Skip to content

Commit 5e3c95d

Browse files
committed
update readme
1 parent d437666 commit 5e3c95d

5 files changed

Lines changed: 15 additions & 253 deletions

File tree

README.md

Lines changed: 7 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1,23 +1,19 @@
1-
# Autogenesis
1+
# Deep Research Agent
22

33
English | [中文说明](README_zh.md)
44

5-
Autogenesis is a self-evolution protocol and runtime for LLM-based agent systems.
5+
Deep Research Agent is a self-evolution protocol and runtime for LLM-based agent systems.
66

7-
Recent agent protocols often under-specify cross-entity **lifecycle/context management**, **version tracking**, and **safe evolution update interfaces**, which encourages monolithic compositions and brittle glue code. Autogenesis addresses this by decoupling **what evolves** from **how evolution occurs**:
7+
Recent agent protocols often under-specify cross-entity **lifecycle/context management**, **version tracking**, and **safe evolution update interfaces**, which encourages monolithic compositions and brittle glue code. Deep Research Agent addresses this by decoupling **what evolves** from **how evolution occurs**:
88

99
- **RSPL (Resource Substrate Protocol Layer)**: models *prompts, agents, tools, environments, and memory* as protocol-registered resources with explicit **state**, **lifecycle**, and **versioned** interfaces.
1010
- **SEPL (Self Evolution Protocol Layer)**: specifies a closed-loop operator interface to **propose**, **assess**, and **commit** improvements with auditable lineage and **rollback**.
1111

12-
Built on Autogenesis, the system includes an **Autogenesis-Agent** style tool-calling agent that can dynamically instantiate/retrieve/refine resources and improve during execution.
13-
14-
## Architecture
15-
16-
![Autogenesis architecture](docs/architecture.png)
12+
Built on Deep Research Agent, the system includes a **Deep Research Agent** style tool-calling agent that can dynamically instantiate/retrieve/refine resources and improve during execution.
1713

1814
## Self-evolution at a glance
1915

20-
At a high level, Autogenesis supports an iterative loop:
16+
At a high level, Deep Research Agent supports an iterative loop:
2117

2218
- **Act**: an agent produces actions/outputs using an LLM and the available tools/environments.
2319
- **Observe**: capture outcomes, traces, intermediate reasoning, and environment feedback.
@@ -43,7 +39,7 @@ At a high level, Autogenesis supports an iterative loop:
4339
## Repository layout
4440

4541
```
46-
Autogenesis/
42+
DeepResearchAgent/
4743
configs/ # config composition (agents/tools/envs/memory/models)
4844
src/
4945
agent/ # agents
@@ -59,10 +55,6 @@ Autogenesis/
5955
workdir/ # runtime artifacts (logs, traces, results, etc.)
6056
```
6157

62-
## Empirical studies
63-
64-
See empirical results and benchmark protocols in `docs/empirical_studies.md`.
65-
6658
## Optional: run a Tool-Calling Agent
6759

6860
Prerequisites:
@@ -81,4 +73,4 @@ Override model/workdir:
8173
python examples/run_tool_calling_agent.py \
8274
--config configs/tool_calling_agent.py \
8375
--cfg-options model_name=openrouter/gpt-4o workdir=workdir/demo tag=demo
84-
```
76+
```

README_zh.md

Lines changed: 8 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -1,32 +1,28 @@
1-
# Autogenesis(中文)
1+
# Deep Research Agent(中文)
22

33
[English](README.md) | 中文
44

5-
Autogenesis 是一个面向 LLM 智能体系统的**自进化协议**与工程化运行时。
5+
Deep Research Agent 是一个面向 LLM 智能体系统的**自进化协议**与工程化运行时。
66

7-
现有不少 agent 协议(例如 A2A、MCP)对跨实体的 **生命周期/上下文管理****版本追踪**、以及**演化过程中的安全更新接口**规定不足,容易导致系统走向大一统的单体组合与脆弱的胶水代码。Autogenesis 的核心思路是将 **演化什么(what evolves)****如何演化(how evolution occurs)** 解耦:
7+
现有不少 agent 协议(例如 A2A、MCP)对跨实体的 **生命周期/上下文管理****版本追踪**、以及**演化过程中的安全更新接口**规定不足,容易导致系统走向"大一统"的单体组合与脆弱的胶水代码。Deep Research Agent 的核心思路是将 **"演化什么(what evolves)"****"如何演化(how evolution occurs)"** 解耦:
88

