Fix bug: broken import from rag.prompts.prompts (#10217)

### What problem does this PR solve?

Fix broken imports

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

---------

Signed-off-by: jinhai <haijin.chn@gmail.com>

agent/canvas.py

@@ -27,7 +27,7 @@ from agent.component import component_class
 from agent.component.base import ComponentBase
 from api.db.services.file_service import FileService
 from api.utils import get_uuid, hash_str2int
-from rag.prompts.prompts import chunks_format
+from rag.prompts.generator import chunks_format
 from rag.utils.redis_conn import REDIS_CONN
 
 class Graph:
@@ -490,7 +490,8 @@ class Canvas(Graph):
 
         r = self.retrieval[-1]
         for ck in chunks_format({"chunks": chunks}):
-            cid = hash_str2int(ck["id"], 100)
+            cid = hash_str2int(ck["id"], 500)
+            # cid = uuid.uuid5(uuid.NAMESPACE_DNS, ck["id"])
             if cid not in r:
                 r["chunks"][cid] = ck
 

agent/component/agent_with_tools.py

@@ -28,9 +28,8 @@ from api.db.services.llm_service import LLMBundle
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.mcp_server_service import MCPServerService
 from api.utils.api_utils import timeout
-from rag.prompts import message_fit_in
-from rag.prompts.prompts import next_step, COMPLETE_TASK, analyze_task, \
-    citation_prompt, reflect, rank_memories, kb_prompt, citation_plus, full_question
+from rag.prompts.generator import next_step, COMPLETE_TASK, analyze_task, \
+    citation_prompt, reflect, rank_memories, kb_prompt, citation_plus, full_question, message_fit_in
 from rag.utils.mcp_tool_call_conn import MCPToolCallSession, mcp_tool_metadata_to_openai_tool
 from agent.component.llm import LLMParam, LLM
 

agent/component/llm.py

@@ -26,8 +26,7 @@ from api.db.services.llm_service import LLMBundle
 from api.db.services.tenant_llm_service import TenantLLMService
 from agent.component.base import ComponentBase, ComponentParamBase
 from api.utils.api_utils import timeout
-from rag.prompts import message_fit_in, citation_prompt
-from rag.prompts.prompts import tool_call_summary
+from rag.prompts.generator import tool_call_summary, message_fit_in, citation_prompt
 
 
 class LLMParam(ComponentParamBase):

agent/component/string_transform.py

@@ -90,6 +90,8 @@ class StringTransform(Message, ABC):
         for k,v in kwargs.items():
             if not v:
                 v = ""
+            k = re.sub(r'\\m', 'm', k)
+            v = re.sub(r'\\m', 'm', v)
             script = re.sub(k, v, script)
 
         self.set_output("result", script)

agent/tools/base.py

@@ -22,7 +22,7 @@ from typing import TypedDict, List, Any
 from agent.component.base import ComponentParamBase, ComponentBase
 from api.utils import hash_str2int
 from rag.llm.chat_model import ToolCallSession
-from rag.prompts.prompts import kb_prompt
+from rag.prompts.generator import kb_prompt
 from rag.utils.mcp_tool_call_conn import MCPToolCallSession
 from timeit import default_timer as timer
 

agent/tools/retrieval.py

@@ -23,8 +23,7 @@ from api.db.services.llm_service import LLMBundle
 from api import settings
 from api.utils.api_utils import timeout
 from rag.app.tag import label_question
-from rag.prompts import kb_prompt
-from rag.prompts.prompts import cross_languages
+from rag.prompts.generator import cross_languages, kb_prompt
 
 
 class RetrievalParam(ToolParamBase):

api/apps/api_app.py

@@ -39,7 +39,7 @@ from api.utils.api_utils import server_error_response, get_data_error_result, ge
 
 from api.utils.file_utils import filename_type, thumbnail
 from rag.app.tag import label_question
-from rag.prompts import keyword_extraction
+from rag.prompts.generator import keyword_extraction
 from rag.utils.storage_factory import STORAGE_IMPL
 
 from api.db.services.canvas_service import UserCanvasService

api/apps/canvas_app.py

@@ -23,7 +23,7 @@ import trio
 from flask import request, Response
 from flask_login import login_required, current_user
 
