gyt/opendan
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix text summary build bug

Browse Source
This commit is contained in:
Liu Zhicong
2023-11-14 16:49:36 -08:00
parent 5fe3073cb6
commit 87fdba9714
4 changed files with 206 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/aios_kernel/agent.py
+119 -47
View File
@@ -21,6 +21,7 @@ from .contact_manager import ContactManager,Contact,FamilyMember
from .compute_kernel import ComputeKernel
from .bus import AIBus
from .workspace_env import WorkspaceEnvironment
from .storage import AIStorage
from knowledge import *
@@ -53,11 +54,36 @@ DEFAULT_AGENT_GOAL_TO_TODO_PROMPT = """
"""
DEFAULT_AGENT_LEARN_PROMPT = """
你拥有非常优秀的资料整理技能。我给你一段内容，你会尝试对其进行摘要，并在已有的资料库中找到合适的位置存放该文章。
1. 结合你的角色和组织的工作目标构建摘要，尽量精简，长度不要超过256个字
2. 资料库以文件系统的形式组织，浏览知识库是成本高昂的操作，应尝试从根目录往子目录深入来找到最合适的信息。必要的情况下，你可以在合适的位置创建新的目录。为了方便浏览，每一层目录的文件夹数不超过32个，名称长度不超过16个字符，目录深度不超过6层
3. 你可以从不同的角度给出最多3个合适的位置
4. 返回一个json来保存摘要和建议保存位置信息
我是一名软件工程师，拥有非常优秀的资料学习能力。下面是我学习和整理资料的方法
1. 结合我的角色为资料产生长度不超过256个Token的摘要;尝试产生不超过16个tag;
2. 现有资料库以文件系统的形式组织，我未来借助资料的摘要来浏览知识库
3. 我将学习过的资料另存在资料库的合适位置（以/开始的完整路径）。保存位置的目录深度不超过5层，文件夹名称长度不超过16个字符。
4. 当存在已知信息时，需参考已知信息的内容来思考结果。
5. 由于LLM的Token限制，我学习的可能只是资料的部分内容，此时我应能产生合适的中间结果，中间结果保存在metadata中。当我决定构建中间结果时，我只构建中间结果。
6. 当我收到最后一部分内容时，我能结合已知的中间结果产生最终结果。
7. 总是以json格式返回思考结果，json格式如下
{
think:"$think_result",
metadata:{...} , # temp result for long content
tags:["tag1","tag2"...],
path:["/graphic/opengl","/database/mysql"], # list of directories to save to.
title:"$article_title",
summary:"$summary",
catalogs: [{ # optional,catalogs is a tree
title:"$catalog_name1",
pos:"$pos:$length"
children:[
{
title:"$catalog_name 1.1",
pos:"$pos:$length"
}
]},
{
title:"$catalog_name2",
pos:"$pos:$length"
}
]
}
"""
DEFAULT_AGENT_LEARN_LONG_CONENT_PROMPT = """
@@ -125,7 +151,7 @@ class AIAgent:
self.goal_to_todo_prompt = None
self.learn_token_limit = 500
self.learn_token_limit = 4000
self.learn_prompt = AgentPrompt(DEFAULT_AGENT_LEARN_PROMPT)
self.chat_db = None
@@ -217,6 +243,9 @@ class AIAgent:
return self.template_id
def get_llm_model_name(self) -> str:
if self.llm_model_name is None:
return AIStorage.get_instance().get_user_config().get_value("llm_model_name")
return self.llm_model_name
def get_max_token_size(self) -> int:
@@ -889,22 +918,32 @@ class AIAgent:
if self.agent_energy <= 0:
break
knowledge = workspace.kb_db.get_knowledge_by_hash(hash)
knowledge = workspace.kb_db.get_knowledge(hash)
if knowledge is None:
continue
if os.path.exists(knowledge.path) is False:
logger.warning(f"do_self_learn: knowledge {knowledge.path} is not exists!")
full_path = knowledge.get("full_path")
if full_path is None:
continue
if os.path.exists(full_path) is False:
logger.warning(f"do_self_learn: knowledge {full_path} is not exists!")
