messages:
@@ -124,7 +125,6 @@ class DanglingToolCallMiddleware(AgentMiddleware[AgentState]):
                     tool_call_ids.add(tc_id)
 
         patched: list = []
-        consumed_tool_msg_ids: set[str] = set()
         patch_count = 0
         for msg in messages:
             if isinstance(msg, ToolMessage) and msg.tool_call_id in tool_call_ids:
@@ -136,13 +136,13 @@ class DanglingToolCallMiddleware(AgentMiddleware[AgentState]):
 
             for tc in self._message_tool_calls(msg):
                 tc_id = tc.get("id")
-                if not tc_id or tc_id in consumed_tool_msg_ids:
+                if not tc_id:
                     continue
 
-                existing_tool_msg = tool_messages_by_id.get(tc_id)
+                tool_msg_queue = tool_messages_by_id.get(tc_id)
+                existing_tool_msg = tool_msg_queue.popleft() if tool_msg_queue else None
                 if existing_tool_msg is not None:
                     patched.append(existing_tool_msg)
-                    consumed_tool_msg_ids.add(tc_id)
                 else:
                     patched.append(
                         ToolMessage(
@@ -152,7 +152,6 @@ class DanglingToolCallMiddleware(AgentMiddleware[AgentState]):
                             status="error",
                         )
                     )
-                    consumed_tool_msg_ids.add(tc_id)
                     patch_count += 1
 
         if patched == messages:

backend/packages/harness/deerflow/persistence/run/sql.py

@@ -227,9 +227,48 @@ class RunRepository(RunStore):
             await session.execute(update(RunRow).where(RunRow.run_id == run_id).values(**values))
             await session.commit()
 
-    async def aggregate_tokens_by_thread(self, thread_id: str) -> dict[str, Any]:
+    async def update_run_progress(
+        self,
+        run_id: str,
+        *,
+        total_input_tokens: int | None = None,
+        total_output_tokens: int | None = None,
+        total_tokens: int | None = None,
+        llm_call_count: int | None = None,
+        lead_agent_tokens: int | None = None,
+        subagent_tokens: int | None = None,
+        middleware_tokens: int | None = None,
+        message_count: int | None = None,
+        last_ai_message: str | None = None,
+        first_human_message: str | None = None,
+    ) -> None:
+        """Update token usage + convenience fields while a run is still active."""
+        values: dict[str, Any] = {"updated_at": datetime.now(UTC)}
+        optional_counters = {
+            "total_input_tokens": total_input_tokens,
+            "total_output_tokens": total_output_tokens,
+            "total_tokens": total_tokens,
+            "llm_call_count": llm_call_count,
+            "lead_agent_tokens": lead_agent_tokens,
+            "subagent_tokens": subagent_tokens,
+            "middleware_tokens": middleware_tokens,
+            "message_count": message_count,
+        }
+        for key, value in optional_counters.items():
+            if value is not None:
+                values[key] = value
+        if last_ai_message is not None:
+            values["last_ai_message"] = last_ai_message[:2000]
+        if first_human_message is not None:
+            values["first_human_message"] = first_human_message[:2000]
+        async with self._sf() as session:
+            await session.execute(update(RunRow).where(RunRow.run_id == run_id, RunRow.status == "running").values(**values))
+            await session.commit()
+
+    async def aggregate_tokens_by_thread(self, thread_id: str, *, include_active: bool = False) -> dict[str, Any]:
         """Aggregate token usage via a single SQL GROUP BY query."""
-        _completed = RunRow.status.in_(("success", "error"))
+        statuses = ("success", "error", "running") if include_active else ("success", "error")
+        _completed = RunRow.status.in_(statuses)
         _thread = RunRow.thread_id == thread_id
         model_name = func.coalesce(RunRow.model_name, "unknown")
 

backend/packages/harness/deerflow/runtime/journal.py

@@ -20,7 +20,7 @@ from __future__ import annotations
 import asyncio
 import logging
 import time
-from collections.abc import Mapping
+from collections.abc import Awaitable, Callable, Mapping
 from datetime import UTC, datetime
 from typing import TYPE_CHECKING, Any, cast
 from uuid import UUID
@@ -46,6 +46,8 @@ class RunJournal(BaseCallbackHandler):
         *,
         track_token_usage: bool = True,
         flush_threshold: int = 20,
+        progress_reporter: Callable[[dict], Awaitable[None]] | None = None,
+        progress_flush_interval: float = 5.0,
     ):
         super().__init__()
         self.run_id = run_id
@@ -53,10 +55,16 @@ class RunJournal(BaseCallbackHandler):
         self._store = event_store
         self._track_tokens = track_token_usage
         self._flush_threshold = flush_threshold
+        self._progress_reporter = progress_reporter
+        self._progress_flush_interval = progress_flush_interval
 
         # Write buffer
         self._buffer: list[dict] = []
         self._pending_flush_tasks: set[asyncio.Task[None]] = set()
+        self._pending_progress_task: asyncio.Task[None] | None = None
+        self._pending_progress_delayed = False
+        self._progress_dirty = False
+        self._last_progress_flush = 0.0
 
         # Token accumulators
         self._total_input_tokens = 0
@@ -294,6 +302,8 @@ class RunJournal(BaseCallbackHandler):
                     else:
                         self._lead_agent_tokens += total_tk
 
+                    self._schedule_progress_flush()
+
         if messages:
             self._counted_message_llm_run_ids.add(str(run_id))
 
@@ -445,6 +455,8 @@ class RunJournal(BaseCallbackHandler):
             else:
                 self._lead_agent_tokens += total_tk
 
+            self._schedule_progress_flush()
+
     def set_first_human_message(self, content: str) -> None:
         """Record the first human message for convenience fields."""
         self._first_human_msg = content[:2000] if content else None
@@ -474,6 +486,14 @@ class RunJournal(BaseCallbackHandler):
         """Force flush remaining buffer. Called in worker's finally block."""
         if self._pending_flush_tasks:
             await asyncio.gather(*tuple(self._pending_flush_tasks), return_exceptions=True)
+        while self._pending_progress_task is not None and not self._pending_progress_task.done():
+            if self._pending_progress_delayed:
+                self._pending_progress_task.cancel()
+                await asyncio.gather(self._pending_progress_task, return_exceptions=True)
+                self._progress_dirty = False
+                self._pending_progress_delayed = False
+                break
+            await asyncio.gather(self._pending_progress_task, return_exceptions=True)
 
         while self._buffer:
             batch = self._buffer[: self._flush_threshold]
@@ -484,6 +504,57 @@ class RunJournal(BaseCallbackHandler):
                 self._buffer = batch + self._buffer
                 raise
 
