✨ Feat: support user to configurate model concurrency limit and timeout seconds

backend/agents/create_agent_info.py

@@ -247,7 +247,9 @@ async def create_model_config_list(tenant_id):
                             ),
                         url=record["base_url"],
                         ssl_verify=record.get("ssl_verify", True),
-                        model_factory=record.get("model_factory")))
+                        model_factory=record.get("model_factory"),
+                        timeout_seconds=record.get("timeout_seconds"),
+                        concurrency_limit=record.get("concurrency_limit")))
     # fit for old version, main_model and sub_model use default model
     main_model_config = tenant_config_manager.get_model_config(
         key=MODEL_CONFIG_MAPPING["llm"], tenant_id=tenant_id)
@@ -258,15 +260,19 @@ async def create_model_config_list(tenant_id):
                         "model_name") else "",
                     url=main_model_config.get("base_url", ""),
                     ssl_verify=main_model_config.get("ssl_verify", True),
-                    model_factory=main_model_config.get("model_factory")))
+                    model_factory=main_model_config.get("model_factory"),
+                    timeout_seconds=main_model_config.get("timeout_seconds"),
+                    concurrency_limit=main_model_config.get("concurrency_limit")))
     model_list.append(
         ModelConfig(cite_name="sub_model",
                     api_key=main_model_config.get("api_key", ""),
                     model_name=get_model_name_from_config(main_model_config) if main_model_config.get(
                         "model_name") else "",
                     url=main_model_config.get("base_url", ""),
                     ssl_verify=main_model_config.get("ssl_verify", True),
-                    model_factory=main_model_config.get("model_factory")))
+                    model_factory=main_model_config.get("model_factory"),
+                    timeout_seconds=main_model_config.get("timeout_seconds"),
+                    concurrency_limit=main_model_config.get("concurrency_limit")))
 
     return model_list
 

backend/apps/config_app.py

@@ -30,6 +30,7 @@
 from apps.a2a_client_app import router as a2a_client_router
 from apps.monitoring_app import router as monitoring_router
 from apps.a2a_server_app import router as a2a_server_router
+from apps.haotian_app import router as haotian_router
 from consts.const import IS_SPEED_MODE
 from services.prompt_template_service import sync_system_default_prompt_template
 
@@ -84,3 +85,4 @@ async def sync_default_prompt_template_on_startup():
 app.include_router(invitation_router)
 app.include_router(a2a_client_router)
 app.include_router(a2a_server_router)
+app.include_router(haotian_router)

backend/consts/agent_unavailable_reasons.py

@@ -0,0 +1,43 @@
+"""
+Agent Unavailable Reason Constants
+
+Centralized definition of all possible reasons why an agent may be unavailable.
+These values are returned to the frontend via the 'unavailable_reasons' field.
+"""
+
+
+class AgentUnavailableReason:
+    """Reason codes for agent unavailability."""
+
+    # Identity conflicts
+    DUPLICATE_NAME = "duplicate_name"
+    DUPLICATE_DISPLAY_NAME = "duplicate_display_name"
+
+    # Model issues
+    MODEL_NOT_CONFIGURED = "model_not_configured"
+    MODEL_UNAVAILABLE = "model_unavailable"
+
+    # Tool issues
+    TOOL_UNAVAILABLE = "tool_unavailable"
+    ALL_TOOLS_DISABLED = "all_tools_disabled"
+
+    # Agent issues
+    AGENT_NOT_FOUND = "agent_not_found"
+
+    @classmethod
+    def all_reasons(cls) -> list[str]:
+        """Return all defined unavailable reason codes."""
+        return [
+            cls.DUPLICATE_NAME,
+            cls.DUPLICATE_DISPLAY_NAME,
+            cls.MODEL_NOT_CONFIGURED,
+            cls.MODEL_UNAVAILABLE,
+            cls.TOOL_UNAVAILABLE,
+            cls.ALL_TOOLS_DISABLED,
+            cls.AGENT_NOT_FOUND,
+        ]
+
+    @classmethod
+    def is_valid_reason(cls, reason: str) -> bool:
+        """Check if a reason string is a valid reason code."""
+        return reason in cls.all_reasons()

backend/consts/model.py

@@ -130,6 +130,8 @@ class ModelRequest(BaseModel):
     # STT specific fields
     model_appid: Optional[str] = None
     access_token: Optional[str] = None
+    timeout_seconds: Optional[int] = None
+    concurrency_limit: Optional[int] = None
 
 
 class ProviderModelRequest(BaseModel):
@@ -830,6 +832,8 @@ class ManageTenantModelCreateRequest(BaseModel):
     # STT specific fields
     model_appid: Optional[str] = Field(None, description="Application ID for STT models (e.g., Volcano Engine)")
     access_token: Optional[str] = Field(None, description="Access token for STT models (e.g., Volcano Engine)")
+    timeout_seconds: Optional[int] = Field(None, description="Request timeout in seconds")
+    concurrency_limit: Optional[int] = Field(None, description="Maximum concurrent requests for this model")
 
