union-app-chat-stream 1.0.3 → 1.0.4

package/PROJECT_OVERVIEW.md

@@ -162,6 +162,22 @@ Deployment files are organized under `deploy/`:
   treat knowledge-base "related functions" as routing hints unless code
   explicitly promotes them to executable tools.
 
+## Streaming Tool-Call Contract
+
+- External tool calls may run for minutes or longer because some tools query
+  large data platforms such as Hive. Do not treat a long-running tool call as a
+  timeout by default.
+- During tool execution, `/chatstream/v1/chat/stream` emits SSE `heartbeat`
+  events every `TOOL_CALL_HEARTBEAT_INTERVAL` seconds. The payload is the normal
+  chat response JSON with a `heartbeat` object, including `type`, `tool`,
+  `elapsedSeconds`, and `message`.
+- Frontends should listen for both `message` and `heartbeat` SSE events.
+  Heartbeat events mean the stream is alive and the current tool is still
+  running; they are not model content and should not be rendered as answer text.
+- Tool execution failures return a `message` event with `finish_reason="error"`
+  and `errorMsg`. If tool-calling reaches `TOOLS_MAX_ROUNDS`, return
+  `finish_reason="error"` and `errorMsg="工具调用轮数达到上限（N轮）"`.
+
 ## Validation
 
 Use the smallest reliable checks that cover the change:

package/app/.env

@@ -1,63 +1,63 @@
 # Flask and request authentication
-SECRET_KEY=<MASKED_SECRET>
+SECRET_KEY=__MASKED_FOR_NPM__
 
 # External API endpoints
-GET_USE_INFO_URL=<MASKED_URL>
-GET_ORG_INFO_URL=<MASKED_URL>
-GET_JIRA_INFO_URL=<MASKED_URL>
-GET_BIGDATA_URL=<MASKED_URL>
-GET_UNION_BASE_URL=<MASKED_URL>
+GET_USE_INFO_URL=__MASKED_FOR_NPM__
+GET_ORG_INFO_URL=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL=__MASKED_FOR_NPM__
+GET_BIGDATA_URL=__MASKED_FOR_NPM__
+GET_UNION_BASE_URL=__MASKED_FOR_NPM__
 
 # Legacy external API tokens
-GET_ORG_INFO_URL_TOKEN=<MASKED_TOKEN>
-GET_JIRA_INFO_URL_TOKEN=<MASKED_TOKEN>
+GET_ORG_INFO_URL_TOKEN=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL_TOKEN=__MASKED_FOR_NPM__
 
 # Authorization and logging
-PERMISSIONS=
-LOG_LEVEL=INFO
-CONSOLE_STDOUT=TRUE
-LOG_DIR=/data/appLogs
+PERMISSIONS=__MASKED_FOR_NPM__
+LOG_LEVEL=__MASKED_FOR_NPM__
+CONSOLE_STDOUT=__MASKED_FOR_NPM__
+LOG_DIR=__MASKED_FOR_NPM__
 
 # Runtime environment and JWT
-FLASK_ENV=dev
-JWT_SECRET_KEY=<MASKED_SECRET>
-JWT_EXPIRATION_SECOND=900
-JWT_RENEW_SECOND=700
+FLASK_ENV=__MASKED_FOR_NPM__
+JWT_SECRET_KEY=__MASKED_FOR_NPM__
+JWT_EXPIRATION_SECOND=__MASKED_FOR_NPM__
+JWT_RENEW_SECOND=__MASKED_FOR_NPM__
 
 # LLM provider
-LLM_URL=<MASKED_URL>
-LLM_KEY=<MASKED_KEY>
-LLM_MODEL=GLM-4.7-Flash
-LLM_MAX_TOKENS=4096
-LLM_TEMPERATURE=0.7
-LLM_TOP_P=0.9
+LLM_URL=__MASKED_FOR_NPM__
+LLM_KEY=__MASKED_FOR_NPM__
+LLM_MODEL=__MASKED_FOR_NPM__
+LLM_MAX_TOKENS=__MASKED_FOR_NPM__
+LLM_TEMPERATURE=__MASKED_FOR_NPM__
+LLM_TOP_P=__MASKED_FOR_NPM__
 
 # Chat behavior
-SYSTEM_PROMPT=<MASKED_BUSINESS_VALUE>
+SYSTEM_PROMPT=__MASKED_FOR_NPM__
 
 # Business filter
-FILTER_ENABLED=false
-FILTER_ALLOWED_KEYWORDS=<MASKED_BUSINESS_VALUE>
-FILTER_REJECTION_MESSAGE=<MASKED_BUSINESS_VALUE>
+FILTER_ENABLED=__MASKED_FOR_NPM__
+FILTER_ALLOWED_KEYWORDS=__MASKED_FOR_NPM__
+FILTER_REJECTION_MESSAGE=__MASKED_FOR_NPM__
 
 # Tool and conversation settings
-TOOLS_MAX_ROUNDS=5
-CONVERSATION_MAX_HISTORY=20
-CONVERSATION_TTL=3600
+TOOLS_MAX_ROUNDS=__MASKED_FOR_NPM__
+CONVERSATION_MAX_HISTORY=__MASKED_FOR_NPM__
+CONVERSATION_TTL=__MASKED_FOR_NPM__
 
 # RAG settings
-RAG_ENABLED=true
-RAG_KNOWLEDGE_DIR=knowledge
-RAG_PERSIST_DIR=.chroma
-RAG_COLLECTION=ops_knowledge
-RAG_EMBEDDING_MODEL=embedding-3
-RAG_EMBEDDING_MAX_CHARS=6000
-RAG_EMBEDDING_BATCH_SIZE=8
-RAG_TOP_K=5
-RAG_SEMANTIC_CANDIDATE_K=40
-RAG_CONTEXT_K=8
-RAG_EXACT_CONTEXT_K=3
-RAG_EXACT_PER_FILE_CONTEXT_K=1
-RAG_PER_FILE_CONTEXT_K=2
-RAG_CHUNK_SIZE=1200
-RAG_REBUILD_ON_STARTUP=false
+RAG_ENABLED=__MASKED_FOR_NPM__
+RAG_KNOWLEDGE_DIR=__MASKED_FOR_NPM__
+RAG_PERSIST_DIR=__MASKED_FOR_NPM__
+RAG_COLLECTION=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MODEL=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MAX_CHARS=__MASKED_FOR_NPM__
+RAG_EMBEDDING_BATCH_SIZE=__MASKED_FOR_NPM__
+RAG_TOP_K=__MASKED_FOR_NPM__
+RAG_SEMANTIC_CANDIDATE_K=__MASKED_FOR_NPM__
+RAG_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_CHUNK_SIZE=__MASKED_FOR_NPM__
+RAG_REBUILD_ON_STARTUP=__MASKED_FOR_NPM__

package/app/.env.dev

@@ -1,63 +1,63 @@
 # Flask and request authentication
-SECRET_KEY=<MASKED_SECRET>
+SECRET_KEY=__MASKED_FOR_NPM__
 