+    def _schedule_progress_flush(self) -> None:
+        """Best-effort throttled progress snapshot for active run visibility."""
+        if self._progress_reporter is None:
+            return
+        now = time.monotonic()
+        elapsed = now - self._last_progress_flush
+        if elapsed < self._progress_flush_interval:
+            self._progress_dirty = True
+            self._schedule_delayed_progress_flush(self._progress_flush_interval - elapsed)
+            return
+        if self._pending_progress_task is not None and not self._pending_progress_task.done():
+            self._progress_dirty = True
+            return
+        try:
+            loop = asyncio.get_running_loop()
+        except RuntimeError:
+            return
+        self._progress_dirty = False
+        self._pending_progress_task = loop.create_task(self._flush_progress_async(snapshot=self.get_completion_data()))
+
+    def _schedule_delayed_progress_flush(self, delay: float) -> None:
+        if self._pending_progress_task is not None and not self._pending_progress_task.done():
+            return
+        try:
+            loop = asyncio.get_running_loop()
+        except RuntimeError:
+            return
+        delay = max(0.0, delay)
+        self._pending_progress_delayed = delay > 0
+        self._pending_progress_task = loop.create_task(self._flush_progress_async(delay=delay))
+
+    async def _flush_progress_async(self, *, snapshot: dict | None = None, delay: float = 0.0) -> None:
+        if self._progress_reporter is None:
+            return
+        if delay > 0:
+            self._pending_progress_delayed = True
+            await asyncio.sleep(delay)
+            self._pending_progress_delayed = False
+        dirty_before_write = self._progress_dirty
+        self._progress_dirty = False
+        snapshot_to_write = snapshot or self.get_completion_data()
+        try:
+            await self._progress_reporter(snapshot_to_write)
+            self._last_progress_flush = time.monotonic()
+        except Exception:
+            logger.warning("Failed to persist progress snapshot for run %s", self.run_id, exc_info=True)
+        if dirty_before_write or self._progress_dirty:
+            self._progress_dirty = False
+            self._pending_progress_task = None
+            self._schedule_delayed_progress_flush(self._progress_flush_interval)
+
     def get_completion_data(self) -> dict:
         """Return accumulated token and message data for run completion."""
         return {

backend/packages/harness/deerflow/runtime/runs/manager.py

@@ -38,6 +38,16 @@ class RunRecord:
     error: str | None = None
     model_name: str | None = None
     store_only: bool = False
+    total_input_tokens: int = 0
+    total_output_tokens: int = 0
+    total_tokens: int = 0
+    llm_call_count: int = 0
+    lead_agent_tokens: int = 0
+    subagent_tokens: int = 0
+    middleware_tokens: int = 0
+    message_count: int = 0
+    last_ai_message: str | None = None
+    first_human_message: str | None = None
 
 
 class RunManager:
@@ -102,16 +112,53 @@ class RunManager:
             error=row.get("error"),
             model_name=row.get("model_name"),
             store_only=True,
+            total_input_tokens=row.get("total_input_tokens") or 0,
+            total_output_tokens=row.get("total_output_tokens") or 0,
+            total_tokens=row.get("total_tokens") or 0,
+            llm_call_count=row.get("llm_call_count") or 0,
+            lead_agent_tokens=row.get("lead_agent_tokens") or 0,
+            subagent_tokens=row.get("subagent_tokens") or 0,
+            middleware_tokens=row.get("middleware_tokens") or 0,
+            message_count=row.get("message_count") or 0,
+            last_ai_message=row.get("last_ai_message"),
+            first_human_message=row.get("first_human_message"),
         )
 
     async def update_run_completion(self, run_id: str, **kwargs) -> None:
         """Persist token usage and completion data to the backing store."""
+        async with self._lock:
+            record = self._runs.get(run_id)
+            if record is not None:
+                for key, value in kwargs.items():
+                    if key == "status":
+                        continue
+                    if hasattr(record, key) and value is not None:
+                        setattr(record, key, value)
+                record.updated_at = _now_iso()
         if self._store is not None:
             try:
                 await self._store.update_run_completion(run_id, **kwargs)
             except Exception:
                 logger.warning("Failed to persist run completion for %s", run_id, exc_info=True)
 
+    async def update_run_progress(self, run_id: str, **kwargs) -> None:
+        """Persist a running token/message snapshot without changing status."""
+        should_persist = True
+        async with self._lock:
+            record = self._runs.get(run_id)
+            if record is not None:
+                should_persist = record.status == RunStatus.running
+            if record is not None and should_persist:
+                for key, value in kwargs.items():
+                    if hasattr(record, key) and value is not None:
+                        setattr(record, key, value)
+                record.updated_at = _now_iso()
+        if should_persist and self._store is not None:
+            try:
+                await self._store.update_run_progress(run_id, **kwargs)
+            except Exception:
+                logger.warning("Failed to persist run progress for %s", run_id, exc_info=True)
+
     async def create(
         self,
         thread_id: str,

backend/packages/harness/deerflow/runtime/runs/store/base.py

@@ -95,12 +95,30 @@ class RunStore(abc.ABC):
     ) -> None:
         pass
 
+    async def update_run_progress(
+        self,
+        run_id: str,
+        *,
+        total_input_tokens: int | None = None,
+        total_output_tokens: int | None = None,
+        total_tokens: int | None = None,
+        llm_call_count: int | None = None,
+        lead_agent_tokens: int | None = None,
+        subagent_tokens: int | None = None,
+        middleware_tokens: int | None = None,
+        message_count: int | None = None,
+        last_ai_message: str | None = None,
+        first_human_message: str | None = None,
+    ) -> None:
+        """Persist a best-effort running snapshot without changing run status."""
+        return None
+
     @abc.abstractmethod
     async def list_pending(self, *, before: str | None = None) -> list[dict[str, Any]]:
         pass
 
     @abc.abstractmethod
-    async def aggregate_tokens_by_thread(self, thread_id: str) -> dict[str, Any]:
+    async def aggregate_tokens_by_thread(self, thread_id: str, *, include_active: bool = False) -> dict[str, Any]:
         """Aggregate token usage for completed runs in a thread.
 
         Returns a dict with keys: total_tokens, total_input_tokens,

backend/packages/harness/deerflow/runtime/runs/store/memory.py

@@ -82,14 +82,22 @@ class MemoryRunStore(RunStore):
                     self._runs[run_id][key] = value
             self._runs[run_id]["updated_at"] = datetime.now(UTC).isoformat()
 
+    async def update_run_progress(self, run_id, **kwargs):
+        if run_id in self._runs and self._runs[run_id].get("status") == "running":
+            for key, value in kwargs.items():
+                if value is not None:
+                    self._runs[run_id][key] = value
+            self._runs[run_id]["updated_at"] = datetime.now(UTC).isoformat()
+
     async def list_pending(self, *, before=None):
         now = before or datetime.now(UTC).isoformat()
         results = [r for r in self._runs.values() if r["status"] == "pending" and r["created_at"] <= now]
         results.sort(key=lambda r: r["created_at"])
         return results
 
-    async def aggregate_tokens_by_thread(self, thread_id: str) -> dict[str, Any]:
-        completed = [r for r in self._runs.values() if r["thread_id"] == thread_id and r.get("status") in ("success", "error")]
+    async def aggregate_tokens_by_thread(self, thread_id: str, *, include_active: bool = False) -> dict[str, Any]:
+        statuses = ("success", "error", "running") if include_active else ("success", "error")
+        completed = [r for r in self._runs.values() if r["thread_id"] == thread_id and r.get("status") in statuses]
         by_model: dict[str, dict] = {}
         for r in completed:
             model = r.get("model_name") or "unknown"

backend/packages/harness/deerflow/runtime/runs/worker.py

@@ -153,8 +153,6 @@ async def run_agent(
 
     journal = None
 
-    journal = None
-
     # Track whether "events" was requested but skipped
     if "events" in requested_modes:
         logger.info(
@@ -177,6 +175,7 @@ async def run_agent(
                 thread_id=thread_id,
                 event_store=event_store,
                 track_token_usage=getattr(run_events_config, "track_token_usage", True),
+                progress_reporter=lambda snapshot: run_manager.update_run_progress(run_id, **snapshot),
             )
 
