auto-coder 0.1.316__py3-none-any.whl → 0.1.317__py3-none-any.whl

autocoder/common/v2/code_auto_generate_editblock.py

@@ -0,0 +1,380 @@
+from typing import List, Dict, Tuple
+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
+import json
+from concurrent.futures import ThreadPoolExecutor
+from autocoder.common.utils_code_auto_generate import chat_with_continue,stream_chat_with_continue,ChatWithContinueResult
+from autocoder.utils.auto_coder_utils.chat_stream_out import stream_out
+from autocoder.common.stream_out_type import CodeGenerateStreamOutType
+from autocoder.common.auto_coder_lang import get_message_with_format
+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.memory.active_context_manager import ActiveContextManager
+
+
+
+class CodeAutoGenerateEditBlock:
+    def __init__(
+        self,
+        llm: byzerllm.ByzerLLM,
+        args: AutoCoderArgs,
+        action=None,
+        fence_0: str = "```",
+        fence_1: str = "```",
+    ) -> None:
+        self.llm = llm
+        self.args = args
+        self.action = action
+        self.fence_0 = fence_0
+        self.fence_1 = fence_1
+        self.generate_times_same_model = args.generate_times_same_model
+        if not self.llm:
+            raise ValueError(
+                "Please provide a valid model instance to use for code generation."
+            )
+        self.llms = self.llm.get_sub_client("code_model") or [self.llm]
+        if not isinstance(self.llms, list):
+            self.llms = [self.llms]
+    
+
+    @byzerllm.prompt()
+    def single_round_instruction(self, instruction: str, 
+                                 content: str, 
+                                 context: str = "",
+                                 package_context: str = ""
+                                 ) -> str:
+        """
+        如果你需要生成代码，对于每个需要更改的文件,你需要按 *SEARCH/REPLACE block* 的格式进行生成。
+
+        # *SEARCH/REPLACE block* Rules:
+
+        Every *SEARCH/REPLACE block* must use this format:
+        1. The opening fence and code language, eg: {{ fence_0 }}python
+        2. The file path alone on a line, starting with "##File:" and verbatim. No bold asterisks, no quotes around it, no escaping of characters, etc.
+        3. The start of search block: <<<<<<< SEARCH
+        4. A contiguous chunk of lines to search for in the existing source code
+        5. The dividing line: =======
+        6. The lines to replace into the source code
+        7. The end of the replace block: >>>>>>> REPLACE
+        8. The closing fence: {{ fence_1 }}
+
+        Every *SEARCH* section must *EXACTLY MATCH* the existing source code, character for character, including all comments, docstrings, etc.
+
+        *SEARCH/REPLACE* blocks will replace *all* matching occurrences.
+        Include enough lines to make the SEARCH blocks unique.
+
+        Include *ALL* the code being searched and replaced!
+
+        To move code within a file, use 2 *SEARCH/REPLACE* blocks: 1 to delete it from its current location, 1 to insert it in the new location.
+
+        If you want to put code in a new file, use a *SEARCH/REPLACE block* with:
+        - A new file path, including dir name if needed
+        - An empty `SEARCH` section
+        - The new file's contents in the `REPLACE` section
+
+        ONLY EVER RETURN CODE IN A *SEARCH/REPLACE BLOCK*!
+
+        下面我们来看一个例子：
+
+        当前项目目录结构：
+        1. 项目根目录： /tmp/projects/mathweb
+        2. 项目子目录/文件列表(类似tree 命令输出)
+        flask/
+            app.py
+            templates/
+                index.html
+            static/
+                style.css
+
+        用户需求： Change get_factorial() to use math.factorial
+
+        回答： To make this change we need to modify `/tmp/projects/mathweb/flask/app.py` to:
+
+        1. Import the math package.
+        2. Remove the existing factorial() function.
+        3. Update get_factorial() to call math.factorial instead.
+
+        Here are the *SEARCH/REPLACE* blocks:
+
+        {{ fence_0 }}python
+        ##File: /tmp/projects/mathweb/flask/app.py
+        <<<<<<< SEARCH
+        from flask import Flask
+        =======
+        import math
+        from flask import Flask
+        >>>>>>> REPLACE
+        {{ fence_1 }}
+
+        {{ fence_0 }}python
+        ##File: /tmp/projects/mathweb/flask/app.py
+        <<<<<<< SEARCH
+        def factorial(n):
+            "compute factorial"
+
+            if n == 0:
+                return 1
+            else:
+                return n * factorial(n-1)
+
+        =======
+        >>>>>>> REPLACE
+        {{ fence_1 }}
+
+        {{ fence_0 }}python
+        ##File: /tmp/projects/mathweb/flask/app.py
+        <<<<<<< SEARCH
+            return str(factorial(n))
+        =======
+            return str(math.factorial(n))
+        >>>>>>> REPLACE
+        {{ fence_1 }}
+
+        用户需求： Refactor hello() into its own file.