 # External API endpoints
-GET_USE_INFO_URL=<MASKED_URL>
-GET_ORG_INFO_URL=<MASKED_URL>
-GET_JIRA_INFO_URL=<MASKED_URL>
-GET_BIGDATA_URL=<MASKED_URL>
-GET_UNION_BASE_URL=<MASKED_URL>
+GET_USE_INFO_URL=__MASKED_FOR_NPM__
+GET_ORG_INFO_URL=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL=__MASKED_FOR_NPM__
+GET_BIGDATA_URL=__MASKED_FOR_NPM__
+GET_UNION_BASE_URL=__MASKED_FOR_NPM__
 
 # Legacy external API tokens
-GET_ORG_INFO_URL_TOKEN=<MASKED_TOKEN>
-GET_JIRA_INFO_URL_TOKEN=<MASKED_TOKEN>
+GET_ORG_INFO_URL_TOKEN=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL_TOKEN=__MASKED_FOR_NPM__
 
 # Authorization and logging
-PERMISSIONS=
-LOG_LEVEL=INFO
-CONSOLE_STDOUT=TRUE
-LOG_DIR=/Users/simon/code/union-py-app/data/appLogs
+PERMISSIONS=__MASKED_FOR_NPM__
+LOG_LEVEL=__MASKED_FOR_NPM__
+CONSOLE_STDOUT=__MASKED_FOR_NPM__
+LOG_DIR=__MASKED_FOR_NPM__
 
 # Runtime environment and JWT
-FLASK_ENV=dev
-JWT_SECRET_KEY=<MASKED_SECRET>
-JWT_EXPIRATION_SECOND=900
-JWT_RENEW_SECOND=700
+FLASK_ENV=__MASKED_FOR_NPM__
+JWT_SECRET_KEY=__MASKED_FOR_NPM__
+JWT_EXPIRATION_SECOND=__MASKED_FOR_NPM__
+JWT_RENEW_SECOND=__MASKED_FOR_NPM__
 
 # LLM provider
-LLM_URL=<MASKED_URL>
-LLM_KEY=<MASKED_KEY>
-LLM_MODEL=GLM-4.7-Flash
-LLM_MAX_TOKENS=4096
-LLM_TEMPERATURE=0.7
-LLM_TOP_P=0.9
+LLM_URL=__MASKED_FOR_NPM__
+LLM_KEY=__MASKED_FOR_NPM__
+LLM_MODEL=__MASKED_FOR_NPM__
+LLM_MAX_TOKENS=__MASKED_FOR_NPM__
+LLM_TEMPERATURE=__MASKED_FOR_NPM__
+LLM_TOP_P=__MASKED_FOR_NPM__
 
 # Chat behavior
-SYSTEM_PROMPT=<MASKED_BUSINESS_VALUE>
+SYSTEM_PROMPT=__MASKED_FOR_NPM__
 
 # Business filter
-FILTER_ENABLED=false
-FILTER_ALLOWED_KEYWORDS=<MASKED_BUSINESS_VALUE>
-FILTER_REJECTION_MESSAGE=<MASKED_BUSINESS_VALUE>
+FILTER_ENABLED=__MASKED_FOR_NPM__
+FILTER_ALLOWED_KEYWORDS=__MASKED_FOR_NPM__
+FILTER_REJECTION_MESSAGE=__MASKED_FOR_NPM__
 
 # Tool and conversation settings
-TOOLS_MAX_ROUNDS=5
-CONVERSATION_MAX_HISTORY=20
-CONVERSATION_TTL=3600
+TOOLS_MAX_ROUNDS=__MASKED_FOR_NPM__
+CONVERSATION_MAX_HISTORY=__MASKED_FOR_NPM__
+CONVERSATION_TTL=__MASKED_FOR_NPM__
 
 # RAG settings
-RAG_ENABLED=true
-RAG_KNOWLEDGE_DIR=knowledge
-RAG_PERSIST_DIR=.chroma
-RAG_COLLECTION=ops_knowledge
-RAG_EMBEDDING_MODEL=embedding-3
-RAG_EMBEDDING_MAX_CHARS=6000
-RAG_EMBEDDING_BATCH_SIZE=8
-RAG_TOP_K=5
-RAG_SEMANTIC_CANDIDATE_K=40
-RAG_CONTEXT_K=8
-RAG_EXACT_CONTEXT_K=3
-RAG_EXACT_PER_FILE_CONTEXT_K=1
-RAG_PER_FILE_CONTEXT_K=2
-RAG_CHUNK_SIZE=1200
-RAG_REBUILD_ON_STARTUP=false
+RAG_ENABLED=__MASKED_FOR_NPM__
+RAG_KNOWLEDGE_DIR=__MASKED_FOR_NPM__
+RAG_PERSIST_DIR=__MASKED_FOR_NPM__
+RAG_COLLECTION=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MODEL=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MAX_CHARS=__MASKED_FOR_NPM__
+RAG_EMBEDDING_BATCH_SIZE=__MASKED_FOR_NPM__
+RAG_TOP_K=__MASKED_FOR_NPM__
+RAG_SEMANTIC_CANDIDATE_K=__MASKED_FOR_NPM__
+RAG_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_CHUNK_SIZE=__MASKED_FOR_NPM__
+RAG_REBUILD_ON_STARTUP=__MASKED_FOR_NPM__

package/app/.env.prod.bj11

@@ -1,63 +1,63 @@
 # Flask and request authentication
-SECRET_KEY=<MASKED_SECRET>
+SECRET_KEY=__MASKED_FOR_NPM__
 