 
 class ManageTenantModelUpdateRequest(BaseModel):
@@ -850,6 +854,8 @@ class ManageTenantModelUpdateRequest(BaseModel):
     # STT specific fields
     model_appid: Optional[str] = Field(None, description="Application ID for STT models")
     access_token: Optional[str] = Field(None, description="Access token for STT models")
+    timeout_seconds: Optional[int] = Field(None, description="Request timeout in seconds")
+    concurrency_limit: Optional[int] = Field(None, description="Maximum concurrent requests for this model")
 
 
 class ManageTenantModelDeleteRequest(BaseModel):

backend/database/db_models.py

@@ -182,6 +182,10 @@ class ModelRecord(TableBase):
         String(100), doc="Application ID for model authentication (used by some STT/TTS providers like Volcano Engine)")
     access_token = Column(
         String(100), doc="Access token for model authentication (used by some STT/TTS providers like Volcano Engine)")
+    timeout_seconds = Column(
+        Integer, doc="Request timeout in seconds for this model. Default is 120 seconds.")
+    concurrency_limit = Column(
+        Integer, doc="Maximum concurrent requests for this model. Default is null (unlimited).")
 
 
 class ModelMonitoringRecord(SimpleTableBase):

backend/database/model_management_db.py

@@ -1,3 +1,4 @@
+import logging
 from typing import Any, Dict, List, Optional
 
 from sqlalchemy import and_, desc, func, insert, select, update
@@ -7,6 +8,8 @@
 from .db_models import ModelRecord
 from .utils import add_creation_tracking, add_update_tracking
 
+logger = logging.getLogger("database.model_management_db")
+
 
 def create_model_record(model_data: Dict[str, Any], user_id: str, tenant_id: str) -> bool:
     """

backend/services/agent_service.py

@@ -20,6 +20,7 @@
 from consts.const import MEMORY_SEARCH_START_MSG, MEMORY_SEARCH_DONE_MSG, MEMORY_SEARCH_FAIL_MSG, TOOL_TYPE_MAPPING, \
     LANGUAGE, MESSAGE_ROLE, MODEL_CONFIG_MAPPING, CAN_EDIT_ALL_USER_ROLES, PERMISSION_EDIT, PERMISSION_READ, PERMISSION_PRIVATE
 from consts.exceptions import MemoryPreparationException
+from consts.agent_unavailable_reasons import AgentUnavailableReason
 from consts.model import (
     AgentInfoRequest,
     AgentRequest,
@@ -1585,8 +1586,8 @@ def _mark_duplicates(groups: dict[str, list[dict]], reason_key: str) -> None:
             for duplicate_entry in sorted_entries[1:]:
                 duplicate_entry["unavailable_reasons"].append(reason_key)
 
-    _mark_duplicates(name_groups, "duplicate_name")
-    _mark_duplicates(display_name_groups, "duplicate_display_name")
+    _mark_duplicates(name_groups, AgentUnavailableReason.DUPLICATE_NAME)
+    _mark_duplicates(display_name_groups, AgentUnavailableReason.DUPLICATE_DISPLAY_NAME)
 
 
 def _collect_model_availability_reasons(agent: dict, tenant_id: str, model_cache: Dict[int, Optional[dict]]) -> list[str]:
@@ -1598,7 +1599,7 @@ def _collect_model_availability_reasons(agent: dict, tenant_id: str, model_cache
         model_id=agent.get("model_id"),
         tenant_id=tenant_id,
         model_cache=model_cache,
-        reason_key="model_unavailable"
+        reason_key=AgentUnavailableReason.MODEL_UNAVAILABLE
     ))
 
     return reasons
@@ -1656,15 +1657,15 @@ def check_agent_availability(
         agent_info = search_agent_info_by_agent_id(agent_id, tenant_id)
 
     if not agent_info:
-        return False, ["agent_not_found"]
+        return False, [AgentUnavailableReason.AGENT_NOT_FOUND]
 
     # Check tool availability
     tool_info = search_tools_for_sub_agent(agent_id=agent_id, tenant_id=tenant_id)
     tool_id_list = [tool["tool_id"] for tool in tool_info if tool.get("tool_id") is not None]
     if tool_id_list:
         tool_statuses = check_tool_is_available(tool_id_list)
         if not all(tool_statuses):
-            unavailable_reasons.append("tool_unavailable")
+            unavailable_reasons.append(AgentUnavailableReason.TOOL_UNAVAILABLE)
 
     # Check model availability
     model_reasons = _collect_model_availability_reasons(

backend/services/agent_version_service.py

@@ -33,6 +33,7 @@
 )
 from database.model_management_db import get_model_by_model_id
 from utils.str_utils import convert_string_to_list
+from consts.agent_unavailable_reasons import AgentUnavailableReason
 
 logger = logging.getLogger("agent_version_service")
 
@@ -337,21 +338,18 @@ def _check_version_snapshot_availability(
 
     # Check if agent info exists
     if not agent_info:
-        return False, ["agent_not_found"]
+        return False, [AgentUnavailableReason.AGENT_NOT_FOUND]
 
     # Check model availability
     model_id = agent_info.get('model_id')
     if model_id is None or model_id == 0:
-        unavailable_reasons.append("model_not_configured")
+        unavailable_reasons.append(AgentUnavailableReason.MODEL_NOT_CONFIGURED)
 
