zhaoxi
/
Pretor
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
60 Commits 3 Branches 1 Tag 4 MiB
Python 70.7%
TypeScript 28.7%
Dockerfile 0.3%
JavaScript 0.2%
Go to file
朝夕 c1af32e604
Feat/deepseek adapter dropin 14224636701039833263 (#47) 
* feat: refactor DeepSeekReasonerAgent as a drop-in replacement for pydantic_ai Agent

This refactor improves the `DeepSeekReasonerAgent` adapter to be a robust, drop-in replacement for native PydanticAI Agents. It brings several enhancements:

1. Re-implemented `run()` to manually inject historical messages (`message_history`) and dependencies, preserving state across workflows.
2. Replaced the simplistic crash loop with an explicit, manual multi-turn retry mechanism. If the Markdown JSON parser fails, it correctly injects the `ValidationError` back into the conversation history and prompts the model to correct its structure up to `retries` times.
3. Designed an elegant proxy `AgentRunResultProxy` to seamlessly wrap `AgentRunResult` outputs. This cleanly passes through downstream calls (e.g., `result.data`, `result.usage()`, `result.all_messages()`) avoiding `AttributeError`s and Monkey-patching.
4. Integrated fallback tool descriptions parsing, dynamically instructing the model on available tools.
5. Adapted `AgentFactory` to correctly propagate `tools` and `retries`.

Co-authored-by: zhaoxi826 <198742034+zhaoxi826@users.noreply.github.com>

* feat: add deepseek provider option to frontend

- Added 'deepseek' option to `ProvidersSettings.tsx` `<select>`
- Updated frontend Typescript interfaces in `index.ts` to allow 'deepseek' as `provider_type`
- Validated frontend build successful

Co-authored-by: zhaoxi826 <198742034+zhaoxi826@users.noreply.github.com>

---------

Co-authored-by: google-labs-jules[bot] <161369871+google-labs-jules[bot]@users.noreply.github.com>
Co-authored-by: zhaoxi826 <198742034+zhaoxi826@users.noreply.github.com>
 		2026-04-28 10:06:33 +08:00
changelogs fix: 修复部分错误 2026-04-28 04:23:57 +08:00
config wip: 优化了control_node consciousness_node和supervisory_node 2026-04-20 16:20:38 +08:00
docs fix: 修复部分错误 2026-04-28 04:23:57 +08:00
frontend Feat/deepseek adapter dropin 14224636701039833263 (#47) 2026-04-28 10:06:33 +08:00
pretor feat: refactor DeepSeekReasonerAgent as a drop-in replacement for pydantic_ai Agent (#46) 2026-04-28 09:58:54 +08:00
tests Refactor Agent Tool Assignments to Database-Driven Static Loading (#45) 2026-04-28 03:53:33 +08:00
.dockerignore Add react frontend 15346720174401384863 (#7) 2026-04-19 18:59:56 +08:00
.env Merge branch 'dev' of https://github.com/zhaoxi826/Pretor into dev 2026-04-24 23:50:27 +08:00
.gitignore wip:更新版权 2026-04-10 10:50:25 +08:00
.python-version feat:项目初始化，实现了workflow_manager 2026-03-22 18:01:05 +08:00
CODE_OF_CONDUCT.md fix: 修复部分错误 2026-04-28 04:23:57 +08:00
Dockerfile Add react frontend 15346720174401384863 (#7) 2026-04-19 18:59:56 +08:00
LICENSE wip: 优化 2026-04-11 18:14:07 +08:00
Makefile wip: 增加了skill_manager 2026-04-17 12:36:36 +08:00
README.md fix: 修复部分错误 2026-04-28 04:23:57 +08:00
docker-compose.yml wip: 同步 2026-04-25 00:01:59 +08:00
main.py fix: 修复了严重bug 2026-04-27 14:13:34 +08:00
pyproject.toml ✨ Enhance skill management, add tool integrations, and overhaul Chat UI (#44) 2026-04-27 19:20:16 +08:00
uv.lock ✨ Enhance skill management, add tool integrations, and overhaul Chat UI (#44) 2026-04-27 19:20:16 +08:00

README.md

Pretor (执政官)

一款基于 Python 的分布式多 Agent 协作系统

Python 3.13+ Ray Pydantic-AI License

项目架构 | 更新日志 | 未来展望

Pretor 是一款基于 Ray 构建的下一代分布式多 Agent 协作系统。项目采用“中心监管 + 边缘执行”的异构集群模式，通过大参数 MoE 模型进行高层逻辑推理，并协同微调后的轻量化模型高效完成具体任务。借助 Pydantic-AI 提供的强类型约束与 FastAPI 异步网关，Pretor 实现了任务从需求拆解、资源调度到自动化执行的全链路闭环，为个人提供可靠的人工智能助手服务。

✨ 核心特性

🧠 异构协作体系

  • 多智能体集群：内置主管 (Supervisory)、意识 (Consciousness)、控制 (Control) 三大核心节点，实现比单 Agent 系统更严谨的决策链。
  • Worker 动态派生：根据任务需求动态拉起 Ordinary 或 Skill 类型的 Worker Individual，实现资源的按需分配。

🚀 分布式性能保障

  • Ray 驱动：底层基于 Ray 构建，支持跨进程、跨机器的 Actor 通讯，轻松应对高并发任务流。
  • 本地化优先：深度适配 vLLM，支持本地私有化模型部署，在保障隐私的同时大幅降低 API 调用成本。

🛠️ 工业级工程设计

  • 强类型契约：基于 Pydantic-AI 实现 Tool 与 Agent 的接口定义，确保 AI 输出的确定性与安全性。
  • 自动化流：内置工作流引擎 (Workflow Engine)，实现从需求发现到自动化执行的闭环。

📦 Pretor 生态子项目 (Sub-projects)

项目名称 代号 功能定位 当前状态
pretor-viceroy 总督 资源管理：负责系统 Skill 的动态安装、元数据解析与全集群分发。 ✅ 已发布
pretor-stardomain 星域 安全沙箱：为 Agent 自动生成的代码提供轻量化的隔离运行环境，防止逃逸。 📅 规划中
pretor-explorer 探索者 网页感知：自动化爬虫引擎，赋予智能体实时互联网信息搜索与内容抓取能力。 📅 规划中
pretor-pioneer 先驱者 知识增强：RAG 检索增强引擎，管理私有知识库的向量化、索引与精准检索。 📅 规划中

快速开始

本项目正在开发中...