 # External API endpoints
-GET_USE_INFO_URL=<MASKED_URL>
-GET_ORG_INFO_URL=<MASKED_URL>
-GET_JIRA_INFO_URL=<MASKED_URL>
-GET_BIGDATA_URL=<MASKED_URL>
-GET_UNION_BASE_URL=<MASKED_URL>
+GET_USE_INFO_URL=__MASKED_FOR_NPM__
+GET_ORG_INFO_URL=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL=__MASKED_FOR_NPM__
+GET_BIGDATA_URL=__MASKED_FOR_NPM__
+GET_UNION_BASE_URL=__MASKED_FOR_NPM__
 
 # Legacy external API tokens
-GET_ORG_INFO_URL_TOKEN=<MASKED_TOKEN>
-GET_JIRA_INFO_URL_TOKEN=<MASKED_TOKEN>
+GET_ORG_INFO_URL_TOKEN=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL_TOKEN=__MASKED_FOR_NPM__
 
 # Authorization and logging
-PERMISSIONS=
-LOG_LEVEL=INFO
-CONSOLE_STDOUT=FALSE
-LOG_DIR=/data/appLogs
+PERMISSIONS=__MASKED_FOR_NPM__
+LOG_LEVEL=__MASKED_FOR_NPM__
+CONSOLE_STDOUT=__MASKED_FOR_NPM__
+LOG_DIR=__MASKED_FOR_NPM__
 
 # Runtime environment and JWT
-FLASK_ENV=prod.bj11
-JWT_SECRET_KEY=<MASKED_SECRET>
-JWT_EXPIRATION_SECOND=900
-JWT_RENEW_SECOND=700
+FLASK_ENV=__MASKED_FOR_NPM__
+JWT_SECRET_KEY=__MASKED_FOR_NPM__
+JWT_EXPIRATION_SECOND=__MASKED_FOR_NPM__
+JWT_RENEW_SECOND=__MASKED_FOR_NPM__
 
 # LLM provider
-LLM_URL=<MASKED_URL>
-LLM_KEY=<MASKED_KEY>
-LLM_MODEL=glm-5
-LLM_MAX_TOKENS=4096
-LLM_TEMPERATURE=0.7
-LLM_TOP_P=0.9
+LLM_URL=__MASKED_FOR_NPM__
+LLM_KEY=__MASKED_FOR_NPM__
+LLM_MODEL=__MASKED_FOR_NPM__
+LLM_MAX_TOKENS=__MASKED_FOR_NPM__
+LLM_TEMPERATURE=__MASKED_FOR_NPM__
+LLM_TOP_P=__MASKED_FOR_NPM__
 
 # Chat behavior
-SYSTEM_PROMPT=<MASKED_BUSINESS_VALUE>
+SYSTEM_PROMPT=__MASKED_FOR_NPM__
 
 # Business filter
-FILTER_ENABLED=false
-FILTER_ALLOWED_KEYWORDS=<MASKED_BUSINESS_VALUE>
-FILTER_REJECTION_MESSAGE=<MASKED_BUSINESS_VALUE>
+FILTER_ENABLED=__MASKED_FOR_NPM__
+FILTER_ALLOWED_KEYWORDS=__MASKED_FOR_NPM__
+FILTER_REJECTION_MESSAGE=__MASKED_FOR_NPM__
 
 # Tool and conversation settings
-TOOLS_MAX_ROUNDS=5
-CONVERSATION_MAX_HISTORY=20
-CONVERSATION_TTL=3600
+TOOLS_MAX_ROUNDS=__MASKED_FOR_NPM__
+CONVERSATION_MAX_HISTORY=__MASKED_FOR_NPM__
+CONVERSATION_TTL=__MASKED_FOR_NPM__
 
 # RAG settings
-RAG_ENABLED=true
-RAG_KNOWLEDGE_DIR=knowledge
-RAG_PERSIST_DIR=.chroma
-RAG_COLLECTION=ops_knowledge
-RAG_EMBEDDING_MODEL=embedding-3
-RAG_EMBEDDING_MAX_CHARS=6000
-RAG_EMBEDDING_BATCH_SIZE=8
-RAG_TOP_K=5
-RAG_SEMANTIC_CANDIDATE_K=40
-RAG_CONTEXT_K=8
-RAG_EXACT_CONTEXT_K=3
-RAG_EXACT_PER_FILE_CONTEXT_K=1
-RAG_PER_FILE_CONTEXT_K=2
-RAG_CHUNK_SIZE=1200
-RAG_REBUILD_ON_STARTUP=false
+RAG_ENABLED=__MASKED_FOR_NPM__
+RAG_KNOWLEDGE_DIR=__MASKED_FOR_NPM__
+RAG_PERSIST_DIR=__MASKED_FOR_NPM__
+RAG_COLLECTION=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MODEL=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MAX_CHARS=__MASKED_FOR_NPM__
+RAG_EMBEDDING_BATCH_SIZE=__MASKED_FOR_NPM__
+RAG_TOP_K=__MASKED_FOR_NPM__
+RAG_SEMANTIC_CANDIDATE_K=__MASKED_FOR_NPM__
+RAG_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_CHUNK_SIZE=__MASKED_FOR_NPM__
+RAG_REBUILD_ON_STARTUP=__MASKED_FOR_NPM__

package/app/.env.prod.sh20

@@ -1,63 +1,63 @@
 # Flask and request authentication
-SECRET_KEY=<MASKED_SECRET>
+SECRET_KEY=__MASKED_FOR_NPM__
 
 # External API endpoints
-GET_USE_INFO_URL=<MASKED_URL>
-GET_ORG_INFO_URL=<MASKED_URL>
-GET_JIRA_INFO_URL=<MASKED_URL>
-GET_BIGDATA_URL=<MASKED_URL>
-GET_UNION_BASE_URL=<MASKED_URL>
+GET_USE_INFO_URL=__MASKED_FOR_NPM__
+GET_ORG_INFO_URL=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL=__MASKED_FOR_NPM__
+GET_BIGDATA_URL=__MASKED_FOR_NPM__
+GET_UNION_BASE_URL=__MASKED_FOR_NPM__
 
 # Legacy external API tokens
-GET_ORG_INFO_URL_TOKEN=<MASKED_TOKEN>
-GET_JIRA_INFO_URL_TOKEN=<MASKED_TOKEN>
+GET_ORG_INFO_URL_TOKEN=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL_TOKEN=__MASKED_FOR_NPM__
 