-from agent.component import LLM
+from agent.component.llm import LLM
 from api.db import CanvasCategory, FileType
 from api.db.services.canvas_service import CanvasTemplateService, UserCanvasService, API4ConversationService
 from api.db.services.document_service import DocumentService
@@ -474,7 +474,7 @@ def sessions(canvas_id):
 @manager.route('/prompts', methods=['GET'])  # noqa: F821
 @login_required
 def prompts():
-    from rag.prompts.prompts import ANALYZE_TASK_SYSTEM, ANALYZE_TASK_USER, NEXT_STEP, REFLECT, CITATION_PROMPT_TEMPLATE
+    from rag.prompts.generator import ANALYZE_TASK_SYSTEM, ANALYZE_TASK_USER, NEXT_STEP, REFLECT, CITATION_PROMPT_TEMPLATE
     return get_json_result(data={
         "task_analysis": ANALYZE_TASK_SYSTEM +"\n\n"+ ANALYZE_TASK_USER,
         "plan_generation": NEXT_STEP,

api/apps/chunk_app.py

@@ -33,8 +33,7 @@ from api.utils.api_utils import get_data_error_result, get_json_result, server_e
 from rag.app.qa import beAdoc, rmPrefix
 from rag.app.tag import label_question
 from rag.nlp import rag_tokenizer, search
-from rag.prompts import cross_languages, keyword_extraction
-from rag.prompts.prompts import gen_meta_filter
+from rag.prompts.generator import gen_meta_filter, cross_languages, keyword_extraction
 from rag.settings import PAGERANK_FLD
 from rag.utils import rmSpace
 

api/apps/conversation_app.py

@@ -29,8 +29,8 @@ from api.db.services.search_service import SearchService
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.user_service import TenantService, UserTenantService
 from api.utils.api_utils import get_data_error_result, get_json_result, server_error_response, validate_request
-from rag.prompts.prompt_template import load_prompt
-from rag.prompts.prompts import chunks_format
+from rag.prompts.template import load_prompt
+from rag.prompts.generator import chunks_format
 
 
 @manager.route("/set", methods=["POST"])  # noqa: F821

api/apps/dataflow_app.py

@@ -24,7 +24,7 @@ from flask import request
 from flask_login import current_user, login_required
 
 from agent.canvas import Canvas
-from agent.component import LLM
+from agent.component.llm import LLM
 from api.db import CanvasCategory, FileType
 from api.db.services.canvas_service import CanvasTemplateService, UserCanvasService
 from api.db.services.document_service import DocumentService

api/apps/sdk/doc.py

@@ -40,7 +40,7 @@ from api.utils.api_utils import check_duplicate_ids, construct_json_result, get_
 from rag.app.qa import beAdoc, rmPrefix
 from rag.app.tag import label_question
 from rag.nlp import rag_tokenizer, search
-from rag.prompts import cross_languages, keyword_extraction
+from rag.prompts.generator import cross_languages, keyword_extraction
 from rag.utils import rmSpace
 from rag.utils.storage_factory import STORAGE_IMPL
 

api/apps/sdk/session.py

@@ -38,9 +38,8 @@ from api.db.services.user_service import UserTenantService
 from api.utils import get_uuid
 from api.utils.api_utils import check_duplicate_ids, get_data_openai, get_error_data_result, get_json_result, get_result, server_error_response, token_required, validate_request
 from rag.app.tag import label_question
-from rag.prompts import chunks_format
-from rag.prompts.prompt_template import load_prompt
-from rag.prompts.prompts import cross_languages, gen_meta_filter, keyword_extraction
+from rag.prompts.template import load_prompt
+from rag.prompts.generator import cross_languages, gen_meta_filter, keyword_extraction, chunks_format
 
 
 @manager.route("/chats/<chat_id>/sessions", methods=["POST"])  # noqa: F821
@@ -183,7 +182,7 @@ def chat_completion_openai_like(tenant_id, chat_id):
     stream = True
     reference = True
 
-    completion = client.chat.completions.create(
+    completion = client.responses.create(
         model=model,
         messages=[
             {"role": "system", "content": "You are a helpful assistant."},

api/db/services/conversation_service.py

@@ -23,7 +23,7 @@ from api.db.services.dialog_service import DialogService, chat
 from api.utils import get_uuid
 import json
 