         # 1. Mark running

backend/tests/test_dangling_tool_call_middleware.py

@@ -218,6 +218,70 @@ class TestBuildPatchedMessagesPatching:
 
         assert mw._build_patched_messages(msgs) is None
 
+    def test_reused_tool_call_ids_across_ai_turns_keep_their_own_tool_results(self):
+        mw = DanglingToolCallMiddleware()
+        msgs = [
+            HumanMessage(content="summary", name="summary", additional_kwargs={"hide_from_ui": True}),
+            _ai_with_tool_calls(
+                [
+                    _tc("web_search", "web_search:11"),
+                    _tc("web_search", "web_search:12"),
+                    _tc("web_search", "web_search:13"),
+                ]
+            ),
+            _tool_msg("web_search:11", "web_search"),
+            _tool_msg("web_search:12", "web_search"),
+            _tool_msg("web_search:13", "web_search"),
+            _ai_with_tool_calls(
+                [
+                    _tc("web_search", "web_search:9"),
+                    _tc("web_search", "web_search:10"),
+                    _tc("web_search", "web_search:11"),
+                ]
+            ),
+            _tool_msg("web_search:9", "web_search"),
+            _tool_msg("web_search:10", "web_search"),
+            _tool_msg("web_search:11", "web_search"),
+        ]
+
+        assert mw._build_patched_messages(msgs) is None
+
+    def test_reused_tool_call_id_patches_second_dangling_occurrence(self):
+        mw = DanglingToolCallMiddleware()
+        msgs = [
+            _ai_with_tool_calls([_tc("web_search", "web_search:11")]),
+            _tool_msg("web_search:11", "web_search"),
+            _ai_with_tool_calls([_tc("web_search", "web_search:11")]),
+        ]
+
+        patched = mw._build_patched_messages(msgs)
+
+        assert patched is not None
+        assert isinstance(patched[1], ToolMessage)
+        assert patched[1].tool_call_id == "web_search:11"
+        assert patched[1].status == "success"
+        assert isinstance(patched[3], ToolMessage)
+        assert patched[3].tool_call_id == "web_search:11"
+        assert patched[3].status == "error"
+
+    def test_reused_tool_call_id_consumes_later_result_for_first_dangling_occurrence(self):
+        mw = DanglingToolCallMiddleware()
+        result = _tool_msg("web_search:11", "web_search")
+        msgs = [
+            _ai_with_tool_calls([_tc("web_search", "web_search:11")]),
+            _ai_with_tool_calls([_tc("web_search", "web_search:11")]),
+            result,
+        ]
+
+        patched = mw._build_patched_messages(msgs)
+
+        assert patched is not None
+        assert patched[1] is result
+        assert patched[1].status == "success"
+        assert isinstance(patched[3], ToolMessage)
+        assert patched[3].tool_call_id == "web_search:11"
+        assert patched[3].status == "error"
+
     def test_tool_results_are_grouped_with_their_own_ai_turn_across_multiple_ai_messages(self):
         mw = DanglingToolCallMiddleware()
         msgs = [

backend/tests/test_run_journal.py

@@ -714,6 +714,110 @@ class TestExternalUsageRecords:
         assert j._subagent_tokens == 0
 
 
+class TestProgressSnapshots:
+    @pytest.mark.anyio
+    async def test_on_llm_end_reports_progress_snapshot(self):
+        snapshots: list[dict] = []
+
+        async def reporter(snapshot: dict) -> None:
+            snapshots.append(snapshot)
+
+        store = MemoryRunEventStore()
+        j = RunJournal(
+            "r1",
+            "t1",
+            store,
+            flush_threshold=100,
+            progress_reporter=reporter,
+            progress_flush_interval=0,
+        )
+        usage = {"input_tokens": 10, "output_tokens": 5, "total_tokens": 15}
+        j.on_llm_end(_make_llm_response("Answer", usage=usage), run_id=uuid4(), parent_run_id=None, tags=["lead_agent"])
+        await j.flush()
+
+        assert snapshots
+        assert snapshots[-1]["total_tokens"] == 15
+        assert snapshots[-1]["llm_call_count"] == 1
+        assert snapshots[-1]["message_count"] == 1
+        assert snapshots[-1]["last_ai_message"] == "Answer"
+
+    @pytest.mark.anyio
+    async def test_throttled_progress_flush_emits_trailing_snapshot(self):
+        snapshots: list[dict] = []
+        trailing_seen = asyncio.Event()
+
+        async def reporter(snapshot: dict) -> None:
+            snapshots.append(snapshot)
+            if snapshot["total_tokens"] == 45:
+                trailing_seen.set()
+
+        store = MemoryRunEventStore()
+        j = RunJournal(
+            "r1",
+            "t1",
+            store,
+            flush_threshold=100,
+            progress_reporter=reporter,
+            progress_flush_interval=0.01,
+        )
+        j.on_llm_end(
+            _make_llm_response("First", usage={"input_tokens": 10, "output_tokens": 5, "total_tokens": 15}),
+            run_id=uuid4(),
+            parent_run_id=None,
+            tags=["lead_agent"],
+        )
+        j.on_llm_end(
+            _make_llm_response("Second", usage={"input_tokens": 20, "output_tokens": 10, "total_tokens": 30}),
+            run_id=uuid4(),
+            parent_run_id=None,
+            tags=["lead_agent"],
+        )
+        await asyncio.wait_for(trailing_seen.wait(), timeout=1.0)
+        await j.flush()
+
+        assert len(snapshots) >= 2
+        assert snapshots[-1]["total_tokens"] == 45
+        assert snapshots[-1]["llm_call_count"] == 2
+        assert snapshots[-1]["last_ai_message"] == "Second"
+
+    @pytest.mark.anyio
+    async def test_flush_cancels_delayed_progress_without_final_progress_write(self):
+        snapshots: list[dict] = []
+
+        async def reporter(snapshot: dict) -> None:
+            snapshots.append(snapshot)
+
+        store = MemoryRunEventStore()
+        j = RunJournal(
+            "r1",
+            "t1",
+            store,
+            flush_threshold=100,
+            progress_reporter=reporter,
+            progress_flush_interval=10.0,
+        )
+        j.on_llm_end(
+            _make_llm_response("First", usage={"input_tokens": 10, "output_tokens": 5, "total_tokens": 15}),
+            run_id=uuid4(),
+            parent_run_id=None,
+            tags=["lead_agent"],
+        )
+        await asyncio.sleep(0)
+        assert snapshots[-1]["total_tokens"] == 15
+        j.on_llm_end(
+            _make_llm_response("Second", usage={"input_tokens": 20, "output_tokens": 10, "total_tokens": 30}),
+            run_id=uuid4(),
+            parent_run_id=None,
+            tags=["lead_agent"],
+        )
+
+        await asyncio.wait_for(j.flush(), timeout=0.2)
+
+        assert snapshots[-1]["total_tokens"] == 15
+        assert snapshots[-1]["llm_call_count"] == 1
+        assert snapshots[-1]["last_ai_message"] == "First"
+
+
 class TestChatModelStartHumanMessage:
     """Tests for on_chat_model_start extracting the first human message."""
 

backend/tests/test_run_repository.py

@@ -10,6 +10,7 @@ from sqlalchemy.dialects import postgresql
 
 from deerflow.persistence.run import RunRepository
 from deerflow.runtime import RunManager, RunStatus
+from deerflow.runtime.runs.store.base import RunStore
 
 
 async def _make_repo(tmp_path):
@@ -26,6 +27,42 @@ async def _cleanup():
     await close_engine()
 