 # Authorization and logging
-PERMISSIONS=
-LOG_LEVEL=INFO
-CONSOLE_STDOUT=FALSE
-LOG_DIR=/data/appLogs
+PERMISSIONS=__MASKED_FOR_NPM__
+LOG_LEVEL=__MASKED_FOR_NPM__
+CONSOLE_STDOUT=__MASKED_FOR_NPM__
+LOG_DIR=__MASKED_FOR_NPM__
 
 # Runtime environment and JWT
-FLASK_ENV=prod.sh20
-JWT_SECRET_KEY=<MASKED_SECRET>
-JWT_EXPIRATION_SECOND=900
-JWT_RENEW_SECOND=700
+FLASK_ENV=__MASKED_FOR_NPM__
+JWT_SECRET_KEY=__MASKED_FOR_NPM__
+JWT_EXPIRATION_SECOND=__MASKED_FOR_NPM__
+JWT_RENEW_SECOND=__MASKED_FOR_NPM__
 
 # LLM provider
-LLM_URL=<MASKED_URL>
-LLM_KEY=<MASKED_KEY>
-LLM_MODEL=glm-5
-LLM_MAX_TOKENS=4096
-LLM_TEMPERATURE=0.7
-LLM_TOP_P=0.9
+LLM_URL=__MASKED_FOR_NPM__
+LLM_KEY=__MASKED_FOR_NPM__
+LLM_MODEL=__MASKED_FOR_NPM__
+LLM_MAX_TOKENS=__MASKED_FOR_NPM__
+LLM_TEMPERATURE=__MASKED_FOR_NPM__
+LLM_TOP_P=__MASKED_FOR_NPM__
 
 # Chat behavior
-SYSTEM_PROMPT=<MASKED_BUSINESS_VALUE>
+SYSTEM_PROMPT=__MASKED_FOR_NPM__
 
 # Business filter
-FILTER_ENABLED=false
-FILTER_ALLOWED_KEYWORDS=<MASKED_BUSINESS_VALUE>
-FILTER_REJECTION_MESSAGE=<MASKED_BUSINESS_VALUE>
+FILTER_ENABLED=__MASKED_FOR_NPM__
+FILTER_ALLOWED_KEYWORDS=__MASKED_FOR_NPM__
+FILTER_REJECTION_MESSAGE=__MASKED_FOR_NPM__
 
 # Tool and conversation settings
-TOOLS_MAX_ROUNDS=5
-CONVERSATION_MAX_HISTORY=20
-CONVERSATION_TTL=3600
+TOOLS_MAX_ROUNDS=__MASKED_FOR_NPM__
+CONVERSATION_MAX_HISTORY=__MASKED_FOR_NPM__
+CONVERSATION_TTL=__MASKED_FOR_NPM__
 
 # RAG settings
-RAG_ENABLED=true
-RAG_KNOWLEDGE_DIR=knowledge
-RAG_PERSIST_DIR=.chroma
-RAG_COLLECTION=ops_knowledge
-RAG_EMBEDDING_MODEL=embedding-3
-RAG_EMBEDDING_MAX_CHARS=6000
-RAG_EMBEDDING_BATCH_SIZE=8
-RAG_TOP_K=5
-RAG_SEMANTIC_CANDIDATE_K=40
-RAG_CONTEXT_K=8
-RAG_EXACT_CONTEXT_K=3
-RAG_EXACT_PER_FILE_CONTEXT_K=1
-RAG_PER_FILE_CONTEXT_K=2
-RAG_CHUNK_SIZE=1200
-RAG_REBUILD_ON_STARTUP=false
+RAG_ENABLED=__MASKED_FOR_NPM__
+RAG_KNOWLEDGE_DIR=__MASKED_FOR_NPM__
+RAG_PERSIST_DIR=__MASKED_FOR_NPM__
+RAG_COLLECTION=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MODEL=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MAX_CHARS=__MASKED_FOR_NPM__
+RAG_EMBEDDING_BATCH_SIZE=__MASKED_FOR_NPM__
+RAG_TOP_K=__MASKED_FOR_NPM__
+RAG_SEMANTIC_CANDIDATE_K=__MASKED_FOR_NPM__
+RAG_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_CHUNK_SIZE=__MASKED_FOR_NPM__
+RAG_REBUILD_ON_STARTUP=__MASKED_FOR_NPM__

package/app/.env.prod.sz31

@@ -1,63 +1,63 @@
 # Flask and request authentication
-SECRET_KEY=<MASKED_SECRET>
+SECRET_KEY=__MASKED_FOR_NPM__
 
 # External API endpoints
-GET_USE_INFO_URL=<MASKED_URL>
-GET_ORG_INFO_URL=<MASKED_URL>
-GET_JIRA_INFO_URL=<MASKED_URL>
-GET_BIGDATA_URL=<MASKED_URL>
-GET_UNION_BASE_URL=<MASKED_URL>
+GET_USE_INFO_URL=__MASKED_FOR_NPM__
+GET_ORG_INFO_URL=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL=__MASKED_FOR_NPM__
+GET_BIGDATA_URL=__MASKED_FOR_NPM__
+GET_UNION_BASE_URL=__MASKED_FOR_NPM__
 
 # Legacy external API tokens
-GET_ORG_INFO_URL_TOKEN=<MASKED_TOKEN>
-GET_JIRA_INFO_URL_TOKEN=<MASKED_TOKEN>
+GET_ORG_INFO_URL_TOKEN=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL_TOKEN=__MASKED_FOR_NPM__
 
 # Authorization and logging
-PERMISSIONS=
-LOG_LEVEL=INFO
-CONSOLE_STDOUT=FALSE
-LOG_DIR=/data/appLogs
+PERMISSIONS=__MASKED_FOR_NPM__
+LOG_LEVEL=__MASKED_FOR_NPM__
+CONSOLE_STDOUT=__MASKED_FOR_NPM__
+LOG_DIR=__MASKED_FOR_NPM__
 
 # Runtime environment and JWT
-FLASK_ENV=prod.sz31
-JWT_SECRET_KEY=<MASKED_SECRET>
-JWT_EXPIRATION_SECOND=900
-JWT_RENEW_SECOND=700
+FLASK_ENV=__MASKED_FOR_NPM__
+JWT_SECRET_KEY=__MASKED_FOR_NPM__
+JWT_EXPIRATION_SECOND=__MASKED_FOR_NPM__
+JWT_RENEW_SECOND=__MASKED_FOR_NPM__
 