-from rag.prompts import chunks_format
+from rag.prompts.generator import chunks_format
 
 
 class ConversationService(CommonService):

api/db/services/dialog_service.py

@@ -39,8 +39,8 @@ from graphrag.general.mind_map_extractor import MindMapExtractor
 from rag.app.resume import forbidden_select_fields4resume
 from rag.app.tag import label_question
 from rag.nlp.search import index_name
-from rag.prompts import chunks_format, citation_prompt, cross_languages, full_question, kb_prompt, keyword_extraction, message_fit_in
-from rag.prompts.prompts import gen_meta_filter, PROMPT_JINJA_ENV, ASK_SUMMARY
+from rag.prompts.generator import chunks_format, citation_prompt, cross_languages, full_question, kb_prompt, keyword_extraction, message_fit_in, \
+    gen_meta_filter, PROMPT_JINJA_ENV, ASK_SUMMARY
 from rag.utils import num_tokens_from_string, rmSpace
 from rag.utils.tavily_conn import Tavily
 
@@ -176,7 +176,7 @@ def chat_solo(dialog, messages, stream=True):
         delta_ans = ""
         for ans in chat_mdl.chat_streamly(prompt_config.get("system", ""), msg, dialog.llm_setting):
             answer = ans
-            delta_ans = ans[len(last_ans) :]
+            delta_ans = ans[len(last_ans):]
             if num_tokens_from_string(delta_ans) < 16:
                 continue
             last_ans = answer
@@ -261,13 +261,13 @@ def convert_conditions(metadata_condition):
         "not is": "≠"
     }
     return [
-    {
-        "op": op_mapping.get(cond["comparison_operator"], cond["comparison_operator"]),
-        "key": cond["name"],
-        "value": cond["value"]
-    }
-    for cond in metadata_condition.get("conditions", [])
-]
+        {
+            "op": op_mapping.get(cond["comparison_operator"], cond["comparison_operator"]),
+            "key": cond["name"],
+            "value": cond["value"]
+        }
+        for cond in metadata_condition.get("conditions", [])
+    ]
 
 
 def meta_filter(metas: dict, filters: list[dict]):
@@ -284,19 +284,19 @@ def meta_filter(metas: dict, filters: list[dict]):
                 value = str(value)
 
             for conds in [
-                    (operator == "contains", str(value).lower() in str(input).lower()),
-                    (operator == "not contains", str(value).lower() not in str(input).lower()),
-                    (operator == "start with", str(input).lower().startswith(str(value).lower())),
-                    (operator == "end with", str(input).lower().endswith(str(value).lower())),
-                    (operator == "empty", not input),
-                    (operator == "not empty", input),
-                    (operator == "=", input == value),
-                    (operator == "≠", input != value),
-                    (operator == ">", input > value),
-                    (operator == "<", input < value),
-                    (operator == "≥", input >= value),
-                    (operator == "≤", input <= value),
-                ]:
+                (operator == "contains", str(value).lower() in str(input).lower()),
+                (operator == "not contains", str(value).lower() not in str(input).lower()),
+                (operator == "start with", str(input).lower().startswith(str(value).lower())),
+                (operator == "end with", str(input).lower().endswith(str(value).lower())),
+                (operator == "empty", not input),
+                (operator == "not empty", input),
+                (operator == "=", input == value),
+                (operator == "≠", input != value),
+                (operator == ">", input > value),
+                (operator == "<", input < value),
+                (operator == "≥", input >= value),
+                (operator == "≤", input <= value),
+            ]:
                 try:
                     if all(conds):
                         ids.extend(docids)
@@ -456,7 +456,8 @@ def chat(dialog, messages, stream=True, **kwargs):
                 kbinfos["chunks"].extend(tav_res["chunks"])
                 kbinfos["doc_aggs"].extend(tav_res["doc_aggs"])
             if prompt_config.get("use_kg"):
-                ck = settings.kg_retrievaler.retrieval(" ".join(questions), tenant_ids, dialog.kb_ids, embd_mdl, LLMBundle(dialog.tenant_id, LLMType.CHAT))
+                ck = settings.kg_retrievaler.retrieval(" ".join(questions), tenant_ids, dialog.kb_ids, embd_mdl,
+                                                       LLMBundle(dialog.tenant_id, LLMType.CHAT))
                 if ck["content_with_weight"]:
                     kbinfos["chunks"].insert(0, ck)
 