-    # Check tools availability
-    if not tool_instances:
-        unavailable_reasons.append("no_tools")
-    else:
-        # Check if at least one tool is enabled
+    # Check tools availability (only when tools are configured)
+    if tool_instances:
         has_enabled_tool = any(t.get('enabled', True) for t in tool_instances)
         if not has_enabled_tool:
-            unavailable_reasons.append("all_tools_disabled")
+            unavailable_reasons.append(AgentUnavailableReason.ALL_TOOLS_DISABLED)
 
     return len(unavailable_reasons) == 0, unavailable_reasons
 

backend/services/conversation_management_service.py

@@ -248,6 +248,8 @@ def call_llm_for_title(question: str, tenant_id: str, language: str = LANGUAGE["
     display_name = model_config.get("display_name", "") if model_config else ""
     set_monitoring_operation("title_generation", display_name=display_name or None)
 
+    timeout_seconds = model_config.get("timeout_seconds") if model_config else None
+
     # Create OpenAIModel instance
     llm = OpenAIModel(
         model_id=get_model_name_from_config(model_config) if model_config.get("model_name") else "",
@@ -256,7 +258,9 @@ def call_llm_for_title(question: str, tenant_id: str, language: str = LANGUAGE["
         temperature=0.7,
         top_p=0.95,
         model_factory=model_config.get("model_factory", None),
-        ssl_verify=model_config.get("ssl_verify", True)
+        ssl_verify=model_config.get("ssl_verify", True),
+        timeout_seconds=timeout_seconds,
+        stream=False,
     )
 
     # Build messages - use new template variable 'question' instead of 'content'

backend/services/file_management_service.py

@@ -352,13 +352,15 @@ def get_llm_model(tenant_id: str):
     # Get the tenant config
     main_model_config = tenant_config_manager.get_model_config(
         key=MODEL_CONFIG_MAPPING["llm"], tenant_id=tenant_id)
+    timeout_seconds = main_model_config.get("timeout_seconds") if main_model_config else None
     long_text_to_text_model = OpenAILongContextModel(
         observer=MessageObserver(),
         model_id=get_model_name_from_config(main_model_config),
         api_base=main_model_config.get("base_url"),
         api_key=main_model_config.get("api_key"),
         max_context_tokens=main_model_config.get("max_tokens"),
         ssl_verify=main_model_config.get("ssl_verify", True),
+        timeout_seconds=timeout_seconds,
     )
     return long_text_to_text_model
 

backend/services/haotian_service.py

@@ -11,6 +11,8 @@
 
 logger = logging.getLogger("haotian_service")
 
+_DEFAULT_KNOWLEDGE_BASE_ID = "a8d68fbf-bd6e-5461-a9d1-cf1bb3522e38"
+
 
 def _normalize_list_payload(raw: Dict[str, Any]) -> Dict[str, Any]:
     """
@@ -24,7 +26,7 @@ def _normalize_list_payload(raw: Dict[str, Any]) -> Dict[str, Any]:
       ]
     }
 
-    This function also filters out knowledge sets with name == "Public".
+    When dify_dataset_id is "null", it is replaced with the default ID.
     """
     knowledge_sets = raw.get("knowledge_sets", [])
     if not isinstance(knowledge_sets, list):
@@ -35,7 +37,7 @@ def _normalize_list_payload(raw: Dict[str, Any]) -> Dict[str, Any]:
         if not isinstance(ks, dict):
             continue
         set_name = str(ks.get("name", "") or "").strip()
-        if not set_name or set_name == "Public":
+        if not set_name:
             continue
 
         bases = ks.get("knowledge_bases", [])
@@ -48,15 +50,18 @@ def _normalize_list_payload(raw: Dict[str, Any]) -> Dict[str, Any]:
                 continue
             dataset_id = str(kb.get("dify_dataset_id", "") or "").strip()
             kb_name = str(kb.get("name", "") or "").strip()
-            if not dataset_id or not kb_name:
+            if not kb_name:
                 continue
+            if dataset_id == "null" or not dataset_id:
+                dataset_id = _DEFAULT_KNOWLEDGE_BASE_ID
             normalized_bases.append(
                 {"dify_dataset_id": dataset_id, "name": kb_name}
             )
 
-        normalized_sets.append(
-            {"name": set_name, "knowledge_bases": normalized_bases}
-        )
+        if normalized_bases:
+            normalized_sets.append(
+                {"name": set_name, "knowledge_bases": normalized_bases}
+            )
 
     return {"knowledge_sets": normalized_sets}
 
@@ -77,7 +82,7 @@ async def fetch_haotian_knowledge_sets_impl(
         )
 
     headers = {"Authorization": external_authorization}
-    async with httpx.AsyncClient(timeout=timeout_s, follow_redirects=True) as client:
+    async with httpx.AsyncClient(timeout=timeout_s, follow_redirects=True, trust_env=False) as client:
         resp = await client.get(list_url, headers=headers)
         if resp.status_code >= 400:
             raise RuntimeError(