 # LLM provider
-LLM_URL=<MASKED_URL>
-LLM_KEY=<MASKED_KEY>
-LLM_MODEL=glm-5
-LLM_MAX_TOKENS=4096
-LLM_TEMPERATURE=0.7
-LLM_TOP_P=0.9
+LLM_URL=__MASKED_FOR_NPM__
+LLM_KEY=__MASKED_FOR_NPM__
+LLM_MODEL=__MASKED_FOR_NPM__
+LLM_MAX_TOKENS=__MASKED_FOR_NPM__
+LLM_TEMPERATURE=__MASKED_FOR_NPM__
+LLM_TOP_P=__MASKED_FOR_NPM__
 
 # Chat behavior
-SYSTEM_PROMPT=<MASKED_BUSINESS_VALUE>
+SYSTEM_PROMPT=__MASKED_FOR_NPM__
 
 # Business filter
-FILTER_ENABLED=false
-FILTER_ALLOWED_KEYWORDS=<MASKED_BUSINESS_VALUE>
-FILTER_REJECTION_MESSAGE=<MASKED_BUSINESS_VALUE>
+FILTER_ENABLED=__MASKED_FOR_NPM__
+FILTER_ALLOWED_KEYWORDS=__MASKED_FOR_NPM__
+FILTER_REJECTION_MESSAGE=__MASKED_FOR_NPM__
 
 # Tool and conversation settings
-TOOLS_MAX_ROUNDS=5
-CONVERSATION_MAX_HISTORY=20
-CONVERSATION_TTL=3600
+TOOLS_MAX_ROUNDS=__MASKED_FOR_NPM__
+CONVERSATION_MAX_HISTORY=__MASKED_FOR_NPM__
+CONVERSATION_TTL=__MASKED_FOR_NPM__
 
 # RAG settings
-RAG_ENABLED=true
-RAG_KNOWLEDGE_DIR=knowledge
-RAG_PERSIST_DIR=.chroma
-RAG_COLLECTION=ops_knowledge
-RAG_EMBEDDING_MODEL=embedding-3
-RAG_EMBEDDING_MAX_CHARS=6000
-RAG_EMBEDDING_BATCH_SIZE=8
-RAG_TOP_K=5
-RAG_SEMANTIC_CANDIDATE_K=40
-RAG_CONTEXT_K=8
-RAG_EXACT_CONTEXT_K=3
-RAG_EXACT_PER_FILE_CONTEXT_K=1
-RAG_PER_FILE_CONTEXT_K=2
-RAG_CHUNK_SIZE=1200
-RAG_REBUILD_ON_STARTUP=false
+RAG_ENABLED=__MASKED_FOR_NPM__
+RAG_KNOWLEDGE_DIR=__MASKED_FOR_NPM__
+RAG_PERSIST_DIR=__MASKED_FOR_NPM__
+RAG_COLLECTION=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MODEL=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MAX_CHARS=__MASKED_FOR_NPM__
+RAG_EMBEDDING_BATCH_SIZE=__MASKED_FOR_NPM__
+RAG_TOP_K=__MASKED_FOR_NPM__
+RAG_SEMANTIC_CANDIDATE_K=__MASKED_FOR_NPM__
+RAG_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_CHUNK_SIZE=__MASKED_FOR_NPM__
+RAG_REBUILD_ON_STARTUP=__MASKED_FOR_NPM__

package/app/.env.test.bj12

@@ -1,63 +1,63 @@
 # Flask and request authentication
-SECRET_KEY=<MASKED_SECRET>
+SECRET_KEY=__MASKED_FOR_NPM__
 
 # External API endpoints
-GET_USE_INFO_URL=<MASKED_URL>
-GET_ORG_INFO_URL=<MASKED_URL>
-GET_JIRA_INFO_URL=<MASKED_URL>
-GET_BIGDATA_URL=<MASKED_URL>
-GET_UNION_BASE_URL=<MASKED_URL>
+GET_USE_INFO_URL=__MASKED_FOR_NPM__
+GET_ORG_INFO_URL=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL=__MASKED_FOR_NPM__
+GET_BIGDATA_URL=__MASKED_FOR_NPM__
+GET_UNION_BASE_URL=__MASKED_FOR_NPM__
 
 # Legacy external API tokens
-GET_ORG_INFO_URL_TOKEN=<MASKED_TOKEN>
-GET_JIRA_INFO_URL_TOKEN=<MASKED_TOKEN>
+GET_ORG_INFO_URL_TOKEN=__MASKED_FOR_NPM__
+GET_JIRA_INFO_URL_TOKEN=__MASKED_FOR_NPM__
 
 # Authorization and logging
-PERMISSIONS=
-LOG_LEVEL=INFO
-CONSOLE_STDOUT=FALSE
-LOG_DIR=/data/appLogs
+PERMISSIONS=__MASKED_FOR_NPM__
+LOG_LEVEL=__MASKED_FOR_NPM__
+CONSOLE_STDOUT=__MASKED_FOR_NPM__
+LOG_DIR=__MASKED_FOR_NPM__
 
 # Runtime environment and JWT
-FLASK_ENV=test.prod.bj12
-JWT_SECRET_KEY=<MASKED_SECRET>
-JWT_EXPIRATION_SECOND=900
-JWT_RENEW_SECOND=700
+FLASK_ENV=__MASKED_FOR_NPM__
+JWT_SECRET_KEY=__MASKED_FOR_NPM__
+JWT_EXPIRATION_SECOND=__MASKED_FOR_NPM__
+JWT_RENEW_SECOND=__MASKED_FOR_NPM__
 
 # LLM provider
-LLM_URL=<MASKED_URL>
-LLM_KEY=<MASKED_KEY>
-LLM_MODEL=glm-5
-LLM_MAX_TOKENS=4096
-LLM_TEMPERATURE=0.7
-LLM_TOP_P=0.9
+LLM_URL=__MASKED_FOR_NPM__
+LLM_KEY=__MASKED_FOR_NPM__
+LLM_MODEL=__MASKED_FOR_NPM__
+LLM_MAX_TOKENS=__MASKED_FOR_NPM__
+LLM_TEMPERATURE=__MASKED_FOR_NPM__
+LLM_TOP_P=__MASKED_FOR_NPM__
 