99
- **RSPL(Resource Substrate Protocol Layer)**:把 *prompt、agent、tool、environment、memory* 建模为协议注册资源,提供显式的**状态****生命周期****版本化接口**
1010
- **SEPL(Self Evolution Protocol Layer)**:定义闭环的 operator 接口,用于提出(propose)、评估(assess)、提交(commit)改进,并具备可审计的 lineage 与**回滚(rollback)**能力。
1111

12-
基于 Autogenesis,系统也提供了 **Autogenesis-Agent** 风格的思考-行动(tool-calling)智能体,能够在执行过程中动态实例化/检索/精炼资源并持续改进。
13-
14-
## 架构图
15-
16-
![Autogenesis 架构](docs/architecture.png)
12+
基于 Deep Research Agent,系统也提供了 **Deep Research Agent** 风格的思考-行动(tool-calling)智能体,能够在执行过程中动态实例化/检索/精炼资源并持续改进。
1713

1814
## 自进化闭环概览
1915

20-
从机制上看,Autogenesis 支持反复迭代的闭环:
16+
从机制上看,Deep Research Agent 支持反复迭代的闭环:
2117

2218
- **行动(Act)**:智能体基于 LLM + 可用工具/环境产生行动与输出。
2319
- **观测(Observe)**:记录结果、轨迹、关键中间信息,以及来自环境的反馈信号。
24-
- **优化(Optimize)**:用优化器把反馈转成可复用的改进(提示词/解法/变量等的更新),例如 Reflection、GRPO、Reinforce++。
20+
- **优化(Optimize)**:用优化器把反馈转成"可复用的改进"(提示词/解法/变量等的更新),例如 Reflection、GRPO、Reinforce++。
2521
- **记忆(Remember)**:把会话事件、总结与洞察写入记忆系统,供后续步骤/会话复用。
2622

2723
## 核心模块
2824

29-
- **Agents(`src/agent/`**:决定下一步做什么的运行逻辑(规划、工具调用、领域 agent 等)。
25+
- **Agents(`src/agent/`**:决定"下一步做什么"的运行逻辑(规划、工具调用、领域 agent 等)。
3026
- **Tools(`src/tool/`**:暴露给智能体调用的能力(工作流工具 + 默认工具)。
3127
- **Environments(`src/environment/`**:可交互的有状态接口(文件系统、回测环境、浏览器/移动端环境等)。
3228
- **Memory(`src/memory/`**:会话/事件记忆系统,用于总结、洞察与长期状态。
@@ -43,7 +39,7 @@ Autogenesis 是一个面向 LLM 智能体系统的**自进化协议**与工程
4339
## 目录结构
4440

4541
```
46-
Autogenesis/
42+
DeepResearchAgent/
4743
configs/ # 组件组合配置(agent/tool/env/memory/model)
4844
src/
4945
agent/ # agents
@@ -59,10 +55,6 @@ Autogenesis/
5955
workdir/ # 运行产物(日志/轨迹/结果等)
6056
```
6157

62-
## 实验结果
63-
64-
实验设置与结果汇总见 `docs/empirical_studies.md`(英文)。
65-
6658
## (可选)运行 Tool-Calling Agent
6759

6860
前置条件:
@@ -82,4 +74,3 @@ python examples/run_tool_calling_agent.py \
8274
--config configs/tool_calling_agent.py \
8375
--cfg-options model_name=openrouter/gpt-4o workdir=workdir/demo tag=demo
8476
```
85-

docs/architecture.png

-503 KB
Binary file not shown.
-388 KB
Binary file not shown.

docs/empirical_studies.md

Lines changed: 0 additions & 221 deletions
This file was deleted.

0 commit comments

Comments
 (0)