auto-coder 0.1.255__py3-none-any.whl → 0.1.257__py3-none-any.whl

autocoder/common/code_auto_generate_diff.py

@@ -2,6 +2,7 @@ from typing import List, Dict, Tuple
 from autocoder.common.types import Mode, CodeGenerateResult
 from autocoder.common import AutoCoderArgs
 import byzerllm
+from autocoder.privacy.model_filter import ModelPathFilter
 from autocoder.utils.queue_communicate import queue_communicate, CommunicateEvent, CommunicateEventType
 from autocoder.common import sys_prompt
 from concurrent.futures import ThreadPoolExecutor
@@ -9,7 +10,8 @@ import json
 from autocoder.common.utils_code_auto_generate import chat_with_continue
 from autocoder.common.printer import Printer
 from autocoder.rag.token_counter import count_tokens
-
+from autocoder.utils import llms as llm_utils
+from autocoder.common import SourceCodeList
 
 class CodeAutoGenerateDiff:
     def __init__(
@@ -301,9 +303,10 @@ class CodeAutoGenerateDiff:
         }
 
     def single_round_run(
-        self, query: str, source_content: str
+        self, query: str, source_code_list: SourceCodeList
     ) -> CodeGenerateResult:
         llm_config = {"human_as_model": self.args.human_as_model}
+        source_content = source_code_list.to_str()
 
         if self.args.template == "common":
             init_prompt = self.single_round_instruction.prompt(
@@ -341,6 +344,9 @@ class CodeAutoGenerateDiff:
         results = []
         input_tokens_count = 0
         generated_tokens_count = 0
+        input_tokens_cost = 0
+        generated_tokens_cost = 0
+        model_names = []
 
         printer = Printer()
         estimated_input_tokens = count_tokens(json.dumps(conversations, ensure_ascii=False))
@@ -354,13 +360,26 @@ class CodeAutoGenerateDiff:
                 futures = []
                 for llm in self.llms:
                     for _ in range(self.generate_times_same_model):
-                        futures.append(executor.submit(
-                            chat_with_continue, llm=llm, conversations=conversations, llm_config=llm_config))
+                        model_names_list = llm_utils.get_llm_names(llm)
+                        model_name = None
+                        if model_names_list:
+                            model_name = model_names_list[0]
+                        
+                        for _ in range(self.generate_times_same_model):
+                            model_names.append(model_name)
+                            futures.append(executor.submit(
+                                chat_with_continue, llm=llm, conversations=conversations, llm_config=llm_config))
+                
                 temp_results = [future.result() for future in futures]
                 for result in temp_results:
                     results.append(result.content)
                     input_tokens_count += result.input_tokens_count
                     generated_tokens_count += result.generated_tokens_count
+                    model_info = llm_utils.get_model_info(model_name, self.args.product_mode)
+                    input_cost = model_info.get("input_price",0) if model_info else 0
+                    output_cost = model_info.get("output_price",0) if model_info else 0
+                    input_tokens_cost += input_cost * result.input_tokens_count / 1000000
+                    generated_tokens_cost += output_cost * result.generated_tokens_count / 1000000
             
             for result in results:
                 conversations_list.append(
@@ -376,7 +395,9 @@ class CodeAutoGenerateDiff:
         
         statistics = {
             "input_tokens_count": input_tokens_count,
-            "generated_tokens_count": generated_tokens_count
+            "generated_tokens_count": generated_tokens_count,
+            "input_tokens_cost": input_tokens_cost,
+            "generated_tokens_cost": generated_tokens_cost
         }        
     
         if self.args.request_id and not self.args.skip_events:
@@ -391,11 +412,28 @@ class CodeAutoGenerateDiff:
         return CodeGenerateResult(contents=results, conversations=conversations_list, metadata=statistics)
 
     def multi_round_run(
-        self, query: str, source_content: str, max_steps: int = 10
+        self, query: str, source_code_list: SourceCodeList, max_steps: int = 10
     ) -> CodeGenerateResult:
+        
+        # Apply model filter for code_llm
+        printer = Printer()
+        for llm in self.llms:
+            model_filter = ModelPathFilter.from_model_object(llm, self.args)
+            filtered_sources = []
+            for source in source_code_list.sources:
+                if model_filter.is_accessible(source.module_name):
+                    filtered_sources.append(source)
+                else:
+                    printer.print_in_terminal("index_file_filtered", 
+                                               style="yellow",
+                                               file_path=source.path, 
+                                               model_name=",".join(llm_utils.get_llm_names(llm)))
+            
+        source_code_list = SourceCodeList(filtered_sources)
+        
         llm_config = {"human_as_model": self.args.human_as_model}
         result = []
-
+        source_content = source_code_list.to_str()
         if self.args.template == "common":
             init_prompt = self.multi_round_instruction.prompt(
                 instruction=query, content=source_content, context=self.args.context