 
+class _CustomRunStoreWithoutProgress(RunStore):
+    async def put(self, *args, **kwargs):
+        return None
+
+    async def get(self, *args, **kwargs):
+        return None
+
+    async def list_by_thread(self, *args, **kwargs):
+        return []
+
+    async def update_status(self, *args, **kwargs):
+        return None
+
+    async def delete(self, *args, **kwargs):
+        return None
+
+    async def update_model_name(self, *args, **kwargs):
+        return None
+
+    async def update_run_completion(self, *args, **kwargs):
+        return None
+
+    async def list_pending(self, *args, **kwargs):
+        return []
+
+    async def aggregate_tokens_by_thread(self, *args, **kwargs):
+        return {}
+
+
+@pytest.mark.anyio
+async def test_update_run_progress_defaults_to_noop_for_custom_store():
+    store = _CustomRunStoreWithoutProgress()
+
+    await store.update_run_progress("r1", total_tokens=1)
+
+
 class TestRunRepository:
     @pytest.mark.anyio
     async def test_put_and_get(self, tmp_path):
@@ -170,6 +207,69 @@ class TestRunRepository:
         assert row["total_tokens"] == 100
         await _cleanup()
 
+    @pytest.mark.anyio
+    async def test_update_run_progress_keeps_status_running(self, tmp_path):
+        repo = await _make_repo(tmp_path)
+        await repo.put("r1", thread_id="t1", status="running")
+        await repo.update_run_progress(
+            "r1",
+            total_input_tokens=40,
+            total_output_tokens=10,
+            total_tokens=50,
+            llm_call_count=1,
+            message_count=2,
+            last_ai_message="partial answer",
+        )
+        row = await repo.get("r1")
+        assert row["status"] == "running"
+        assert row["total_tokens"] == 50
+        assert row["llm_call_count"] == 1
+        assert row["message_count"] == 2
+        assert row["last_ai_message"] == "partial answer"
+        await _cleanup()
+
+    @pytest.mark.anyio
+    async def test_update_run_progress_preserves_omitted_fields(self, tmp_path):
+        repo = await _make_repo(tmp_path)
+        await repo.put("r1", thread_id="t1", status="running")
+        await repo.update_run_progress(
+            "r1",
+            total_input_tokens=40,
+            total_output_tokens=10,
+            total_tokens=50,
+            llm_call_count=1,
+            lead_agent_tokens=30,
+            subagent_tokens=20,
+            message_count=2,
+        )
+
+        await repo.update_run_progress("r1", total_tokens=60, last_ai_message="updated")
+
+        row = await repo.get("r1")
+        assert row["total_input_tokens"] == 40
+        assert row["total_output_tokens"] == 10
+        assert row["total_tokens"] == 60
+        assert row["llm_call_count"] == 1
+        assert row["lead_agent_tokens"] == 30
+        assert row["subagent_tokens"] == 20
+        assert row["message_count"] == 2
+        assert row["last_ai_message"] == "updated"
+        await _cleanup()
+
+    @pytest.mark.anyio
+    async def test_update_run_progress_skips_terminal_runs(self, tmp_path):
+        repo = await _make_repo(tmp_path)
+        await repo.put("r1", thread_id="t1", status="running")
+        await repo.update_run_completion("r1", status="success", total_tokens=100, llm_call_count=1)
+
+        await repo.update_run_progress("r1", total_tokens=200, llm_call_count=2)
+
+        row = await repo.get("r1")
+        assert row["status"] == "success"
+        assert row["total_tokens"] == 100
+        assert row["llm_call_count"] == 1
+        await _cleanup()
+
     @pytest.mark.anyio
     async def test_aggregate_tokens_by_thread_counts_completed_runs_only(self, tmp_path):
         repo = await _make_repo(tmp_path)
@@ -225,6 +325,28 @@ class TestRunRepository:
         }
         await _cleanup()
 
+    @pytest.mark.anyio
+    async def test_aggregate_tokens_by_thread_can_include_active_runs(self, tmp_path):
+        repo = await _make_repo(tmp_path)
+        await repo.put("success-run", thread_id="t1", status="running")
+        await repo.update_run_completion("success-run", status="success", total_tokens=100, lead_agent_tokens=100)
+        await repo.put("running-run", thread_id="t1", status="running")
+        await repo.update_run_progress("running-run", total_tokens=25, lead_agent_tokens=20, subagent_tokens=5)
+
+        without_active = await repo.aggregate_tokens_by_thread("t1")
+        with_active = await repo.aggregate_tokens_by_thread("t1", include_active=True)
+
+        assert without_active["total_tokens"] == 100
+        assert without_active["total_runs"] == 1
+        assert with_active["total_tokens"] == 125
+        assert with_active["total_runs"] == 2
+        assert with_active["by_caller"] == {
+            "lead_agent": 120,
+            "subagent": 5,
+            "middleware": 0,
+        }
+        await _cleanup()
+
     @pytest.mark.anyio
     async def test_list_by_thread_ordered_desc(self, tmp_path):
         """list_by_thread returns newest first."""

backend/tests/test_thread_token_usage.py

@@ -53,3 +53,30 @@ def test_thread_token_usage_returns_stable_shape():
         },
     }
     run_store.aggregate_tokens_by_thread.assert_awaited_once_with("thread-1")
+
+
+def test_thread_token_usage_can_include_active_runs():
+    run_store = MagicMock()
+    run_store.aggregate_tokens_by_thread = AsyncMock(
+        return_value={
+            "total_tokens": 175,
+            "total_input_tokens": 120,
+            "total_output_tokens": 55,
+            "total_runs": 3,
+            "by_model": {"unknown": {"tokens": 175, "runs": 3}},
+            "by_caller": {
+                "lead_agent": 145,
+                "subagent": 25,
+                "middleware": 5,
+            },
+        },
+    )
+    app = _make_app(run_store)
+
+    with TestClient(app) as client:
+        response = client.get("/api/threads/thread-1/token-usage?include_active=true")
+
+    assert response.status_code == 200
+    assert response.json()["total_tokens"] == 175
+    assert response.json()["total_runs"] == 3
+    run_store.aggregate_tokens_by_thread.assert_awaited_once_with("thread-1", include_active=True)

frontend/Makefile

@@ -18,3 +18,7 @@ lint:
 
 format:
 	pnpm format:write
+
+build-static:
+	NEXT_CONFIG_BUILD_OUTPUT=standalone SKIP_ENV_VALIDATION=1 NEXT_PUBLIC_STATIC_WEBSITE_ONLY=true pnpm build
+	@if [ -d .next/static ]; then mkdir -p .next/standalone/.next && cp -R .next/static .next/standalone/.next/static; fi

frontend/next.config.js

@@ -16,6 +16,10 @@ const withNextra = nextra({});
 
 /** @type {import("next").NextConfig} */
 const config = {
+  output:
+    process.env.NEXT_CONFIG_BUILD_OUTPUT === "standalone"
+      ? "standalone"
+      : undefined,
   i18n: {
     locales: ["en", "zh"],
     defaultLocale: "en",

frontend/public/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-master-photography-article.md

@@ -32,7 +32,7 @@ Even with digital Leicas, photographers often emulate film characteristics: natu
 
 ### Image 1: Parisian Decisive Moment
 
-![Paris Decisive Moment](/frontend/public/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-paris-decisive-moment.jpg)
+![Paris Decisive Moment](/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-paris-decisive-moment.jpg)
 
 This image captures the essence of Cartier-Bresson's philosophy. A woman in a red coat leaps over a puddle while a cyclist passes in perfect synchrony. The composition follows the rule of thirds, with the subject positioned at the intersection of grid lines. Shot with a simulated Leica M11 and 35mm Summicron lens at f/2.8, the image features shallow depth of field, natural film grain, and the warm, muted color palette characteristic of Leica photography.
 
@@ -40,7 +40,7 @@ The "decisive moment" here isn't just about timing—it's about the alignment of
 
 ### Image 2: Tokyo Night Reflections
 
-![Tokyo Night Scene](/frontend/public/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-tokyo-night.jpg)
+![Tokyo Night Scene](/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-tokyo-night.jpg)
 
 Moving to Shinjuku, Tokyo, this image explores the atmospheric possibilities of Leica's legendary Noctilux lens. Simulating a Leica M10-P with a 50mm f/0.95 Noctilux wide open, the image creates extremely shallow depth of field with beautiful bokeh balls from neon signs reflected in wet pavement.
 
