⚙️
Agent工作流
优化引擎
基于 Elixir · 无代码搭建完整 AI 自动化流程
英文名：OptimalEngine
7.5AI 综合评分
elixiraiworkflow
✦ AI Skill Hub 推荐
经 AI Skill Hub 精选评估，优化引擎 获评「推荐使用」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色，AI 评分 7.5 分，适合有一定技术背景的用户使用。
📚 深度解析
优化引擎 是一套完整的 AI Agent 自动化工作流方案。随着 AI 能力的不断提升，基于 Agent 的自动化工作流正在成为提升个人和团队效率的核心方式。区别于传统的 RPA 自动化（模拟鼠标键盘操作），AI Agent 工作流通过理解任务意图、动态规划执行路径，能够处理更复杂的非结构化任务。
优化引擎 工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。
在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 7.5 分，是同类 Agent 工作流中的精选推荐。
优化引擎 工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。
在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 7.5 分，是同类 Agent 工作流中的精选推荐。
📋 工具概览
优化引擎 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
GitHub Stars
⭐ 6
开发语言
Elixir
支持平台
Windows / macOS / Linux
维护状态
轻量级项目，按需更新
开源协议
未公布
AI 综合评分
7.5 分
工具类型
Agent工作流
Forks
—
📖 中文文档
以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。
优化引擎 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
📌 核心特色
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
🎯 主要使用场景
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。
安装命令
# 克隆仓库 git clone https://github.com/Miosa-osa/OptimalEngine cd OptimalEngine # 查看安装说明 cat README.md # 按 README 完成环境依赖安装后即可使用
📋 安装步骤说明
- 访问 GitHub 仓库获取工作流文件
- 在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
- 上传工作流文件
- 按照提示配置必要的环境变量和 API Key
- 运行测试确认流程正常后投入使用
以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。
常用命令 / 代码示例
# 查看帮助 optimalengine --help # 基本运行 optimalengine [options] <input> # 详细使用说明请查阅文档 # https://github.com/Miosa-osa/OptimalEngine
以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。
配置示例
# optimalengine 配置说明 # 查看配置选项 optimalengine --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量（覆盖配置文件） export OPTIMALENGINE_CONFIG="/path/to/config.yml"
以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。
Optimal Engine
Optimal Engine is a self-hosted operating engine for human and AI workspaces.
It lets a person or organization define a workspace, organize that workspace into typed Nodes, preserve source evidence, classify incoming Signals, promote reviewed knowledge into Facts and Memory Objects, assemble governed context for humans and agents, and project the same state into markdown, APIs, dashboards, CLI tools, and agent runtimes.
The short version:
Human defines the world:
Workspace -> Nodes -> Relationships -> Policies
Data enters the world:
Source Package -> Signal -> Route -> Claim -> Fact -> Memory Object
Humans and agents use the world:
Query or task -> Context Package -> Active Memory Pool -> Action
Work improves the world:
Observation -> Pending Claim -> Reviewed Fact -> Memory -> Workflow -> SkillSystem flow:
flowchart LR
Human[Human / Agent / App / Connector] --> Topology[Workspace + Node Scope]
Topology --> Source[Source Package]
Source --> Signal[Signal Pipeline]
Signal --> Claim[Claim Candidate]
Claim --> Review[Review Queue]
Review --> Policy[Stale / Conflict / Supersession Policy]
Policy --> Fact[Accepted Fact]
Policy --> Reject[Rejected or Conflict Response]
Policy --> Supersede[Supersede Prior Fact]
Supersede --> Fact
Fact --> Memory[Memory Object]
Memory --> Retrieval[Retrieval Coordinator]
Retrieval --> Context[Context Package]
Context --> Pool[Active Memory Pool]
Pool --> Action[Human / Agent Action]
Action --> Observation[Observation]
Observation --> Claim
Layer ownership:
flowchart TB
Surfaces[Markdown / CLI / App / Agent / Connector]
Gateway[Command Gateway]
Topology["Workspace Topology
workspaces, Nodes, relationships"] Memory["Memory Core
sources, Claims, Facts, memories, ledger"] Retrieval["Retrieval + Context
packages, filters, audit"] Active["Active Work
pools, observations, pending Claims"] Workflow["Workflow + Skill
traces, procedures, Skill Packages"] Governance["Tool / Model Governance
registrations, calls, audit"] Evaluation["Evaluation
runs, cases, scores"] Store["Shared SQLite/Postgres store
separate table ownership"] Exports[Markdown / API / UI projections] Surfaces --> Gateway Gateway --> Topology Gateway --> Memory Gateway --> Retrieval Gateway --> Active Gateway --> Workflow Gateway --> Governance Gateway --> Evaluation Topology --> Store Memory --> Store Retrieval --> Store Active --> Store Workflow --> Store Governance --> Store Evaluation --> Store Store --> Exports
workspaces, Nodes, relationships"] Memory["Memory Core
sources, Claims, Facts, memories, ledger"] Retrieval["Retrieval + Context