autocoder/common/code_auto_generate_editblock.py

@@ -3,6 +3,7 @@ from autocoder.common.types import Mode, CodeGenerateResult
 from autocoder.common import AutoCoderArgs
 import byzerllm
 from autocoder.common import sys_prompt
+from autocoder.privacy.model_filter import ModelPathFilter
 from autocoder.utils.queue_communicate import (
     queue_communicate,
     CommunicateEvent,
@@ -13,6 +14,8 @@ from concurrent.futures import ThreadPoolExecutor
 from autocoder.common.utils_code_auto_generate import chat_with_continue
 from autocoder.common.printer import Printer
 from autocoder.rag.token_counter import count_tokens
+from autocoder.utils import llms as llm_utils
+from autocoder.common import SourceCodeList
 
 
 class CodeAutoGenerateEditBlock:
@@ -383,10 +386,29 @@ class CodeAutoGenerateEditBlock:
         }
 
     def single_round_run(
-        self, query: str, source_content: str
+        self, query: str, source_code_list: SourceCodeList
     ) -> CodeGenerateResult:
+        
+        # Apply model filter for code_llm
+        printer = Printer()
+        for llm in self.llms:
+            model_filter = ModelPathFilter.from_model_object(llm, self.args)
+            filtered_sources = []
+            for source in source_code_list.sources:
+                if model_filter.is_accessible(source.module_name): 
+                    filtered_sources.append(source)
+                else:
+                    printer.print_in_terminal("index_file_filtered", 
+                                               style="yellow",
+                                               file_path=source.module_name, 
+                                               model_name=",".join(llm_utils.get_llm_names(llm)))
+
+        source_code_list = SourceCodeList(filtered_sources)
+        
         llm_config = {"human_as_model": self.args.human_as_model}
 
+        source_content = source_code_list.to_str()
+
         if self.args.template == "common":
             init_prompt = self.single_round_instruction.prompt(
                 instruction=query, content=source_content, context=self.args.context
@@ -424,6 +446,11 @@ class CodeAutoGenerateEditBlock:
         input_tokens_count = 0
         generated_tokens_count = 0
 
+        input_tokens_cost = 0
+        generated_tokens_cost = 0
+
+        model_names = []
+
         printer = Printer()
         estimated_input_tokens = count_tokens(
             json.dumps(conversations, ensure_ascii=False))
@@ -437,14 +464,28 @@ class CodeAutoGenerateEditBlock:
             with ThreadPoolExecutor(max_workers=len(self.llms) * self.generate_times_same_model) as executor:
                 futures = []
                 for llm in self.llms:
+                    
+                    model_names_list = llm_utils.get_llm_names(llm) 
+                    model_name = None
+                    if model_names_list:
+                        model_name = model_names_list[0]                    
+
                     for _ in range(self.generate_times_same_model):
+                        model_names.append(model_name)
                         futures.append(executor.submit(
                             chat_with_continue, llm=llm, conversations=conversations, llm_config=llm_config))
+                
                 temp_results = [future.result() for future in futures]
-                for result in temp_results:
+                
+                for result,model_name in zip(temp_results,model_names):
                     results.append(result.content)
                     input_tokens_count += result.input_tokens_count
                     generated_tokens_count += result.generated_tokens_count
+                    model_info = llm_utils.get_model_info(model_name,self.args.product_mode)                    
+                    input_cost = model_info.get("input_price", 0) if model_info else 0
+                    output_cost = model_info.get("output_price", 0) if model_info else 0
+                    input_tokens_cost += input_cost * result.input_tokens_count / 1000000
+                    generated_tokens_cost += output_cost * result.generated_tokens_count / 1000000
 
             for result in results:
                 conversations_list.append(
@@ -461,7 +502,9 @@ class CodeAutoGenerateEditBlock:
 
         statistics = {
             "input_tokens_count": input_tokens_count,
-            "generated_tokens_count": generated_tokens_count
+            "generated_tokens_count": generated_tokens_count,
+            "input_tokens_cost": input_tokens_cost,
+            "generated_tokens_cost": generated_tokens_cost
         }
 