+
+        回答：To make this change we need to modify `main.py` and make a new file `hello.py`:
+
+        1. Make a new hello.py file with hello() in it.
+        2. Remove hello() from main.py and replace it with an import.
+
+        Here are the *SEARCH/REPLACE* blocks:
+
+        {{ fence_0 }}python
+        ##File: /tmp/projects/mathweb/hello.py
+        <<<<<<< SEARCH
+        =======
+        def hello():
+            "print a greeting"
+
+            print("hello")
+        >>>>>>> REPLACE
+        {{ fence_1 }}
+
+        {{ fence_0 }}python
+        ##File: /tmp/projects/mathweb/main.py
+        <<<<<<< SEARCH
+        def hello():
+            "print a greeting"
+
+            print("hello")
+        =======
+        from hello import hello
+        >>>>>>> REPLACE
+        {{ fence_1 }}
+
+        现在让我们开始一个新的任务:
+
+        {%- if structure %}
+        {{ structure }}
+        {%- endif %}
+
+        {%- if content %}
+        下面是一些文件路径以及每个文件对应的源码：
+        <files>
+        {{ content }}
+        </files>
+        {%- endif %}
+
+        {%- if package_context %}
+        下面是上面文件的一些信息（包括最近的变更情况）：
+        <package_context>
+        {{ package_context }}
+        </package_context>
+        {%- endif %}
+
+        {%- if context %}
+        <extra_context>
+        {{ context }}
+        </extra_context>
+        {%- endif %}     
+
+        下面是用户的需求：
+
+        {{ instruction }}
+
+        """
+        if not self.args.include_project_structure:
+            return {
+                "structure": "",
+                "fence_0": self.fence_0,
+                "fence_1": self.fence_1,
+            }
+
+        return {
+            "structure": (
+                self.action.pp.get_tree_like_directory_structure()
+                if self.action
+                else ""
+            ),
+            "fence_0": self.fence_0,
+            "fence_1": self.fence_1,
+        }
+
+    def single_round_run(
+        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()
+
+        active_context_manager = ActiveContextManager(self.llm, self.args.source_dir)
+       
+        # 获取包上下文信息
+        package_context = ""
+        
+        if self.args.enable_active_context:
+            # 获取活动上下文信息
+            result = active_context_manager.load_active_contexts_for_files(
+                [source.module_name for source in source_code_list.sources]
+            )
+            # 将活动上下文信息格式化为文本
+            if result.contexts:
+                package_context_parts = []
+                for dir_path, context in result.contexts.items():
+                    package_context_parts.append(f"<package_info>{context.content}</package_info>")
+                
+                package_context = "\n".join(package_context_parts)
+
+        init_prompt = self.single_round_instruction.prompt(
+            instruction=query, content=source_content, context=self.args.context, 
+            package_context=package_context
+        )        
+
+        with open(self.args.target_file, "w",encoding="utf-8") as file:
+            file.write(init_prompt)
+
+        conversations = []
+
+        if self.args.system_prompt and self.args.system_prompt.strip() == "claude":
+            conversations.append(
+                {"role": "system", "content": sys_prompt.claude_sys_prompt.prompt()})
+        elif self.args.system_prompt:
+            conversations.append(
+                {"role": "system", "content": self.args.system_prompt})
+
+        conversations.append({"role": "user", "content": init_prompt})        
+
+        conversations_list = []
+        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))
+        printer.print_in_terminal("estimated_input_tokens_in_generate",
+                                  style="yellow",
+                                  estimated_input_tokens_in_generate=estimated_input_tokens,
+                                  generate_mode="editblock"
+                                  )
+
+        if not self.args.human_as_model:
+            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 i in range(self.generate_times_same_model):
+                        model_names.append(model_name)
+                        if i==0:
+                            def job():
+                                stream_generator = stream_chat_with_continue(
+                                    llm=llm, 
+                                    conversations=conversations, 
+                                    llm_config=llm_config,
+                                    args=self.args
+                                )
+                                full_response, last_meta = stream_out(
+                                stream_generator,
+                                model_name=model_name,
+                                title=get_message_with_format(
+                                    "code_generate_title", model_name=model_name),
+                                args=self.args,
+                                extra_meta={
+                                    "stream_out_type": CodeGenerateStreamOutType.CODE_GENERATE.value
+                                })
+                                return ChatWithContinueResult(
+                                    content=full_response,
+                                    input_tokens_count=last_meta.input_tokens_count,
+                                    generated_tokens_count=last_meta.generated_tokens_count
+                                )
+                            futures.append(executor.submit(job))
+                        else:                                
+                            futures.append(executor.submit(
+                                chat_with_continue, 
+                                llm=llm, 
+                                conversations=conversations, 
+                                llm_config=llm_config,
+                                args=self.args
+                            ))
+                