@@ -467,7 +468,8 @@ def chat(dialog, messages, stream=True, **kwargs):
     retrieval_ts = timer()
     if not knowledges and prompt_config.get("empty_response"):
         empty_res = prompt_config["empty_response"]
-        yield {"answer": empty_res, "reference": kbinfos, "prompt": "\n\n### Query:\n%s" % " ".join(questions), "audio_binary": tts(tts_mdl, empty_res)}
+        yield {"answer": empty_res, "reference": kbinfos, "prompt": "\n\n### Query:\n%s" % " ".join(questions),
+               "audio_binary": tts(tts_mdl, empty_res)}
         return {"answer": prompt_config["empty_response"], "reference": kbinfos}
 
     kwargs["knowledge"] = "\n------\n" + "\n\n------\n\n".join(knowledges)
@@ -565,7 +567,8 @@ def chat(dialog, messages, stream=True, **kwargs):
 
     if langfuse_tracer:
         langfuse_generation = langfuse_tracer.start_generation(
-            trace_context=trace_context, name="chat", model=llm_model_config["llm_name"], input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg}
+            trace_context=trace_context, name="chat", model=llm_model_config["llm_name"],
+            input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg}
         )
 
     if stream:
@@ -575,12 +578,12 @@ def chat(dialog, messages, stream=True, **kwargs):
             if thought:
                 ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
             answer = ans
-            delta_ans = ans[len(last_ans) :]
+            delta_ans = ans[len(last_ans):]
             if num_tokens_from_string(delta_ans) < 16:
                 continue
             last_ans = answer
             yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
-        delta_ans = answer[len(last_ans) :]
+        delta_ans = answer[len(last_ans):]
         if delta_ans:
             yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
         yield decorate_answer(thought + answer)
@@ -676,7 +679,9 @@ Please write the SQL, only SQL, without any other explanations or text.
 
     # compose Markdown table
     columns = (
-        "|" + "|".join([re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"], tbl["columns"][i]["name"])) for i in column_idx]) + ("|Source|" if docid_idx and docid_idx else "|")
+            "|" + "|".join(
+        [re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"], tbl["columns"][i]["name"])) for i in column_idx]) + (
+                "|Source|" if docid_idx and docid_idx else "|")
     )
 
     line = "|" + "|".join(["------" for _ in range(len(column_idx))]) + ("|------|" if docid_idx and docid_idx else "")
@@ -753,7 +758,7 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
                 doc_ids = None
 
     kbinfos = retriever.retrieval(
-        question = question,
+        question=question,
         embd_mdl=embd_mdl,
         tenant_ids=tenant_ids,
         kb_ids=kb_ids,
@@ -775,7 +780,8 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
 
     def decorate_answer(answer):
         nonlocal knowledges, kbinfos, sys_prompt
-        answer, idx = retriever.insert_citations(answer, [ck["content_ltks"] for ck in kbinfos["chunks"]], [ck["vector"] for ck in kbinfos["chunks"]], embd_mdl, tkweight=0.7, vtweight=0.3)
+        answer, idx = retriever.insert_citations(answer, [ck["content_ltks"] for ck in kbinfos["chunks"]], [ck["vector"] for ck in kbinfos["chunks"]],
+                                                 embd_mdl, tkweight=0.7, vtweight=0.3)
         idx = set([kbinfos["chunks"][int(i)]["doc_id"] for i in idx])
         recall_docs = [d for d in kbinfos["doc_aggs"] if d["doc_id"] in idx]
         if not recall_docs:

conf/llm_factories.json

@@ -622,6 +622,13 @@
                     "tags": "SPEECH2TEXT,8k",
                     "max_tokens": 8000,
                     "model_type": "speech2text"
+                },
+                {
+                    "llm_name": "qianwen-deepresearch-30b-a3b-131k",
+                    "tags": "LLM,CHAT,1M,AGENT,DEEPRESEARCH",
+                    "max_tokens": 1000000,
+                    "model_type": "chat",
+                    "is_tools": true
                 }
             ]
         },

deepdoc/parser/figure_parser.py

@@ -19,7 +19,7 @@ from PIL import Image
 
 from api.utils.api_utils import timeout
 from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
-from rag.prompts import vision_llm_figure_describe_prompt
+from rag.prompts.generator import vision_llm_figure_describe_prompt
 
 
 def vision_figure_parser_figure_data_wrapper(figures_data_without_positions):

deepdoc/parser/pdf_parser.py

@@ -37,7 +37,7 @@ from api.utils.file_utils import get_project_base_directory
 from deepdoc.vision import OCR, AscendLayoutRecognizer, LayoutRecognizer, Recognizer, TableStructureRecognizer
 from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
 from rag.nlp import rag_tokenizer
-from rag.prompts import vision_llm_describe_prompt
+from rag.prompts.generator import vision_llm_describe_prompt
 from rag.settings import PARALLEL_DEVICES
 