packages, filters, audit"] Active["Active Work
pools, observations, pending Claims"] Workflow["Workflow + Skill
traces, procedures, Skill Packages"] Governance["Tool / Model Governance
registrations, calls, audit"] Evaluation["Evaluation
runs, cases, scores"] Store["Shared SQLite/Postgres store
separate table ownership"] Exports[Markdown / API / UI projections] Surfaces --> Gateway Gateway --> Topology Gateway --> Memory Gateway --> Retrieval Gateway --> Active Gateway --> Workflow Gateway --> Governance Gateway --> Evaluation Topology --> Store Memory --> Store Retrieval --> Store Active --> Store Workflow --> Store Governance --> Store Evaluation --> Store Store --> Exports
Current Build
The current build moves the engine from a file/search-first system toward a governed workspace runtime.
Built and verified now:
| Area | Status |
|---|---|
| Workspace topology | Workspaces, Projects-as-Nodes, typed Nodes, Node Types, Node relationships, membership, and projection records. |
| Source evidence | Raw text becomes Source Packages with hash, workspace scope, trust label, security labels, partitions, and metadata. |
| Governed assets | Raw multimodal files can be preserved as Source Packages, workspace-scoped asset rows, and derivation ledger entries. |
| Pipeline asset governance | Pipeline.run/2 preserves parser-produced assets through Memory Core before enrichment, decomposition, and embedding. |
| Indexer asset governance | Binary indexing can pass workspace scope into the governed pipeline so indexed assets do not fall back to the default workspace. |
| API asset uploads | POST /api/assets preserves JSON-uploaded or local-path files through Memory Core and can optionally run a governed multimodal adapter. |
| Connector asset ingestion | Connectors.preserve_payload_assets/4 preserves connector attachments/files through Memory Core with connector origin metadata and per-attachment errors. |
| Connector sync asset preservation | Connector sync may return raw payloads with attachments/files; the runner preserves them through Memory Core before completing the run. |
| Multimodal tool registry | Open-source adapter targets are cataloged for document intelligence, OCR, audio, video, visual reasoning, visual document retrieval, and cross-modal embeddings, with adapter profiles for primary role, output formats, claimability, install profile, and default runtime arguments. |
| Multimodal adapter runs | Adapter attempts and outputs can be recorded as governed derived artifacts linked to assets, Source Packages, scopes, hashes, and derivation ledger entries. |
| Multimodal adapter runner | Configured local adapter commands can execute against governed assets, with completed, failed, and unavailable runs recorded through Memory Core. Runtime defaults now come from adapter profiles instead of hardcoded runner branches. |
| Structured multimodal extraction parsing | Nested transcript segments, document pages/elements/tables, and video frame observations/detections can be normalized into typed extraction projection rows. |
| Adapter-output Claims | Completed adapter outputs can be preserved as derived Source Packages and converted into pending Claims. Failed or unavailable adapter runs cannot become Claims. |
| Asset extraction projections | Completed adapter runs can be normalized into asset_extractions plus typed transcript, OCR span, visual observation, and embedding-ref projection tables. The adapter runner now auto-projects supported completed runs, and text-bearing extractions can become derived Source Packages and pending Claims. |
| Claim/Fact separation | Extracted text becomes an unreviewed Claim first. A Claim becomes a Fact only through the truth-promotion lifecycle. Stale Claims are blocked by default, conflicting current Facts require explicit supersession, and supersession closes the old Fact, linked Memory Objects, and affected Context Packages with a typed edge and ledger entry. |
| Claim review queue | MemoryCore.claim_review_queue/1 and GET /api/memory-core/claim-review return review/lifecycle counts plus filterable Claim rows for UI and agent review workflows. |
| Memory Objects | Accepted Facts can be wrapped into source-backed Memory Objects with evidence links and confidence/precision metadata. |
| Derivation Ledger | Source-to-Claim, Claim-to-Fact, and Fact-to-Memory steps write lineage entries. |