+                temp_results = [future.result() for future in futures]
+                
+                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(
+                    conversations + [{"role": "assistant", "content": result}])
+        else:
+            for _ in range(self.args.human_model_num):
+                single_result = chat_with_continue(
+                    llm=self.llms[0], 
+                    conversations=conversations, 
+                    llm_config=llm_config,
+                    args=self.args
+                )
+                results.append(single_result.content)
+                input_tokens_count += single_result.input_tokens_count
+                generated_tokens_count += single_result.generated_tokens_count
+                conversations_list.append(
+                    conversations + [{"role": "assistant", "content": single_result.content}])
+
+        statistics = {
+            "input_tokens_count": input_tokens_count,
+            "generated_tokens_count": generated_tokens_count,
+            "input_tokens_cost": input_tokens_cost,
+            "generated_tokens_cost": generated_tokens_cost
+        }
+
+        return CodeGenerateResult(contents=results, conversations=conversations_list, metadata=statistics)
+
+    

autocoder/common/v2/code_auto_generate_strict_diff.py

@@ -0,0 +1,269 @@
+from typing import List, Dict, Tuple
+from autocoder.common.types import Mode, CodeGenerateResult
+from autocoder.common import AutoCoderArgs
+import byzerllm
+from autocoder.common import sys_prompt
+from autocoder.common.v2.code_auto_generate import CodeAutoGenerate
+from autocoder.common import SourceCodeList
+
+class CodeAutoGenerateStrictDiff(CodeAutoGenerate):
+    """
+    A class that handles code generation in strict diff format.
+    """
+
+    @byzerllm.prompt(llm=lambda self: self.llm)
+    def single_round_instruction(
+        self, instruction: str, content: str, context: str = "", package_context: str = ""
+    ) -> str:
+        """
+        如果你需要生成代码，对于每个需要更改的文件，写出类似于 unified diff 的更改，就像`diff -U0`会产生的那样。
+        下面是一些生成diff的要求：
+        Make sure you include the first 2 lines with the file paths.
+        Don't include timestamps with the file paths.
+
+        Start each hunk of changes with a `@@ ... @@` line.
+        Must include line numbers like `diff -U0` does.
+        The user's patch tool need them.
+
+        The user's patch tool needs CORRECT patches that apply cleanly against the current contents of the file!
+        Think carefully and make sure you include and mark all lines that need to be removed or changed as `-` lines.
+        Make sure you mark all new or modified lines with `+`.
+        Don't leave out any lines or the diff patch won't apply correctly.
+
+        Indentation matters in the diffs!
+
+        To make a new file, show a diff from `--- /dev/null` to `+++ path/to/new/file.ext`.
+        The code part of the diff content should not contains any line number.
+
+        The path start with `---` or `+++` should be the absolute path of the file or relative path from the project root.
+
+        下面我们来看一个例子：
+
+        当前项目目录结构：
+        1. 项目根目录： /tmp/projects/mathweb
+        2. 项目子目录/文件列表(类似tree 命令输出)
+        flask/
+            app.py
+            templates/
+                index.html
+            static/
+                style.css
+
+        用户需求： 请将下面的代码中的is_prime()函数替换为sympy。
+        回答：
+        好的，我会先罗列出需要的修改步骤，然后再列出diff。
+        修改步骤：
+        1. 添加sympy的import 语句。
+        2. 删除is_prime()函数。
+        3. 将现有对is_prime()的调用替换为sympy.isprime()。
+
+        下面是这些变更的diff：
+
+        ```diff
+        --- /tmp/projects/mathweb/flask/app.py
+        +++ /tmp/projects/mathweb/flask/app.py
+        @@ ... @@
+        -class MathWeb:
+        +import sympy
+        +
+        +class MathWeb:
+        @@ ... @@
+        -def is_prime(x):
+        -    if x < 2:
+        -        return False
+        -    for i in range(2, int(math.sqrt(x)) + 1):
+        -        if x % i == 0:
+        -            return False
+        -    return True
+        @@ ... @@
+        -@app.route('/prime/<int:n>')
+        -def nth_prime(n):
+        -    count = 0
+        -    num = 1
+        -    while count < n:
+        -        num += 1
+        -        if is_prime(num):
+        -            count += 1
+        -    return str(num)
+        +@app.route('/prime/<int:n>')
+        +def nth_prime(n):
+        +    count = 0
+        +    num = 1
+        +    while count < n:
+        +        num += 1
+        +        if sympy.isprime(num):
+        +            count += 1
+        +    return str(num)
+        ```
+
+        现在让我们开始一个新的任务:
+
+        {%- if structure %}
+        {{ structure }}
+        {%- endif %}
+
+        {%- if content %}
+        下面是一些文件路径以及每个文件对应的源码：
+        <files>
+        {{ content }}
+        </files>
+        {%- endif %}
+
+        {%- if package_context %}
+        下面是上面文件的一些信息（包括最近的变更情况）：
+        <package_context>
+        {{ package_context }}
+        </package_context>
+        {%- endif %}
+
+        {%- if context %}
+        <extra_context>
+        {{ context }}
+        </extra_context>
+        {%- endif %}
+
+        下面是用户的需求：
+
+        {{ instruction }}
+        """
+        
+        if not self.args.include_project_structure:
+            return {
+                "structure": "",                
+            }
+
+        return {
+            "structure": (
+                self.action.pp.get_tree_like_directory_structure()
+                if self.action
+                else ""
+            )
+        }
+
+    @byzerllm.prompt(llm=lambda self: self.llm)
+    def multi_round_instruction(
+        self, instruction: str, content: str, context: str = "", package_context: str = ""
+    ) -> str:
+        """
+        如果你需要生成代码，对于每个需要更改的文件，写出类似于 unified diff 的更改，就像`diff -U0`会产生的那样。
+        下面是一些生成diff的要求：
+        Make sure you include the first 2 lines with the file paths.