@@ -48,7 +48,7 @@ A salaryman waits under glowing kanji signs, steam rising from a nearby ramen sh
 
 ### Image 3: New York City Candid
 
-![NYC Candid Scene](/frontend/public/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-nyc-candid.jpg)
+![NYC Candid Scene](/demo/threads/7f9dc56c-e49c-4671-a3d2-c492ff4dce0c/user-data/outputs/leica-nyc-candid.jpg)
 
 This Chinatown scene demonstrates the documentary power of Leica's Q2 camera with its fixed 28mm Summilux lens. The wide angle captures environmental context while maintaining intimate proximity to the subjects. A fishmonger hands a live fish to a customer while tourists photograph the scene—a moment of cultural contrast and authentic urban life.
 

frontend/src/app/workspace/chats/[thread_id]/layout.tsx

@@ -1,19 +1,19 @@
-"use client";
+import { isStaticWebsiteOnly } from "@/core/static-mode";
+import { DEMO_THREAD_IDS } from "@/core/threads/static-demo";
 
-import { PromptInputProvider } from "@/components/ai-elements/prompt-input";
-import { ArtifactsProvider } from "@/components/workspace/artifacts";
-import { SubtasksProvider } from "@/core/tasks/context";
+import { ChatProviders } from "./providers";
+
+export function generateStaticParams() {
+  if (!isStaticWebsiteOnly()) {
+    return [];
+  }
+  return DEMO_THREAD_IDS.map((thread_id) => ({ thread_id }));
+}
 
 export default function ChatLayout({
   children,
 }: {
   children: React.ReactNode;
 }) {
-  return (
-    <SubtasksProvider>
-      <ArtifactsProvider>
-        <PromptInputProvider>{children}</PromptInputProvider>
-      </ArtifactsProvider>
-    </SubtasksProvider>
-  );
+  return <ChatProviders>{children}</ChatProviders>;
 }

frontend/src/app/workspace/chats/[thread_id]/page.tsx

@@ -227,6 +227,7 @@ export default function ChatPage() {
                       isWelcomeMode && <Welcome mode={settings.context.mode} />
                     }
                     disabled={
+                      isMock ||
                       env.NEXT_PUBLIC_STATIC_WEBSITE_ONLY === "true" ||
                       isUploading
                     }

frontend/src/app/workspace/chats/[thread_id]/providers.tsx

@@ -0,0 +1,15 @@
+"use client";
+
+import { PromptInputProvider } from "@/components/ai-elements/prompt-input";
+import { ArtifactsProvider } from "@/components/workspace/artifacts";
+import { SubtasksProvider } from "@/core/tasks/context";
+
+export function ChatProviders({ children }: { children: React.ReactNode }) {
+  return (
+    <SubtasksProvider>
+      <ArtifactsProvider>
+        <PromptInputProvider>{children}</PromptInputProvider>
+      </ArtifactsProvider>
+    </SubtasksProvider>
+  );
+}

frontend/src/app/workspace/layout.tsx

@@ -43,12 +43,14 @@ export default async function WorkspaceLayout({
             >
               Retry
             </Link>
-            <Link
-              href="/api/v1/auth/logout"
-              className="text-muted-foreground hover:bg-muted rounded-md border px-4 py-2 text-sm"
-            >
-              Logout &amp; Reset
-            </Link>
+            <form action="/api/v1/auth/logout" method="post">
+              <button
+                type="submit"
+                className="text-muted-foreground hover:bg-muted rounded-md border px-4 py-2 text-sm"
+              >
+                Logout &amp; Reset
+              </button>
+            </form>
           </div>
         </div>
       );

frontend/src/components/workspace/artifacts/artifact-file-detail.tsx

@@ -83,7 +83,7 @@ export function ArtifactFileDetail({
   const isSupportPreview = useMemo(() => {
     return language === "html" || language === "markdown";
   }, [language]);
-  const { content } = useArtifactContent({
+  const { content, url } = useArtifactContent({
     threadId,
     filepath: filepathFromProps,
     enabled: isCodeFile && !isWriteFile,
@@ -254,7 +254,9 @@ export function ArtifactFileDetail({
           (language === "markdown" || language === "html") && (
             <ArtifactFilePreview
               content={displayContent}
+              isWriteFile={isWriteFile}
               language={language ?? "text"}
+              url={url}
             />
           )}
         {isCodeFile && viewMode === "code" && (
@@ -277,27 +279,33 @@ export function ArtifactFileDetail({
 
 export function ArtifactFilePreview({
   content,
+  isWriteFile,
   language,
+  url,
 }: {
   content: string;
+  isWriteFile: boolean;
   language: string;
+  url?: string;
 }) {
   const [htmlPreviewUrl, setHtmlPreviewUrl] = useState<string>();
 
   useEffect(() => {
-    if (language !== "html") {
+    if (language !== "html" || isWriteFile) {
       setHtmlPreviewUrl(undefined);
       return;
     }
 
-    const blob = new Blob([content ?? ""], { type: "text/html" });
-    const url = URL.createObjectURL(blob);
-    setHtmlPreviewUrl(url);
+    const blob = new Blob([htmlWithBaseHref(content ?? "", url)], {
+      type: "text/html",
+    });
+    const objectUrl = URL.createObjectURL(blob);
+    setHtmlPreviewUrl(objectUrl);
 
     return () => {
-      URL.revokeObjectURL(url);
+      URL.revokeObjectURL(objectUrl);
     };
-  }, [content, language]);
+  }, [content, isWriteFile, language, url]);
 
   if (language === "markdown") {
     return (
@@ -318,9 +326,35 @@ export function ArtifactFilePreview({
         className="size-full"
         title="Artifact preview"
         sandbox="allow-scripts allow-forms"
-        src={htmlPreviewUrl}
+        src={isWriteFile ? undefined : htmlPreviewUrl}
+        srcDoc={isWriteFile ? content : undefined}
       />
     );
   }
   return null;
 }
+
+function htmlWithBaseHref(content: string, url?: string) {
+  if (!url || /<base\s/i.exec(content)) {
+    return content;
+  }
+
+  const baseHref = htmlBaseHref(url);
+  const baseElement = `<base href="${escapeHtmlAttribute(baseHref)}">`;
+  if (/<head[^>]*>/i.exec(content)) {
+    return content.replace(/<head([^>]*)>/i, `<head$1>${baseElement}`);
+  }
+  return `${baseElement}${content}`;
+}
+
+function htmlBaseHref(url: string) {
+  const baseUrl = new URL(url, window.location.href);
+  baseUrl.pathname = baseUrl.pathname.replace(/\/[^/]*$/, "/");
+  baseUrl.search = "";
+  baseUrl.hash = "";
+  return baseUrl.toString();
+}
+
+function escapeHtmlAttribute(value: string) {
+  return value.replaceAll("&", "&amp;").replaceAll('"', "&quot;");
+}

frontend/src/content/en/index.mdx

@@ -20,27 +20,27 @@ If you want to understand how DeerFlow works, start with the Introduction. If yo
 
 Start with the conceptual overview first.
 
-- [Introduction](/docs/introduction)
-- [Why DeerFlow](/docs/introduction/why-deerflow)
-- [Harness vs App](/docs/introduction/harness-vs-app)
+- [Introduction](./docs/introduction)
+- [Why DeerFlow](./docs/introduction/why-deerflow)
+- [Harness vs App](./docs/introduction/harness-vs-app)
 
 ### If you want to build with DeerFlow
 
 Start with the Harness section. This path is for teams who want to integrate DeerFlow capabilities into their own system or build a custom agent product on top of the DeerFlow runtime.
 