 # Chat behavior
-SYSTEM_PROMPT=<MASKED_BUSINESS_VALUE>
+SYSTEM_PROMPT=__MASKED_FOR_NPM__
 
 # Business filter
-FILTER_ENABLED=false
-FILTER_ALLOWED_KEYWORDS=<MASKED_BUSINESS_VALUE>
-FILTER_REJECTION_MESSAGE=<MASKED_BUSINESS_VALUE>
+FILTER_ENABLED=__MASKED_FOR_NPM__
+FILTER_ALLOWED_KEYWORDS=__MASKED_FOR_NPM__
+FILTER_REJECTION_MESSAGE=__MASKED_FOR_NPM__
 
 # Tool and conversation settings
-TOOLS_MAX_ROUNDS=5
-CONVERSATION_MAX_HISTORY=20
-CONVERSATION_TTL=3600
+TOOLS_MAX_ROUNDS=__MASKED_FOR_NPM__
+CONVERSATION_MAX_HISTORY=__MASKED_FOR_NPM__
+CONVERSATION_TTL=__MASKED_FOR_NPM__
 
 # RAG settings
-RAG_ENABLED=true
-RAG_KNOWLEDGE_DIR=knowledge
-RAG_PERSIST_DIR=.chroma
-RAG_COLLECTION=ops_knowledge
-RAG_EMBEDDING_MODEL=embedding-3
-RAG_EMBEDDING_MAX_CHARS=6000
-RAG_EMBEDDING_BATCH_SIZE=8
-RAG_TOP_K=5
-RAG_SEMANTIC_CANDIDATE_K=40
-RAG_CONTEXT_K=8
-RAG_EXACT_CONTEXT_K=3
-RAG_EXACT_PER_FILE_CONTEXT_K=1
-RAG_PER_FILE_CONTEXT_K=2
-RAG_CHUNK_SIZE=1200
-RAG_REBUILD_ON_STARTUP=false
+RAG_ENABLED=__MASKED_FOR_NPM__
+RAG_KNOWLEDGE_DIR=__MASKED_FOR_NPM__
+RAG_PERSIST_DIR=__MASKED_FOR_NPM__
+RAG_COLLECTION=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MODEL=__MASKED_FOR_NPM__
+RAG_EMBEDDING_MAX_CHARS=__MASKED_FOR_NPM__
+RAG_EMBEDDING_BATCH_SIZE=__MASKED_FOR_NPM__
+RAG_TOP_K=__MASKED_FOR_NPM__
+RAG_SEMANTIC_CANDIDATE_K=__MASKED_FOR_NPM__
+RAG_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_EXACT_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_PER_FILE_CONTEXT_K=__MASKED_FOR_NPM__
+RAG_CHUNK_SIZE=__MASKED_FOR_NPM__
+RAG_REBUILD_ON_STARTUP=__MASKED_FOR_NPM__

package/app/config/env_config.py

@@ -71,6 +71,12 @@ class Config:
     )
 
     TOOLS_MAX_ROUNDS = _env_int("TOOLS_MAX_ROUNDS", 5)
+    TOOL_CALL_HEARTBEAT_INTERVAL = _env_float("TOOL_CALL_HEARTBEAT_INTERVAL", 15.0)
+    CHAT_OPENING_QUESTIONS = _env_list("CHAT_OPENING_QUESTIONS", [
+        "上周全链路运行质量如何",
+        "最近有哪些成员机构交易异常",
+        "当前系统运行风险点有哪些",
+    ])
 
     CONVERSATION_MAX_HISTORY = _env_int("CONVERSATION_MAX_HISTORY", 20)
     CONVERSATION_TTL = _env_int("CONVERSATION_TTL", 3600)

package/app/manager/chatstream_manager.py

@@ -14,6 +14,8 @@ class ChatstreamManager:
         self._chat_service = chat_service
         self._rag_service = rag_service
         self._conversations: Dict[str, Dict] = {}
+        # ponytail: process-local guard; use shared storage only if workers need cross-process cancellation.
+        self._active_streams: Dict[str, Dict] = {}
         self._max_history = config["CONVERSATION_MAX_HISTORY"]
         self._ttl = config["CONVERSATION_TTL"]
         self._lock = threading.Lock()
@@ -62,6 +64,24 @@ class ChatstreamManager:
             if len(conversation["messages"]) > max_messages:
                 conversation["messages"] = conversation["messages"][-max_messages:]
 
+    def _start_stream(self, jsessionid: str, conversation_id: str) -> threading.Event:
+        abort_event = threading.Event()
+        with self._lock:
+            active = self._active_streams.get(jsessionid)
+            if active and active["conversation_id"] != conversation_id:
+                active["abort_event"].set()
+            self._active_streams[jsessionid] = {
+                "conversation_id": conversation_id,
+                "abort_event": abort_event,
+            }
+        return abort_event
+
+    def _finish_stream(self, jsessionid: str, abort_event: threading.Event):
+        with self._lock:
+            active = self._active_streams.get(jsessionid)
+            if active and active["abort_event"] is abort_event:
+                del self._active_streams[jsessionid]
+
     def chat_stream(
         self,
         conversation_id: Optional[str],
@@ -69,22 +89,34 @@ class ChatstreamManager:
         jsessionid: str,
     ) -> Generator[ChatResponse, None, None]:
         normalized_conversation_id = self.normalize_conversation_id(conversation_id)
+        abort_event = self._start_stream(jsessionid, normalized_conversation_id)
         history = self._get_history(normalized_conversation_id)
         answer_parts: List[str] = []
-        stopped = False
+        saved = False
 
-        for chunk in self._chat_service.tool_call_stream(
-            normalized_conversation_id,
-            question,
-            tools,
-            history,
-            jsessionid,
-        ):
-            if chunk.content:
-                answer_parts.append(chunk.content)
-            if chunk.finish_reason == "stop":
-                stopped = True
-            yield chunk
+        try:
+            for chunk in self._chat_service.tool_call_stream(
+                normalized_conversation_id,
+                question,
+                tools,
+                history,
+                jsessionid,
+            ):
+                if abort_event.is_set():
+                    yield ChatResponse(
+                        conversationId=normalized_conversation_id,
+                        content="当前对话已被新的对话替换，已停止。",
+                        finish_reason="abort",
+                    )
+                    return
+                if chunk.content:
+                    answer_parts.append(chunk.content)
+                if chunk.finish_reason and answer_parts and not saved:
+                    self._append_exchange(normalized_conversation_id, question, "".join(answer_parts))
+                    saved = True
+                yield chunk
 