+        Don't include timestamps with the file paths.
+
+        Start each hunk of changes with a `@@ ... @@` line.
+        Must include line numbers like `diff -U0` does.
+        The user's patch tool need them.
+
+        The user's patch tool needs CORRECT patches that apply cleanly against the current contents of the file!
+        Think carefully and make sure you include and mark all lines that need to be removed or changed as `-` lines.
+        Make sure you mark all new or modified lines with `+`.
+        Don't leave out any lines or the diff patch won't apply correctly.
+
+        Indentation matters in the diffs!
+
+        To make a new file, show a diff from `--- /dev/null` to `+++ path/to/new/file.ext`.
+
+        下面我们来看一个例子：
+
+        当前项目目录结构：
+        1. 项目根目录： /tmp/projects/mathweb
+        2. 项目子目录/文件列表(类似tree 命令输出)
+        flask/
+            app.py
+            templates/
+                index.html
+            static/
+                style.css
+
+        用户需求： 请将下面的代码中的is_prime()函数替换为sympy。
+        回答：
+        好的，我会先罗列出需要的修改步骤，然后再列出diff。
+        修改步骤：
+        1. 添加sympy的import 语句。
+        2. 删除is_prime()函数。
+        3. 将现有对is_prime()的调用替换为sympy.isprime()。
+
+        下面是这些变更的diff：
+
+        ```diff
+        --- /tmp/projects/mathweb/flask/app.py
+        +++ /tmp/projects/mathweb/flask/app.py
+        @@ ... @@
+        -class MathWeb:
+        +import sympy
+        +
+        +class MathWeb:
+        @@ ... @@
+        -def is_prime(x):
+        -    if x < 2:
+        -        return False
+        -    for i in range(2, int(math.sqrt(x)) + 1):
+        -        if x % i == 0:
+        -            return False
+        -    return True
+        @@ ... @@
+        -@app.route('/prime/<int:n>')
+        -def nth_prime(n):
+        -    count = 0
+        -    num = 1
+        -    while count < n:
+        -        num += 1
+        -        if is_prime(num):
+        -            count += 1
+        -    return str(num)
+        +@app.route('/prime/<int:n>')
+        +def nth_prime(n):
+        +    count = 0
+        +    num = 1
+        +    while count < n:
+        +        num += 1
+        +        if sympy.isprime(num):
+        +            count += 1
+        +    return str(num)
+        ```
+
+        现在让我们开始一个新的任务:
+
+        {%- if structure %}
+        {{ structure }}
+        {%- endif %}
+
+        {%- if content %}
+        下面是一些文件路径以及每个文件对应的源码：
+        <files>
+        {{ content }}
+        </files>
+        {%- endif %}
+
+        {%- if package_context %}
+        下面是上面文件的一些信息（包括最近的变更情况）：
+        <package_context>
+        {{ package_context }}
+        </package_context>
+        {%- endif %}
+
+        {%- if context %}
+        <extra_context>
+        {{ context }}
+        </extra_context>
+        {%- endif %}
+
+        下面是用户的需求：
+
+        {{ instruction }}
+
+        每次生成一个文件的diff，然后询问我是否继续，当我回复继续，继续生成下一个文件的diff。当没有后续任务时，请回复 "__完成__" 或者 "__EOF__"。
+        """
+        
+        if not self.args.include_project_structure:
+            return {
+                "structure": "",                
+            }
+
+        return {
+            "structure": (
+                self.action.pp.get_tree_like_directory_structure()
+                if self.action
+                else ""
+            )
+        }