         if self.args.request_id and not self.args.skip_events:
@@ -476,10 +519,11 @@ class CodeAutoGenerateEditBlock:
         return CodeGenerateResult(contents=results, conversations=conversations_list, metadata=statistics)
 
     def multi_round_run(
-        self, query: str, source_content: str, max_steps: int = 10
+        self, query: str, source_code_list: SourceCodeList, max_steps: int = 10
     ) -> CodeGenerateResult:
         llm_config = {"human_as_model": self.args.human_as_model}
         result = []
+        source_content = source_code_list.to_str()
 
         if self.args.template == "common":
             init_prompt = self.multi_round_instruction.prompt(

autocoder/common/code_auto_generate_strict_diff.py

@@ -9,7 +9,9 @@ import json
 from autocoder.common.utils_code_auto_generate import chat_with_continue
 from autocoder.common.printer import Printer
 from autocoder.rag.token_counter import count_tokens
-
+from autocoder.utils import llms as llm_utils
+from autocoder.common import SourceCodeList
+from autocoder.privacy.model_filter import ModelPathFilter
 class CodeAutoGenerateStrictDiff:
     def __init__(
         self, llm: byzerllm.ByzerLLM, args: AutoCoderArgs, action=None
@@ -271,9 +273,10 @@ class CodeAutoGenerateStrictDiff:
         }
 
     def single_round_run(
-        self, query: str, source_content: str
+        self, query: str, source_code_list: SourceCodeList
     ) -> CodeGenerateResult:
         llm_config = {"human_as_model": self.args.human_as_model}
+        source_content = source_code_list.to_str()
 
         if self.args.template == "common":
             init_prompt = self.single_round_instruction.prompt(
@@ -311,6 +314,9 @@ class CodeAutoGenerateStrictDiff:
         results = []
         input_tokens_count = 0
         generated_tokens_count = 0
+        input_tokens_cost = 0
+        generated_tokens_cost = 0
+        model_names = []
 
         printer = Printer()
         estimated_input_tokens = count_tokens(json.dumps(conversations, ensure_ascii=False))
@@ -324,14 +330,27 @@ class CodeAutoGenerateStrictDiff:
                 futures = []
                 for llm in self.llms:
                     for _ in range(self.generate_times_same_model):
-                        futures.append(executor.submit(
-                            chat_with_continue, llm=llm, conversations=conversations, llm_config=llm_config))
+                        
+                        model_names_list = llm_utils.get_llm_names(llm)
+                        model_name = None
+                        if model_names_list:
+                            model_name = model_names_list[0]                                                    
+                        
+                        for _ in range(self.generate_times_same_model):
+                            model_names.append(model_name)
+                            futures.append(executor.submit(
+                                chat_with_continue, llm=llm, conversations=conversations, llm_config=llm_config))
+                            
                 temp_results = [future.result() for future in futures]
                 for result in temp_results:
                     results.append(result.content)
                     input_tokens_count += result.input_tokens_count
                     generated_tokens_count += result.generated_tokens_count
-            
+                    model_info = llm_utils.get_model_info(model_name, self.args.product_mode)
+                    input_cost = model_info.get("input_price", 0) if model_info else 0
+                    output_cost = model_info.get("output_price", 0) if model_info else 0
+                    input_tokens_cost += input_cost * result.input_tokens_count / 1000000
+                    generated_tokens_cost += output_cost * result.generated_tokens_count / 1000000
             for result in results:
                 conversations_list.append(
                     conversations + [{"role": "assistant", "content": result}])
@@ -345,7 +364,9 @@ class CodeAutoGenerateStrictDiff:
         
         statistics = {
             "input_tokens_count": input_tokens_count,
-            "generated_tokens_count": generated_tokens_count
+            "generated_tokens_count": generated_tokens_count,
+            "input_tokens_cost": input_tokens_cost,
+            "generated_tokens_cost": generated_tokens_cost
         }        
     
         if self.args.request_id and not self.args.skip_events:
@@ -360,10 +381,28 @@ class CodeAutoGenerateStrictDiff:
         return CodeGenerateResult(contents=results, conversations=conversations_list, metadata=statistics)
 