 LOCK_KEY_pdfplumber = "global_shared_lock_pdfplumber"

graphrag/general/extractor.py

@@ -37,7 +37,7 @@ from graphrag.utils import (
     split_string_by_multi_markers,
 )
 from rag.llm.chat_model import Base as CompletionLLM
-from rag.prompts import message_fit_in
+from rag.prompts.generator import message_fit_in
 from rag.utils import truncate
 
 GRAPH_FIELD_SEP = "<SEP>"

rag/flow/chunker/chunker.py

@@ -23,7 +23,7 @@ from graphrag.utils import chat_limiter, get_llm_cache, set_llm_cache
 from rag.flow.base import ProcessBase, ProcessParamBase
 from rag.flow.chunker.schema import ChunkerFromUpstream
 from rag.nlp import naive_merge, naive_merge_with_images
-from rag.prompts.prompts import keyword_extraction, question_proposal
+from rag.prompts.generator import keyword_extraction, question_proposal
 
 
 class ChunkerParam(ProcessParamBase):

rag/llm/chat_model.py

@@ -143,7 +143,8 @@ class Base(ABC):
         logging.info("[HISTORY]" + json.dumps(history, ensure_ascii=False, indent=2))
         if self.model_name.lower().find("qwen3") >= 0:
             kwargs["extra_body"] = {"enable_thinking": False}
-        response = self.client.chat.completions.create(model=self.model_name, messages=history, **gen_conf, **kwargs)
+        
+        response = self.client.responses.create(model=self.model_name, messages=history, **gen_conf, **kwargs)
 
         if any([not response.choices, not response.choices[0].message, not response.choices[0].message.content]):
             return "", 0
@@ -155,10 +156,12 @@ class Base(ABC):
     def _chat_streamly(self, history, gen_conf, **kwargs):
         logging.info("[HISTORY STREAMLY]" + json.dumps(history, ensure_ascii=False, indent=4))
         reasoning_start = False
+        
         if kwargs.get("stop") or "stop" in gen_conf:
-            response = self.client.chat.completions.create(model=self.model_name, messages=history, stream=True, **gen_conf, stop=kwargs.get("stop"))
+            response = self.client.responses.create(model=self.model_name, messages=history, stream=True, **gen_conf, stop=kwargs.get("stop"))
         else:
-            response = self.client.chat.completions.create(model=self.model_name, messages=history, stream=True, **gen_conf)
+            response = self.client.responses.create(model=self.model_name, messages=history, stream=True, **gen_conf)
+        
         for resp in response:
             if not resp.choices:
                 continue
@@ -254,7 +257,7 @@ class Base(ABC):
             try:
                 for _ in range(self.max_rounds + 1):
                     logging.info(f"{self.tools=}")
-                    response = self.client.chat.completions.create(model=self.model_name, messages=history, tools=self.tools, tool_choice="auto", **gen_conf)
+                    response = self.client.responses.create(model=self.model_name, messages=history, tools=self.tools, tool_choice="auto", **gen_conf)
                     tk_count += self.total_token_count(response)
                     if any([not response.choices, not response.choices[0].message]):
                         raise Exception(f"500 response structure error. Response: {response}")
@@ -339,7 +342,7 @@ class Base(ABC):
                 for _ in range(self.max_rounds + 1):
                     reasoning_start = False
                     logging.info(f"{tools=}")
-                    response = self.client.chat.completions.create(model=self.model_name, messages=history, stream=True, tools=tools, tool_choice="auto", **gen_conf)
+                    response = self.client.responses.create(model=self.model_name, messages=history, stream=True, tools=tools, tool_choice="auto", **gen_conf)
                     final_tool_calls = {}
                     answer = ""
                     for resp in response:
@@ -402,7 +405,7 @@ class Base(ABC):
 
                 logging.warning(f"Exceed max rounds: {self.max_rounds}")
                 history.append({"role": "user", "content": f"Exceed max rounds: {self.max_rounds}"})
-                response = self.client.chat.completions.create(model=self.model_name, messages=history, stream=True, **gen_conf)
+                response = self.client.responses.create(model=self.model_name, messages=history, stream=True, **gen_conf)
                 for resp in response:
                     if any([not resp.choices, not resp.choices[0].delta, not hasattr(resp.choices[0].delta, "content")]):
                         raise Exception("500 response structure error.")
@@ -556,7 +559,7 @@ class BaiChuanChat(Base):
         }
 