-- [DeerFlow Harness](/docs/harness)
-- [Quick Start](/docs/harness/quick-start)
-- [Configuration](/docs/harness/configuration)
-- [Customization](/docs/harness/customization)
+- [DeerFlow Harness](./docs/harness)
+- [Quick Start](./docs/harness/quick-start)
+- [Configuration](./docs/harness/configuration)
+- [Customization](./docs/harness/customization)
 
 ### If you want to deploy and use DeerFlow
 
 Start with the App section. This path is for teams who want to run DeerFlow as a complete application and understand how to configure, operate, and use it in practice.
 
-- [DeerFlow App](/docs/app)
-- [Quick Start](/docs/app/quick-start)
-- [Deployment Guide](/docs/app/deployment-guide)
-- [Workspace Usage](/docs/app/workspace-usage)
+- [DeerFlow App](./docs/app)
+- [Quick Start](./docs/app/quick-start)
+- [Deployment Guide](./docs/app/deployment-guide)
+- [Workspace Usage](./docs/app/workspace-usage)
 
 ## Documentation structure
 
@@ -79,17 +79,17 @@ The App section is written for teams who want to deploy DeerFlow as a usable pro
 
 The Tutorials section is for hands-on, task-oriented learning.
 
-- [Tutorials](/docs/tutorials)
+- [Tutorials](./docs/tutorials)
 
 ### Reference
 
 The Reference section is for detailed lookup material, including configuration, runtime modes, APIs, and source-oriented mapping.
 
-- [Reference](/docs/reference)
+- [Reference](./docs/reference)
 
 ## Choose the right path
 
-- If you are **evaluating the project**, start with [Introduction](/docs/introduction).
-- If you are **building your own agent system**, start with [DeerFlow Harness](/docs/harness).
-- If you are **deploying DeerFlow for users**, start with [DeerFlow App](/docs/app).
-- If you want to **learn by doing**, go to [Tutorials](/docs/tutorials).
+- If you are **evaluating the project**, start with [Introduction](./docs/introduction).
+- If you are **building your own agent system**, start with [DeerFlow Harness](./docs/harness).
+- If you are **deploying DeerFlow for users**, start with [DeerFlow App](./docs/app).
+- If you want to **learn by doing**, go to [Tutorials](./docs/tutorials).

frontend/src/content/en/posts/releases/2_0_rc.mdx

@@ -0,0 +1,9 @@
+---
+title: DeerFlow 2.0 M1
+description: DeerFlow 2.0 M1 is officially in RC. Here's what you need to know.
+date: 2026-05-30
+tags:
+  - Release
+---
+
+## DeerFlow 2.0 M1 Release

frontend/src/content/zh/index.mdx

@@ -20,27 +20,27 @@ DeerFlow 是一个用于构建和运行 Agent 系统的框架。它提供了一
 
 先从概念概述开始。
 
-- [简介](/docs/introduction)
-- [为什么选择 DeerFlow](/docs/introduction/why-deerflow)
-- [Harness 与应用的区别](/docs/introduction/harness-vs-app)
+- [简介](./docs/introduction)
+- [为什么选择 DeerFlow](./docs/introduction/why-deerflow)
+- [Harness 与应用的区别](./docs/introduction/harness-vs-app)
 
 ### 如果你想基于 DeerFlow 进行开发
 
 从 Harness 章节开始。这条路径适合想将 DeerFlow 功能集成到自己系统中，或基于 DeerFlow 运行时构建自定义 Agent 产品的团队。
 
-- [DeerFlow Harness](/docs/harness)
-- [快速上手](/docs/harness/quick-start)
-- [配置](/docs/harness/configuration)
-- [自定义与扩展](/docs/harness/customization)
+- [DeerFlow Harness](./docs/harness)
+- [快速上手](./docs/harness/quick-start)
+- [配置](./docs/harness/configuration)
+- [自定义与扩展](./docs/harness/customization)
 
 ### 如果你想部署和使用 DeerFlow
 
 从应用章节开始。这条路径适合想将 DeerFlow 作为完整应用运行，并了解如何配置、运维和实际使用的团队。
 
-- [DeerFlow 应用](/docs/application)
-- [快速上手](/docs/application/quick-start)
-- [部署指南](/docs/application/deployment-guide)
-- [工作区使用](/docs/application/workspace-usage)
+- [DeerFlow 应用](./docs/application)
+- [快速上手](./docs/application/quick-start)
+- [部署指南](./docs/application/deployment-guide)
+- [工作区使用](./docs/application/workspace-usage)
 
 ## 文档结构
 

frontend/src/core/api/api-client.ts

@@ -3,6 +3,13 @@
 import { Client as LangGraphClient } from "@langchain/langgraph-sdk/client";
 
 import { getLangGraphBaseURL } from "../config";
+import { isStaticWebsiteOnly } from "../static-mode";
+import {
+  loadStaticDemoThread,
+  loadStaticDemoThreads,
+  staticDemoThreadState,
+} from "../threads/static-demo";
+import type { AgentThreadState } from "../threads/types";
 
 import { isStateChangingMethod, readCsrfCookie } from "./fetcher";
 import { sanitizeRunStreamOptions } from "./stream-mode";
@@ -32,6 +39,10 @@ function injectCsrfHeader(_url: URL, init: RequestInit): RequestInit {
 }
 
 function createCompatibleClient(isMock?: boolean): LangGraphClient {
+  if (isStaticWebsiteOnly() && !isMock) {
+    return createStaticClient();
+  }
+
   const apiUrl = getLangGraphBaseURL(isMock);
   console.log(`Creating API client with base URL: ${apiUrl}`);
   const client = new LangGraphClient({
@@ -58,6 +69,44 @@ function createCompatibleClient(isMock?: boolean): LangGraphClient {
   return client;
 }
 
+function createStaticClient(): LangGraphClient {
+  const apiUrl =
+    typeof window === "undefined"
+      ? "http://localhost:3000"
+      : window.location.origin;
+  const client = new LangGraphClient({ apiUrl });
+
+  client.threads.search = (async (query) => {
+    return loadStaticDemoThreads(query);
+  }) as typeof client.threads.search;
+
+  client.threads.get = (async (threadId) => {
+    return loadStaticDemoThread(threadId);
+  }) as typeof client.threads.get;
+
+  client.threads.getState = (async (threadId) => {
+    return staticDemoThreadState(await loadStaticDemoThread(threadId));
+  }) as typeof client.threads.getState;
+
+  client.threads.getHistory = (async (threadId) => {
+    return [staticDemoThreadState(await loadStaticDemoThread(threadId))];
+  }) as typeof client.threads.getHistory;
+
+  client.threads.update = (async (threadId) => {
+    return loadStaticDemoThread(threadId);
+  }) as typeof client.threads.update;
+
+  client.runs.list = (async () => []) as typeof client.runs.list;
+  client.runs.stream = async function* () {
+    /* empty */
+  } as typeof client.runs.stream;
+  client.runs.joinStream = async function* () {
+    /* empty */
+  } as typeof client.runs.joinStream;
+
+  return client as LangGraphClient<AgentThreadState>;
+}
+
 const _clients = new Map<string, LangGraphClient>();
 export function getAPIClient(isMock?: boolean): LangGraphClient {
   const cacheKey = isMock ? "mock" : "default";