     def multi_round_run(
-        self, query: str, source_content: str, max_steps: int = 10
+        self, query: str, source_code_list: SourceCodeList, max_steps: int = 10
     ) -> CodeGenerateResult:
+        
+        # Apply model filter for code_llm
+        printer = Printer()
+        for llm in self.llms:
+            model_filter = ModelPathFilter.from_model_object(llm, self.args)
+            filtered_sources = []
+            for source in source_code_list.sources:
+                if model_filter.is_accessible(source.module_name): 
+                    filtered_sources.append(source)
+                else:
+                    printer.print_in_terminal("index_file_filtered", 
+                                               style="yellow",
+                                               file_path=source.module_name, 
+                                               model_name=",".join(llm_utils.get_llm_names(llm)))
+        
+        source_code_list = SourceCodeList(filtered_sources)
+        
         llm_config = {"human_as_model": self.args.human_as_model}
         result = []
+        source_content = source_code_list.to_str()
 
         if self.args.template == "common":
             init_prompt = self.multi_round_instruction.prompt(

autocoder/common/code_modification_ranker.py

@@ -8,8 +8,8 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
 import traceback
 from autocoder.common.utils_code_auto_generate import chat_with_continue
 from byzerllm.utils.str2model import to_model
+from autocoder.utils.llms import get_llm_names, get_model_info
 
-from autocoder.utils.llms import get_llm_names
 class RankResult(BaseModel):
     rank_result: List[int]
 
@@ -97,13 +97,42 @@ class CodeModificationRanker:
 
                 # Collect all results
                 results = []
-                for future in as_completed(futures):
+                # 获取模型名称列表
+                model_names = []
+                for llm in self.llms:
+                    # 获取当前llm实例对应的模型名称
+                    names = get_llm_names(llm)  
+                    model_names.extend(names)
+
+                # 获取模型价格信息
+                model_info_map = {}
+                for name in model_names:
+                    # 第二个参数是产品模式,从args中获取
+                    info = get_model_info(name, self.args.product_mode)  
+                    if info:
+                        model_info_map[name] = {
+                            "input_cost": info.get("input_price", 0.0),  # 每百万tokens成本
+                            "output_cost": info.get("output_price", 0.0) # 每百万tokens成本 
+                        }
+
+                # 计算总成本
+                total_input_cost = 0.0
+                total_output_cost = 0.0
+
+                for future, model_name in zip(futures, model_names):
                     try:
                         result = future.result()
                         input_tokens_count += result.input_tokens_count
                         generated_tokens_count += result.generated_tokens_count
                         v = to_model(result.content,RankResult)                        
                         results.append(v.rank_result)
+
+                        # 计算成本                        
+                        info = model_info_map.get(model_name, {})
+                        # 计算公式:token数 * 单价 / 1000000
+                        total_input_cost += (result.input_tokens_count * info.get("input_cost", 0.0)) / 1000000
+                        total_output_cost += (result.generated_tokens_count * info.get("output_cost", 0.0)) / 1000000
+
                     except Exception as e:
                         self.printer.print_in_terminal(
                             "ranking_failed_request", style="yellow", error=str(e))                        
@@ -113,6 +142,10 @@ class CodeModificationRanker:
                     raise Exception(
                         self.printer.get_message_from_key("ranking_all_failed"))
 
+                # 四舍五入到4位小数
+                total_input_cost = round(total_input_cost, 4)
+                total_output_cost = round(total_output_cost, 4)
+
                 # Calculate scores for each candidate
                 candidate_scores = defaultdict(float)
                 for rank_result in results:
@@ -137,7 +170,10 @@ class CodeModificationRanker:
                     best_candidate=sorted_candidates[0],
                     scores=score_details,
                     input_tokens=input_tokens_count,
-                    output_tokens=generated_tokens_count
+                    output_tokens=generated_tokens_count,
+                    input_cost=total_input_cost,
+                    output_cost=total_output_cost,
+                    model_names=", ".join(model_names)
                 )
 
                 rerank_contents = [generate_result.contents[i]

autocoder/dispacher/actions/action.py

@@ -27,6 +27,8 @@ from loguru import logger
 import time
 from autocoder.common.printer import Printer
 from autocoder.utils.llms import get_llm_names
+from autocoder.privacy.model_filter import ModelPathFilter
+from autocoder.common import SourceCodeList
 
 
 class BaseAction:    
@@ -55,14 +57,15 @@ class ActionTSProject(BaseAction):
         self.pp = pp
         pp.run()
 
-        source_code = pp.output()
+        # source_code = pp.output()
+        source_code_list = SourceCodeList(pp.sources)
         if self.llm:
             if args.in_code_apply:
                 old_query = args.query
                 args.query = (args.context or "") + "\n\n" + args.query
-            source_code = build_index_and_filter_files(
+            source_code_list = build_index_and_filter_files(
                 llm=self.llm, args=args, sources=pp.sources
-            )
+            )            
             if args.in_code_apply:
                 args.query = old_query
 