     def _chat(self, history, gen_conf={}, **kwargs):
-        response = self.client.chat.completions.create(
+        response = self.client.responses.create(
             model=self.model_name,
             messages=history,
             extra_body={"tools": [{"type": "web_search", "web_search": {"enable": True, "search_mode": "performance_first"}}]},
@@ -578,7 +581,7 @@ class BaiChuanChat(Base):
         ans = ""
         total_tokens = 0
         try:
-            response = self.client.chat.completions.create(
+            response = self.client.responses.create(
                 model=self.model_name,
                 messages=history,
                 extra_body={"tools": [{"type": "web_search", "web_search": {"enable": True, "search_mode": "performance_first"}}]},
@@ -648,7 +651,7 @@ class ZhipuChat(Base):
         tk_count = 0
         try:
             logging.info(json.dumps(history, ensure_ascii=False, indent=2))
-            response = self.client.chat.completions.create(model=self.model_name, messages=history, stream=True, **gen_conf)
+            response = self.client.responses.create(model=self.model_name, messages=history, stream=True, **gen_conf)
             for resp in response:
                 if not resp.choices[0].delta.content:
                     continue
@@ -1361,7 +1364,7 @@ class LiteLLMBase(ABC):
             drop_params=True,
             timeout=self.timeout,
         )
-        # response = self.client.chat.completions.create(model=self.model_name, messages=history, **gen_conf, **kwargs)
+        # response = self.client.responses.create(model=self.model_name, messages=history, **gen_conf, **kwargs)
 
         if any([not response.choices, not response.choices[0].message, not response.choices[0].message.content]):
             return "", 0

rag/llm/cv_model.py

@@ -25,7 +25,7 @@ from openai import OpenAI
 from openai.lib.azure import AzureOpenAI
 from zhipuai import ZhipuAI
 from rag.nlp import is_english
-from rag.prompts import vision_llm_describe_prompt
+from rag.prompts.generator import vision_llm_describe_prompt
 from rag.utils import num_tokens_from_string
 
 
@@ -75,7 +75,7 @@ class Base(ABC):
 
     def chat(self, system, history, gen_conf, images=[], **kwargs):
         try:
-            response = self.client.chat.completions.create(
+            response = self.client.responses.create(
                 model=self.model_name,
                 messages=self._form_history(system, history, images)
             )
@@ -87,7 +87,7 @@ class Base(ABC):
         ans = ""
         tk_count = 0
         try:
-            response = self.client.chat.completions.create(
+            response = self.client.responses.create(
                 model=self.model_name,
                 messages=self._form_history(system, history, images),
                 stream=True
@@ -174,7 +174,8 @@ class GptV4(Base):
 
     def describe(self, image):
         b64 = self.image2base64(image)
-        res = self.client.chat.completions.create(
+        # Check if this is a GPT-5 model and use responses.create API
+        res = self.client.responses.create(
             model=self.model_name,
             messages=self.prompt(b64),
         )
@@ -182,7 +183,7 @@ class GptV4(Base):
 
     def describe_with_prompt(self, image, prompt=None):
         b64 = self.image2base64(image)
-        res = self.client.chat.completions.create(
+        res = self.client.responses.create(
             model=self.model_name,
             messages=self.vision_llm_prompt(b64, prompt),
         )

rag/prompts/__init__.py

@@ -1,6 +1,6 @@
-from . import prompts
+from . import generator
 
-__all__ = [name for name in dir(prompts)
+__all__ = [name for name in dir(generator)
            if not name.startswith('_')]
 
-globals().update({name: getattr(prompts, name) for name in __all__})
+globals().update({name: getattr(generator, name) for name in __all__})

rag/prompts/generator.py

@@ -22,7 +22,7 @@ from typing import Tuple
 import jinja2
 import json_repair
 from api.utils import hash_str2int
-from rag.prompts.prompt_template import load_prompt
+from rag.prompts.template import load_prompt
 from rag.settings import TAG_FLD
 from rag.utils import encoder, num_tokens_from_string
 

rag/svr/task_executor.py

@@ -27,7 +27,7 @@ from api.utils.log_utils import init_root_logger, get_project_base_directory
 from graphrag.general.index import run_graphrag
 from graphrag.utils import get_llm_cache, set_llm_cache, get_tags_from_cache, set_tags_to_cache
 from rag.flow.pipeline import Pipeline
-from rag.prompts import keyword_extraction, question_proposal, content_tagging
+from rag.prompts.generator import keyword_extraction, question_proposal, content_tagging
 
 import logging
 import os