frontend/src/core/artifacts/utils.ts

@@ -1,4 +1,5 @@
 import { getBackendBaseURL } from "../config";
+import { isStaticWebsiteOnly } from "../static-mode";
 import type { AgentThread } from "../threads";
 
 export function urlOfArtifact({
@@ -12,6 +13,9 @@ export function urlOfArtifact({
   download?: boolean;
   isMock?: boolean;
 }) {
+  if (isStaticWebsiteOnly()) {
+    return staticDemoArtifactURL({ filepath, threadId, download });
+  }
   if (isMock) {
     return `${getBackendBaseURL()}/mock/api/threads/${threadId}/artifacts${filepath}${download ? "?download=true" : ""}`;
   }
@@ -23,5 +27,21 @@ export function extractArtifactsFromThread(thread: AgentThread) {
 }
 
 export function resolveArtifactURL(absolutePath: string, threadId: string) {
+  if (isStaticWebsiteOnly()) {
+    return staticDemoArtifactURL({ filepath: absolutePath, threadId });
+  }
   return `${getBackendBaseURL()}/api/threads/${threadId}/artifacts${absolutePath}`;
 }
+
+function staticDemoArtifactURL({
+  filepath,
+  threadId,
+  download = false,
+}: {
+  filepath: string;
+  threadId: string;
+  download?: boolean;
+}) {
+  const demoPath = filepath.replace(/^\/mnt\//, "/");
+  return `${getBackendBaseURL()}/demo/threads/${threadId}${demoPath}${download ? "?download=true" : ""}`;
+}

frontend/src/core/auth/AuthProvider.tsx

@@ -10,6 +10,8 @@ import React, {
   type ReactNode,
 } from "react";
 
+import { isStaticWebsiteOnly } from "../static-mode";
+
 import { type User, buildLoginUrl } from "./types";
 
 // Re-export for consumers
@@ -46,6 +48,7 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
   const [isLoading, setIsLoading] = useState(false);
   const router = useRouter();
   const pathname = usePathname();
+  const staticMode = isStaticWebsiteOnly();
 
   const isAuthenticated = user !== null;
 
@@ -54,6 +57,8 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
    * Used when initialUser might be stale (e.g., after tab was inactive)
    */
   const refreshUser = useCallback(async () => {
+    if (staticMode) return;
+
     try {
       setIsLoading(true);
       const res = await fetch("/api/v1/auth/me", {
@@ -77,7 +82,7 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
     } finally {
       setIsLoading(false);
     }
-  }, [pathname, router]);
+  }, [staticMode, pathname, router]);
 
   /**
    * Logout - call FastAPI logout endpoint and clear local state
@@ -87,6 +92,11 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
     // Immediately clear local state to prevent UI flicker
     setUser(null);
 
+    if (staticMode) {
+      router.push("/");
+      return;
+    }
+
     try {
       await fetch("/api/v1/auth/logout", {
         method: "POST",
@@ -99,7 +109,7 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
 
     // Redirect to home page
     router.push("/");
-  }, [router]);
+  }, [staticMode, router]);
 
   /**
    * Handle visibility change - refresh user when tab becomes visible again.
@@ -108,6 +118,8 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
   const lastCheckRef = React.useRef(0);
 
   useEffect(() => {
+    if (staticMode) return;
+
     const handleVisibilityChange = () => {
       if (document.visibilityState !== "visible" || user === null) return;
       const now = Date.now();
@@ -120,7 +132,7 @@ export function AuthProvider({ children, initialUser }: AuthProviderProps) {
     return () => {
       document.removeEventListener("visibilitychange", handleVisibilityChange);
     };
-  }, [user, refreshUser]);
+  }, [staticMode, user, refreshUser]);
 
   const value: AuthContextType = {
     user,
@@ -155,6 +167,8 @@ export function useRequireAuth(): AuthContextType {
   const pathname = usePathname();
 
   useEffect(() => {
+    if (isStaticWebsiteOnly()) return;
+
     // Only redirect if we're sure user is not authenticated (not just loading)
     if (!auth.isLoading && !auth.isAuthenticated) {
       router.push(buildLoginUrl(pathname || "/workspace"));

frontend/src/core/auth/server.ts

@@ -1,6 +1,9 @@
 import { cookies } from "next/headers";
 
+import { isStaticWebsiteOnly } from "../static-mode";
+
 import { getGatewayConfig } from "./gateway-config";
+import { STATIC_WEBSITE_USER } from "./static-user";
 import { type AuthResult, userSchema } from "./types";
 
 const SSR_AUTH_TIMEOUT_MS = 5_000;
@@ -10,6 +13,13 @@ const SSR_AUTH_TIMEOUT_MS = 5_000;
  * Returns a tagged AuthResult — callers use exhaustive switch, no try/catch.
  */
 export async function getServerSideUser(): Promise<AuthResult> {
+  if (isStaticWebsiteOnly()) {
+    return {
+      tag: "authenticated",
+      user: STATIC_WEBSITE_USER,
+    };
+  }
+
   if (process.env.DEER_FLOW_AUTH_DISABLED === "1") {
     return {
       tag: "authenticated",

frontend/src/core/auth/static-user.ts

@@ -0,0 +1,8 @@
+import type { User } from "./types";
+
+export const STATIC_WEBSITE_USER: User = {
+  id: "static-website-user",
+  email: "static@example.local",
+  system_role: "admin",
+  needs_setup: false,
+};

frontend/src/core/models/api.ts

@@ -1,8 +1,18 @@
 import { getBackendBaseURL } from "../config";
+import { isStaticWebsiteOnly } from "../static-mode";
 
 import type { ModelsResponse } from "./types";
 
+const STATIC_MODELS_RESPONSE: ModelsResponse = {
+  models: [],
+  token_usage: { enabled: false },
+};
+
 export async function loadModels(): Promise<ModelsResponse> {
+  if (isStaticWebsiteOnly()) {
+    return STATIC_MODELS_RESPONSE;
+  }
+
   const res = await fetch(`${getBackendBaseURL()}/api/models`);
   const data = (await res.json()) as Partial<ModelsResponse>;
   return {

frontend/src/core/static-mode.ts

@@ -0,0 +1,5 @@
+import { env } from "@/env";
+
+export function isStaticWebsiteOnly() {
+  return env.NEXT_PUBLIC_STATIC_WEBSITE_ONLY === "true";
+}