continue
#TODO 可以用v-db 对不同目录的名字进行选择后，先进行一次快速的插入。有时间再慢慢用LLM整理
llm_result = await self._llm_read_article(knowledge)
result_obj = await self._llm_read_article(knowledge,full_path)
#根据结果更新knowledge
if llm_result is not None:
workspace.kb_db.update_knowledge_by_hash(hash,llm_result)
if result_obj is not None:
workspace.kb_db.set_knowledge_llm_result(hash,result_obj)
# 在知识库中创建软链接
path_list = result_obj.get("path")
new_title = result_obj.get("title")
if path_list:
for new_path in path_list:
full_new_path = f"/knowledge{new_path}/{new_title}"
await workspace.symlink(full_path,full_new_path)
logger.info(f"create soft link {full_path} -> {full_new_path}")
self.agent_energy -= 1
@@ -958,13 +997,11 @@ class AIAgent:
def parser_learn_llm_result(self,llm_result:LLMResult):
pass
async def gen_known_info_for_knowledge_prompt(self,knowledge_item:dict,need_catalogs = False) -> AgentPrompt:
#已知信息：
# 组织的工作总结（如有）待完成
# 现在知识库的结构（注意大小控制）gen_kb_tree_prompt (当为空的时候应该让LLM生成一个合适的初始目录结构)
# 原始路径，现在标题，摘要，目录
async def gen_known_info_for_knowledge_prompt(self,knowledge_item:dict,temp_meta = None,need_catalogs = False) -> AgentPrompt:
workspace =self.get_workspace_by_msg(None)
kb_tree = await workspace.get_knowledege_catalog()
known_obj = {}
title = knowledge_item.get("title")
if title:
@@ -979,37 +1016,41 @@ class AIAgent:
catalogs = knowledge_item.get("catalogs")
if catalogs:
known_obj["catalogs"] = catalogs
org_path = knowledge_item.get("path")
if temp_meta:
for key in temp_meta.keys():
known_obj[key] = temp_meta[key]
org_path = knowledge_item.get("full_path")
known_obj["orginal_path"] = org_path
know_info_str = f"# Known information\n{json.dumps(known_obj)}\n"
know_info_str = f"# Known information:\n## Current directory structure:\n{kb_tree}\n## Knowlege Metadata:\n{json.dumps(known_obj)}\n"
return AgentPrompt(know_info_str)
async def _llm_read_article(self,knowledge_item:dict,full_path:str) -> ComputeTaskResult:
# Objectives:
# Obtain better titles, abstracts, table of contents (if necessary), tags
# Determine the appropriate place to put it (in line with the organization's goals)
# Known information:
# The reason why the target service's learn_prompt is being sorted
# Summary of the organization's work (if any)
# The current structure of the knowledge base (note the size control) gen_kb_tree_prompt (when empty, LLM should generate an appropriate initial directory structure)
# Original path, current title, abstract, table of contents
# Sorting long files (general tricks)
# Indicate that the input is part of the content, let LLM generate intermediate results for the task
# Enter the content in sequence, when the last content block is input, LLM gets the result
async def _llm_read_article(self,knowledge_item:dict) -> ComputeTaskResult:
#目标：
# 得到更好的标题，摘要，目录 （如有必要）,tags
# 应放的合适的位置 （结合组织的目标）
#已知信息：
# 整理是为什么目标服务的 learn_prompt
# 组织的工作总结（如有）
# 现在知识库的结构（注意大小控制）gen_kb_tree_prompt (当为空的时候应该让LLM生成一个合适的初始目录结构)
# 原始路径，现在标题，摘要，目录
# 整理长文件（通用技巧）
# 告诉输入的是部分内容，让LLM为任务产生中间结果
# 依次输入内容，在最后一个内容块输入时，LLM得到结果
#full_content = item.get_article_full_content()
workspace = self.get_workspace_by_msg(None)
full_content = await workspace.load_knowledge_content(knowledge_item["hash"])
full_content = await workspace.load_knowledge_content(full_path)
if full_content is None:
return
return None
if len(full_content) < 16:
logger.warning(f"llm_read_article: article {knowledge_item['path']} is too short,just read summary!")
return None
full_content_len = self.token_len(full_content)
if full_content_len < self.get_llm_learn_token_limit():
@@ -1021,9 +1062,9 @@ class AIAgent:
prompt.append(known_info_prompt)
content_prompt = AgentPrompt(full_content)
prompt.append(content_prompt)
env_functions = workspace.get_knowledge_base_ai_functions()
task_result:ComputeTaskResult = await self._do_llm_complection(prompt,env_functions)
env_functions = None
#env_functions,function_len = workspace.get_knowledge_base_ai_functions()
task_result:ComputeTaskResult = await self._do_llm_complection(prompt,env_functions,None,True)
if task_result.result_code != ComputeTaskResultCode.OK:
result_obj = {}
result_obj["error_str"] = task_result.error_str
@@ -1033,10 +1074,41 @@ class AIAgent:
return result_obj
else:
logger.warning(f"llm_read_article: article {knowledge_item['path']} is too long,just read summary!")
result_obj = {}
result_obj["error_str"] = f"llm_read_article: article {knowledge_item['path']} is too long,just read summary!"
return result_obj
logger.warning(f"llm_read_article: article {full_path} use LLM loop learn!")
pos = 0
read_len = int(self.get_llm_learn_token_limit() * 1.5)
temp_meta_data = {}
is_final = False
while pos < full_content_len:
_content = full_content[pos:pos+read_len]
if len(_content) < read_len:
# last chunk
is_final = True
part_content = f"<<Final Part:start at {pos}>>\n{_content}"
else:
part_content = f"<<Part:start at {pos}>>\n{_content}"
pos = pos + read_len
prompt = self.get_learn_prompt()
known_info_prompt = await self.gen_known_info_for_knowledge_prompt(knowledge_item,temp_meta_data)
prompt.append(known_info_prompt)
content_prompt = AgentPrompt(part_content)
prompt.append(content_prompt)
env_functions = None
#env_functions,function_len = workspace.get_knowledge_base_ai_functions()
task_result:ComputeTaskResult = await self._do_llm_complection(prompt,env_functions,None,True)
if task_result.result_code != ComputeTaskResultCode.OK:
result_obj = {}
result_obj["error_str"] = task_result.error_str
return result_obj
result_obj = json.loads(task_result.result_str)
temp_meta_data = result_obj.get("metadata")
if is_final:
return result_obj
return None
async def do_self_think(self):