@@ -81,17 +84,21 @@ class ActionTSProject(BaseAction):
                 html_path=html_path,
                 max_iter=self.args.image_max_iter,
             )
-
+            html_code = ""
             with open(html_path, "r") as f:
                 html_code = f.read()
-                source_code = f"##File: {html_path}\n{html_code}\n\n" + source_code
+            
+            source_code_list.sources.append(SourceCode( 
+                module_name=html_path,
+                source_code=html_code,
+                tag="IMAGE"))            
 
-        self.process_content(source_code)
+        self.process_content(source_code_list)
         return True
 
-    def process_content(self, content: str):
+    def process_content(self, source_code_list: SourceCodeList):
         args = self.args
-
+        content = source_code_list.to_str()
         if args.execute and self.llm and not args.human_as_model:
             content_length = self._get_content_length(content)
             if content_length > self.args.model_max_input_length:
@@ -116,22 +123,27 @@ class ActionTSProject(BaseAction):
                 )
             else:
                 generate = CodeAutoGenerate(llm=self.llm, args=self.args, action=self)
+                        
             if self.args.enable_multi_round_generate:
                 generate_result = generate.multi_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
             else:
                 generate_result = generate.single_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
             elapsed_time = time.time() - start_time
-            speed = generate_result.metadata.get('generated_tokens_count', 0) / elapsed_time if elapsed_time > 0 else 0            
-            model_names = ",".join(get_llm_names(self.llm))
+            speed = generate_result.metadata.get('generated_tokens_count', 0) / elapsed_time if elapsed_time > 0 else 0
+            input_tokens_cost = generate_result.metadata.get('input_tokens_cost', 0)
+            generated_tokens_cost = generate_result.metadata.get('generated_tokens_cost', 0)
+            model_names = ",".join(get_llm_names(generate.llms))
             self.printer.print_in_terminal(
                 "code_generation_complete",
                 duration=elapsed_time,
                 input_tokens=generate_result.metadata.get('input_tokens_count', 0),
                 output_tokens=generate_result.metadata.get('generated_tokens_count', 0),
+                input_cost=input_tokens_cost,
+                output_cost=generated_tokens_cost,
                 speed=round(speed, 2),
                 model_names=model_names
             )
@@ -187,11 +199,12 @@ class ActionPyScriptProject(BaseAction):
         pp = Level1PyProject(
             script_path=args.script_path, package_name=args.package_name
         )
-        content = pp.run()
-        self.process_content(content)
+        pp.run()
+        source_code_list = SourceCodeList(pp.sources)
+        self.process_content(source_code_list)
         return True
 
-    def process_content(self, content: str):
+    def process_content(self, source_code_list: SourceCodeList):
         args = self.args
         if args.execute:
             self.printer.print_in_terminal("code_generation_start")
@@ -212,21 +225,25 @@ class ActionPyScriptProject(BaseAction):
                 generate = CodeAutoGenerate(llm=self.llm, args=self.args, action=self)
             if self.args.enable_multi_round_generate:
                 generate_result = generate.multi_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
             else:
                 generate_result = generate.single_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
 
             elapsed_time = time.time() - start_time
             speed = generate_result.metadata.get('generated_tokens_count', 0) / elapsed_time if elapsed_time > 0 else 0
-            model_names = ",".join(get_llm_names(self.llm))
+            model_names = ",".join(get_llm_names(generate.llms))
+            input_tokens_cost = generate_result.metadata.get('input_tokens_cost', 0)
+            generated_tokens_cost = generate_result.metadata.get('generated_tokens_cost', 0)
             self.printer.print_in_terminal(
                 "code_generation_complete",
                 duration=elapsed_time,
                 input_tokens=generate_result.metadata.get('input_tokens_count', 0),
                 output_tokens=generate_result.metadata.get('generated_tokens_count', 0),
+                input_cost=input_tokens_cost,
+                output_cost=generated_tokens_cost,
                 speed=round(speed, 2),
                 model_names=model_names
             )
@@ -264,13 +281,7 @@ class ActionPyScriptProject(BaseAction):
                     model=self.llm.default_model_name,
                 )
 
-            end_time = time.time()
-            self.printer.print_in_terminal(
-                "code_generation_complete",
-                duration=end_time - start_time,
-                input_tokens=generate_result.metadata.get('input_tokens_count', 0),
-                output_tokens=generate_result.metadata.get('generated_tokens_count', 0)
-            )
+            end_time = time.time()            
             with open(self.args.target_file, "w") as file:
                 file.write(content)
 