frontend/src/core/threads/static-demo.ts

@@ -0,0 +1,87 @@
+import type { ThreadState } from "@langchain/langgraph-sdk";
+import type { ThreadsClient } from "@langchain/langgraph-sdk/client";
+
+import type { AgentThread, AgentThreadState } from "./types";
+
+export const DEMO_THREAD_IDS = [
+  "21cfea46-34bd-4aa6-9e1f-3009452fbeb9",
+  "3823e443-4e2b-4679-b496-a9506eae462b",
+  "4f3e55ee-f853-43db-bfb3-7d1a411f03cb",
+  "5aa47db1-d0cb-4eb9-aea5-3dac1b371c5a",
+  "7cfa5f8f-a2f8-47ad-acbd-da7137baf990",
+  "7f9dc56c-e49c-4671-a3d2-c492ff4dce0c",
+  "90040b36-7eba-4b97-ba89-02c3ad47a8b9",
+  "ad76c455-5bf9-4335-8517-fc03834ab828",
+  "b83fbb2a-4e36-4d82-9de0-7b2a02c2092a",
+  "c02bb4d5-4202-490e-ae8f-ff4864fc0d2e",
+  "d3e5adaf-084c-4dd5-9d29-94f1d6bccd98",
+  "f4125791-0128-402a-8ca9-50e0947557e4",
+  "fe3f7974-1bcb-4a01-a950-79673baafefd",
+] as const;
+
+export type ThreadSearchParams = NonNullable<
+  Parameters<ThreadsClient["search"]>[0]
+>;
+
+export async function loadStaticDemoThreads(
+  params: ThreadSearchParams = {},
+): Promise<AgentThread[]> {
+  const threads = await Promise.all(
+    DEMO_THREAD_IDS.map((threadId) => loadStaticDemoThread(threadId)),
+  );
+
+  const sortBy = params.sortBy ?? "updated_at";
+  const sortOrder = params.sortOrder ?? "desc";
+  const sortedThreads = [...threads].sort((a, b) => {
+    const aTimestamp = (a as unknown as Record<string, unknown>)[sortBy];
+    const bTimestamp = (b as unknown as Record<string, unknown>)[sortBy];
+    const aParsed = typeof aTimestamp === "string" ? Date.parse(aTimestamp) : 0;
+    const bParsed = typeof bTimestamp === "string" ? Date.parse(bTimestamp) : 0;
+    const aValue = Number.isNaN(aParsed) ? 0 : aParsed;
+    const bValue = Number.isNaN(bParsed) ? 0 : bParsed;
+    return sortOrder === "asc" ? aValue - bValue : bValue - aValue;
+  });
+
+  const offset = Math.max(0, Math.floor(params.offset ?? 0));
+  const limit =
+    typeof params.limit === "number"
+      ? Math.max(0, Math.floor(params.limit))
+      : sortedThreads.length;
+  return sortedThreads.slice(offset, offset + limit);
+}
+
+export async function loadStaticDemoThread(
+  threadId: string,
+): Promise<AgentThread> {
+  const response = await globalThis.fetch(
+    `/demo/threads/${encodeURIComponent(threadId)}/thread.json`,
+  );
+  if (!response.ok) {
+    throw new Error(`Failed to load demo thread ${threadId}`);
+  }
+  const thread = (await response.json()) as AgentThread;
+  return {
+    ...thread,
+    thread_id: threadId,
+    updated_at: thread.updated_at ?? thread.created_at,
+  };
+}
+
+export function staticDemoThreadState(
+  thread: AgentThread,
+): ThreadState<AgentThreadState> {
+  return {
+    values: thread.values,
+    next: [],
+    checkpoint: {
+      thread_id: thread.thread_id,
+      checkpoint_ns: "",
+      checkpoint_id: null,
+      checkpoint_map: null,
+    },
+    metadata: thread.metadata ?? null,
+    created_at: thread.updated_at ?? thread.created_at ?? null,
+    parent_checkpoint: null,
+    tasks: [],
+  };
+}

frontend/tests/unit/core/artifacts/utils.test.ts

@@ -0,0 +1,69 @@
+import { afterEach, beforeEach, describe, expect, test, vi } from "vitest";
+
+const ENV_KEYS = [
+  "NEXT_PUBLIC_BACKEND_BASE_URL",
+  "NEXT_PUBLIC_STATIC_WEBSITE_ONLY",
+] as const;
+
+type EnvSnapshot = Partial<
+  Record<(typeof ENV_KEYS)[number], string | undefined>
+>;
+
+function snapshotEnv(): EnvSnapshot {
+  const snapshot: EnvSnapshot = {};
+  for (const key of ENV_KEYS) {
+    snapshot[key] = process.env[key];
+  }
+  return snapshot;
+}
+
+function setEnv(key: (typeof ENV_KEYS)[number], value: string | undefined) {
+  const env = process.env as Record<string, string | undefined>;
+  if (value === undefined) {
+    delete env[key];
+  } else {
+    env[key] = value;
+  }
+}
+
+function restoreEnv(snapshot: EnvSnapshot) {
+  for (const key of ENV_KEYS) {
+    setEnv(key, snapshot[key]);
+  }
+}
+
+async function loadFreshArtifactUtils() {
+  vi.resetModules();
+  return await import("@/core/artifacts/utils");
+}
+
+describe("artifact URL helpers", () => {
+  let saved: EnvSnapshot;
+
+  beforeEach(() => {
+    saved = snapshotEnv();
+    setEnv("NEXT_PUBLIC_BACKEND_BASE_URL", undefined);
+    setEnv("NEXT_PUBLIC_STATIC_WEBSITE_ONLY", undefined);
+  });
+
+  afterEach(() => {
+    restoreEnv(saved);
+  });
+
+  test("maps static demo artifact paths to bundled public files", async () => {
+    setEnv("NEXT_PUBLIC_STATIC_WEBSITE_ONLY", "true");
+
+    const { resolveArtifactURL, urlOfArtifact } =
+      await loadFreshArtifactUtils();
+
+    expect(
+      urlOfArtifact({
+        filepath: "/mnt/user-data/outputs/index.html",
+        threadId: "thread-1",
+      }),
+    ).toBe("/demo/threads/thread-1/user-data/outputs/index.html");
+    expect(
+      resolveArtifactURL("/mnt/user-data/outputs/style.css", "thread-1"),
+    ).toBe("/demo/threads/thread-1/user-data/outputs/style.css");
+  });
+});

frontend/tests/unit/core/auth/server.test.ts

@@ -0,0 +1,77 @@
+import { afterEach, beforeEach, describe, expect, test, vi } from "vitest";
+
+import { STATIC_WEBSITE_USER } from "@/core/auth/static-user";
+
+vi.mock("next/headers", () => ({
+  cookies: vi.fn(() => {
+    throw new Error("cookies should not be read in static website mode");
+  }),
+}));
+
+const ENV_KEYS = [
+  "DEER_FLOW_AUTH_DISABLED",
+  "NEXT_PUBLIC_STATIC_WEBSITE_ONLY",
+] as const;
+
+type EnvSnapshot = Partial<
+  Record<(typeof ENV_KEYS)[number], string | undefined>
+>;
+
+function snapshotEnv(): EnvSnapshot {
+  const snapshot: EnvSnapshot = {};
+  for (const key of ENV_KEYS) {
+    snapshot[key] = process.env[key];
+  }
+  return snapshot;
+}
+
+function setEnv(key: (typeof ENV_KEYS)[number], value: string | undefined) {
+  const env = process.env as Record<string, string | undefined>;
+  if (value === undefined) {
+    delete env[key];
+  } else {
+    env[key] = value;
+  }
+}
+
+function restoreEnv(snapshot: EnvSnapshot) {
+  for (const key of ENV_KEYS) {
+    setEnv(key, snapshot[key]);
+  }
+}
+
+async function loadFreshServerAuth() {
+  vi.resetModules();
+  return await import("@/core/auth/server");
+}
+
+describe("getServerSideUser", () => {
+  let saved: EnvSnapshot;
+
+  beforeEach(() => {
+    saved = snapshotEnv();
+    setEnv("DEER_FLOW_AUTH_DISABLED", undefined);
+    setEnv("NEXT_PUBLIC_STATIC_WEBSITE_ONLY", undefined);
+  });
+
+  afterEach(() => {
+    restoreEnv(saved);
+    vi.unstubAllGlobals();
+  });
+
+  test("bypasses gateway auth in static website mode", async () => {
+    setEnv("NEXT_PUBLIC_STATIC_WEBSITE_ONLY", "true");
+    const fetchSpy = vi.fn(() => {
+      throw new Error("fetch should not be called in static website mode");
+    });
+    vi.stubGlobal("fetch", fetchSpy);
+
+    const { getServerSideUser } = await loadFreshServerAuth();
+
+    await expect(getServerSideUser()).resolves.toEqual({
+      tag: "authenticated",
+      user: STATIC_WEBSITE_USER,
+    });
+    expect(fetchSpy).not.toHaveBeenCalled();
+  });
+});