-        if stopped:
-            self._append_exchange(normalized_conversation_id, question, "".join(answer_parts))
+            if answer_parts and not saved:
+                self._append_exchange(normalized_conversation_id, question, "".join(answer_parts))
+        finally:
+            self._finish_stream(jsessionid, abort_event)

package/app/models/schemas.py

@@ -1,11 +1,11 @@
-from typing import Literal, Optional
+from typing import Any, Dict, Literal, Optional
 
 from pydantic import BaseModel, ConfigDict, Field
 
 
 class ChatRequest(BaseModel):
     """聊天请求模型"""
-    model_config = ConfigDict(extra="forbid")
+    model_config = ConfigDict(extra="ignore")
 
     conversation_id: Optional[str] = Field(
         default=None,
@@ -24,7 +24,9 @@ class ChatResponse(BaseModel):
     reasoning_content: Optional[str] = Field(default=None, description="模型推理内容增量")
     tool_call: Optional[str] = Field(default=None, description="工具调用信息")
     tool_result: Optional[str] = Field(default=None, description="工具执行结果")
-    finish_reason: Optional[Literal["stop", "error", "rejected", "done"]] = Field(
+    heartbeat: Optional[Dict[str, Any]] = Field(default=None, description="长耗时工具调用心跳")
+    error_msg: Optional[str] = Field(default=None, alias="errorMsg", description="错误信息")
+    finish_reason: Optional[Literal["stop", "error", "rejected", "done", "abort"]] = Field(
         default=None,
         description="结束原因；中间流式增量为空",
     )

package/app/service/chat_service.py

@@ -1,3 +1,7 @@
+import json
+import time
+from concurrent.futures import ThreadPoolExecutor, TimeoutError as FutureTimeoutError
+from datetime import datetime, timedelta, timezone
 from pathlib import Path
 from typing import Dict, Generator, List
 
@@ -14,6 +18,33 @@ def _preview(text: str, limit: int = 300) -> str:
     return str(text).replace("\n", " ")[:limit]
 
 
+def _public_tool_result(tool_result: str) -> str:
+    try:
+        payload = json.loads(tool_result)
+    except json.JSONDecodeError:
+        payload = {}
+
+    if not isinstance(payload, dict):
+        payload = {}
+
+    public_payload = {
+        key: payload[key]
+        for key in ("tool_name", "display_name", "arguments", "message")
+        if key in payload
+    }
+    public_payload.setdefault("message", "工具调用完成")
+    return json.dumps(public_payload, ensure_ascii=False)
+
+
+def _time_context() -> str:
+    now = datetime.now(timezone.utc).astimezone(timezone(timedelta(hours=8)))
+    return (
+        "当前基准时间："
+        f"{now:%Y-%m-%d %H:%M:%S} UTC+8。"
+        "用户问题中的“当前、近期、上周、昨天、今天”等相对时间，均按该基准时间计算。"
+    )
+
+
 def _load_prompts() -> Dict[str, str]:
     path = Path(__file__).resolve().parents[2] / "tools" / "prompts.yaml"
     if not path.exists():
@@ -55,6 +86,7 @@ class ChatService:
         self._top_p = config["LLM_TOP_P"]
         self._system_prompt = config["SYSTEM_PROMPT"]
         self._tools_max_rounds = config["TOOLS_MAX_ROUNDS"]
+        self._tool_heartbeat_interval = float(config.get("TOOL_CALL_HEARTBEAT_INTERVAL", 15.0))
         self._rag = rag_service
         self._union_service = union_service
 
@@ -76,6 +108,7 @@ class ChatService:
         messages = []
         if self._system_prompt:
             messages.append({"role": "system", "content": self._system_prompt})
+        messages.append({"role": "system", "content": _time_context()})
         messages.extend(history)
         messages.append({"role": "user", "content": user_question})
         return messages
@@ -118,12 +151,32 @@ class ChatService:
             return ChatResponse(conversationId=conversation_id, tool_call=tool_call)
 
         def tool_result_event(tool_result: str) -> ChatResponse:
-            return ChatResponse(conversationId=conversation_id, tool_result=tool_result)
+            return ChatResponse(conversationId=conversation_id, tool_result=_public_tool_result(tool_result))
+
+        def heartbeat_event(tool_name: str, elapsed_seconds: float) -> ChatResponse:
+            return ChatResponse(
+                conversationId=conversation_id,
+                heartbeat={
+                    "type": "tool_call_running",
+                    "tool": tool_name,
+                    "elapsedSeconds": round(elapsed_seconds, 3),
+                    "message": f"工具 {tool_name} 仍在执行，请继续等待。",
+                },
+            )
+
+        def error_event(error_msg: str) -> ChatResponse:
+            return ChatResponse(
+                conversationId=conversation_id,
+                content=f"[错误] {error_msg}",
+                errorMsg=error_msg,
+                finish_reason="error",
+            )
 
         try:
             messages = self._build_tool_messages(history, question)
             max_rounds = self._tools_max_rounds
             final_answer = ""
+            completed_without_tool_call = False
 
             logger.info(f"开始模型流式调用。conversation_id={conversation_id} model={self._model} question={_preview(question, 120)}")
             for round_idx in range(max_rounds):
@@ -168,6 +221,7 @@ class ChatService:
 
                 if not tool_calls_map:
                     final_answer = current_content
+                    completed_without_tool_call = True
                     break
 
                 assistant_tool_calls = [tool_calls_map[i] for i in sorted(tool_calls_map)]
@@ -189,16 +243,33 @@ class ChatService:
                         rag_service=self._rag,
                         jsessionid=jsessionid,
                     )
-                    result = call_function(name, args, tool_context)
+                    result = yield from self._call_function_with_heartbeats(
+                        name,
+                        args,
+                        tool_context,
+                        heartbeat_event,
+                    )
                     logger.info(f"工具调用完成。conversation_id={conversation_id} tool={name} result_preview={_preview(result, 300)}")
                     yield tool_result_event(result)
 
+                    tool_error = self._extract_tool_error(result)
+                    if tool_error:
+                        logger.error(f"工具调用失败。conversation_id={conversation_id} tool={name} error={tool_error}")
+                        yield error_event(f"工具调用失败: {tool_error}")
+                        return
+
                     messages.append({
                         "role": "tool",
                         "content": result,
                         "tool_call_id": tc["id"],
                     })
 