@@ -291,24 +302,24 @@ class ActionPyProject(BaseAction):
         pp = PyProject(args=self.args, llm=self.llm)
         self.pp = pp
         pp.run(packages=args.py_packages.split(",") if args.py_packages else [])            
-        source_code = pp.output()                
+        source_code_list = SourceCodeList(pp.sources)                
 
         if self.llm:
             old_query = args.query
             if args.in_code_apply:
                 args.query = (args.context or "") + "\n\n" + args.query
-            source_code = build_index_and_filter_files(
+            source_code_list = build_index_and_filter_files(
                 llm=self.llm, args=args, sources=pp.sources
             )
             if args.in_code_apply:
                 args.query = old_query
 
-        self.process_content(source_code)
+        self.process_content(source_code_list)
         return True
 
-    def process_content(self, content: str):
+    def process_content(self, source_code_list: SourceCodeList):
         args = self.args
-
+        content = source_code_list.to_str()
         if args.execute and self.llm and not args.human_as_model:
             content_length = self._get_content_length(content)
             if content_length > self.args.model_max_input_length:
@@ -340,20 +351,24 @@ class ActionPyProject(BaseAction):
 
             if self.args.enable_multi_round_generate:
                 generate_result = generate.multi_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
             else:
                 generate_result = generate.single_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
             elapsed_time = time.time() - start_time
             speed = generate_result.metadata.get('generated_tokens_count', 0) / elapsed_time if elapsed_time > 0 else 0
-            model_names = ",".join(get_llm_names(self.llm))
+            model_names = ",".join(get_llm_names(generate.llms))
+            input_tokens_cost = generate_result.metadata.get('input_tokens_cost', 0)
+            generated_tokens_cost = generate_result.metadata.get('generated_tokens_cost', 0)
             self.printer.print_in_terminal(
                 "code_generation_complete",
                 duration=elapsed_time,
                 input_tokens=generate_result.metadata.get('input_tokens_count', 0),
                 output_tokens=generate_result.metadata.get('generated_tokens_count', 0),
+                input_cost=input_tokens_cost,
+                output_cost=generated_tokens_cost,
                 speed=round(speed, 2),
                 model_names=model_names
             )
@@ -408,20 +423,21 @@ class ActionSuffixProject(BaseAction):
         pp = SuffixProject(args=args, llm=self.llm)
         self.pp = pp
         pp.run()
-        source_code = pp.output()
+        source_code_list = SourceCodeList(pp.sources)
         if self.llm:
             if args.in_code_apply:
                 old_query = args.query
                 args.query = (args.context or "") + "\n\n" + args.query
-            source_code = build_index_and_filter_files(
+            source_code_list = build_index_and_filter_files(
                 llm=self.llm, args=args, sources=pp.sources
             )
             if args.in_code_apply:
                 args.query = old_query
-        self.process_content(source_code)
+        self.process_content(source_code_list)
 
-    def process_content(self, content: str):
+    def process_content(self, source_code_list: SourceCodeList):
         args = self.args
+        content = source_code_list.to_str()
 
         if args.execute and self.llm and not args.human_as_model:
             content_length = self._get_content_length(content)
@@ -449,21 +465,25 @@ class ActionSuffixProject(BaseAction):
                 generate = CodeAutoGenerate(llm=self.llm, args=self.args, action=self)
             if self.args.enable_multi_round_generate:
                 generate_result = generate.multi_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
             else:
                 generate_result = generate.single_round_run(
-                    query=args.query, source_content=content
+                    query=args.query, source_code_list=source_code_list
                 )
               
         elapsed_time = time.time() - start_time
         speed = generate_result.metadata.get('generated_tokens_count', 0) / elapsed_time if elapsed_time > 0 else 0
-        model_names = ",".join(get_llm_names(self.llm))
+        model_names = ",".join(get_llm_names(generate.llms))
+        input_tokens_cost = generate_result.metadata.get('input_tokens_cost', 0)
+        generated_tokens_cost = generate_result.metadata.get('generated_tokens_cost', 0)
         self.printer.print_in_terminal(
             "code_generation_complete",
             duration=elapsed_time,
             input_tokens=generate_result.metadata.get('input_tokens_count', 0),
             output_tokens=generate_result.metadata.get('generated_tokens_count', 0),
+            input_cost=input_tokens_cost,
+            output_cost=generated_tokens_cost,
             speed=round(speed, 2),
             model_names=model_names
         )