+            if not completed_without_tool_call:
+                error_msg = f"工具调用轮数达到上限（{max_rounds}轮）"
+                logger.error(f"对话异常结束。conversation_id={conversation_id} error={error_msg}")
+                yield error_event(error_msg)
+                return
+
             logger.info(f"对话完成。conversation_id={conversation_id} final_answer_chars={len(final_answer)} final_answer_preview={_preview(final_answer)}")
             yield ChatResponse(conversationId=conversation_id, finish_reason="stop")
 
@@ -207,9 +278,52 @@ class ChatService:
             yield ChatResponse(
                 conversationId=conversation_id,
                 content=f"[错误] 模型调用异常: {str(e)}",
+                errorMsg=f"模型调用异常: {str(e)}",
                 finish_reason="error",
             )
 
+    def _call_function_with_heartbeats(
+        self,
+        name: str,
+        args: str,
+        tool_context: ToolContext,
+        heartbeat_event,
+    ) -> Generator[ChatResponse, None, str]:
+        interval = self._tool_heartbeat_interval
+        if interval <= 0:
+            return call_function(name, args, tool_context)
+
+        executor = ThreadPoolExecutor(max_workers=1)
+        future = executor.submit(call_function, name, args, tool_context)
+        started_at = time.monotonic()
+        try:
+            while True:
+                try:
+                    return future.result(timeout=interval)
+                except FutureTimeoutError:
+                    yield heartbeat_event(name, time.monotonic() - started_at)
+        finally:
+            executor.shutdown(wait=False, cancel_futures=True)
+
+    @staticmethod
+    def _extract_tool_error(result: str) -> str:
+        try:
+            payload = json.loads(result)
+        except json.JSONDecodeError:
+            return ""
+        if not isinstance(payload, dict):
+            return ""
+
+        error = payload.get("error")
+        if error:
+            return str(error)
+
+        status = payload.get("status")
+        if status and status != "success":
+            message = payload.get("message") or f"工具返回状态: {status}"
+            return str(message)
+        return ""
+
     @staticmethod
     def _merge_tool_call_delta(tool_calls_map: Dict[int, Dict], tc) -> None:
         """将单个流式 tool_call 增量按 index 合并到累积字典中"""

package/app/views/view_chatstream.py

@@ -1,3 +1,6 @@
+import re
+from itertools import chain
+
 from pydantic import ValidationError
 from flask import Blueprint, current_app, request, Response, jsonify, stream_with_context, g
 
@@ -10,6 +13,25 @@ def _sse_event(event: str, data: str) -> str:
     return f"event: {event}\ndata: {data}\n\n"
 
 
+def _error_status(error_msg: str) -> int:
+    match = re.search(r"Error code: (\d{3})", error_msg)
+    if match:
+        status = int(match.group(1))
+        if 400 <= status <= 599:
+            return status
+    return 500
+
+
+def _error_payload(chunk: ChatResponse) -> dict:
+    error_msg = chunk.error_msg or chunk.content or "聊天流异常"
+    return {
+        "conversationId": chunk.conversation_id,
+        "detail": error_msg,
+        "errorMsg": error_msg,
+        "finish_reason": "error",
+    }
+
+
 def _chatstream_manager():
     return current_app.extensions["chatstream_manager"]
 
@@ -27,6 +49,13 @@ def rag_force_rebuild():
         return jsonify({"detail": str(exc)}), 500
 
 
+@chatstream.post("/chat/opening")
+def chat_opening():
+    return jsonify({
+        "suggestedQuestions": current_app.config.get("CHAT_OPENING_QUESTIONS", [])
+    })
+
+
 @chatstream.route("/chat/stream", methods=["OPTIONS", "POST"])
 def chat_stream_endpoint():
     if request.method == "OPTIONS":
@@ -37,12 +66,40 @@ def chat_stream_endpoint():
     except ValidationError as exc:
         return jsonify({"detail": exc.errors()}), 422
 
+    conversation_id = chat_request.conversation_id
+    jsessionid = g.current_user["jsessionid"]
+    stream = _chatstream_manager().chat_stream(conversation_id, chat_request.question, jsessionid)
+
+    try:
+        first_chunk = next(stream)
+    except StopIteration:
+        first_chunk = ChatResponse(conversationId=conversation_id, finish_reason="done")
+    except Exception as exc:
+        error_msg = f"聊天流异常: {exc}"
+        return jsonify({"detail": error_msg, "errorMsg": error_msg, "finish_reason": "error"}), 500
+
+    if first_chunk.finish_reason == "error":
+        return jsonify(_error_payload(first_chunk)), _error_status(first_chunk.error_msg or first_chunk.content or "")
+
     def event_generator():
-        conversation_id = chat_request.conversation_id
-        jsessionid = g.current_user["jsessionid"]
-        for chunk in _chatstream_manager().chat_stream(conversation_id, chat_request.question, jsessionid):
-            conversation_id = chunk.conversation_id
-            yield _sse_event("message", chunk.model_dump_json(by_alias=True))
+        nonlocal conversation_id
+        try:
+            for chunk in chain([first_chunk], stream):
+                conversation_id = chunk.conversation_id
+                event = "heartbeat" if chunk.heartbeat else "message"
+                yield _sse_event(event, chunk.model_dump_json(by_alias=True))
+                if chunk.finish_reason == "error":
+                    return
+        except Exception as exc:
+            error_msg = f"聊天流异常: {exc}"
+            error = ChatResponse(
+                conversationId=conversation_id,
+                content=f"[错误] {error_msg}",
+                errorMsg=error_msg,
+                finish_reason="error",
+            )
+            yield _sse_event("message", error.model_dump_json(by_alias=True))
+            return
 
         done = ChatResponse(conversationId=conversation_id, finish_reason="done")
         yield _sse_event("done", done.model_dump_json(by_alias=True))

package/app/wsgi.py

@@ -34,4 +34,4 @@ except Exception:
 
 if __name__ == '__main__':
     if 'dev' in app.config["FLASK_ENV"]:
-        app.run(debug=True, host='0.0.0.0', port=8000)
+        app.run(debug=True, host='0.0.0.0', port=8083)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "union-app-chat-stream",
-  "version": "1.0.3",
+  "version": "1.0.4",
   "description": "Union operations chat stream Flask application package.",
   "license": "UNLICENSED",
   "files": [
@@ -8,6 +8,8 @@
     "deploy",
     "!deploy/offline-packages",
     "!deploy/offline-packages/**",
+    "!app/**/__pycache__",
+    "!app/**/*.pyc",
     "knowledge",
     "tools",
     "PROJECT_OVERVIEW.md",