| Governed retrieval | Retrieval returns Context Packages, not loose chunks, writes explicit retrieval-plan metadata, applies structured subject/action/object, asset, workflow, and skill filters, filters Facts, Memory Objects, asset extraction projections, workflow candidates, and approved/enabled Skill Packages by partition/security scope before package assembly, marks affected packages stale when returned Facts or Memory Objects are superseded, can refresh a stale package from its original request scope, can batch refresh stale packages, and exposes API, CLI/cron, and supervised scheduler triggers for app/agent/runtime workflows. |
| Active Memory Pools | Task-scoped working memory can load Context Packages, refresh stale loaded Context Packages, publish observations as pending Claims, and expose the open/retrieve/refresh/observe/close loop through API routes. |
| Smart memory intake | Memory.remember/2 and POST /api/memory/remember can gate low-salience writes, skip semantic duplicates, update superseded memories, and attach intake metadata before the governed Source Package and pending Claim bridge runs. |
| Tool/model governance | Registered tools and model operations can enforce privileges, partitions, required inputs, required outputs, audit links, and the first governed execution path. |
| Connector governance | Connector sync runs through the governed tool-call surface by default, blocking unauthorized runs before connector execution and recording both connector-run and tool-call audit rows when allowed. Raw sync requires an explicit governed: false bypass. |
| Evaluation runner | Benchmark/evaluation runs can execute against governed retrieval, load JSON/JSONL datasets, assemble Context Packages per question, produce deterministic local answer surfaces, judge expected-answer matches, persist per-case scores, and update aggregate run scores. mix optimal.eval.run exposes the flow for CLI/cron use, and external answerer/judge callbacks can be plugged in later without changing the storage contract. |
| Workspace export | Markdown/files are projections and editing surfaces, not the only source of truth. |
| Reality check | mix optimal.reality_check covers store counts, topology, evidence/truth lifecycle, recall packages, retrieval, connectors, evaluation records, dataset-runner execution, wiki, and compliance probes. |
Current verification:
mix test test/memory_core/spine_test.exs test/workspace_export_test.exs test/topology/workspace_topology_test.exs test/topology/node_member_test.exs test/topology/node_test.exs test/topology/workspace_surface_spine_test.exs test/signal/dispatcher_test.exs test/connectors/runner_test.exs --seed 0
64 tests, 0 failures
mix test test/memory_core/asset_store_test.exs test/pipeline/pipeline_asset_store_test.exs test/pipeline/indexer_asset_store_test.exs --seed 0
5 tests, 0 failures
mix test test/pipeline/multimodal_tool_registry_test.exs test/memory_core/asset_store_test.exs test/pipeline/pipeline_asset_store_test.exs test/pipeline/indexer_asset_store_test.exs --seed 0
10 tests, 0 failures
mix test test/pipeline/multimodal_adapter_runner_test.exs test/pipeline/multimodal_tool_registry_test.exs test/memory_core/asset_store_test.exs test/pipeline/pipeline_asset_store_test.exs test/pipeline/indexer_asset_store_test.exs --seed 0
21 tests, 0 failures
mix test test/memory_core/spine_test.exs test/pipeline/multimodal_adapter_runner_test.exs test/memory_core/asset_store_test.exs --seed 0
28 tests, 0 failures
mix test test/api/router_test.exs --seed 0
32 tests, 0 failures
mix test test/connectors/asset_ingest_test.exs --seed 0
3 tests, 0 failures
mix test test/connectors/runner_test.exs test/connectors/asset_ingest_test.exs --seed 0
14 tests, 0 failures
mix test test/evaluation_test.exs --seed 0
4 tests, 0 failures
mix test test/memory_core/claim_review_test.exs test/api/router_test.exs --seed 0
34 tests, 0 failures
mix test test/evaluation_test.exs test/connectors/runner_test.exs test/connectors/asset_ingest_test.exs test/api/router_test.exs test/memory_core/spine_test.exs test/pipeline/multimodal_adapter_runner_test.exs test/memory_core/asset_store_test.exs test/memory_core/claim_review_test.exs --seed 0
87 tests, 0 failures
mix test test/memory_core/context_refresh_scheduler_test.exs --seed 0
3 tests, 0 failures
mix test test/memory_core/spine_test.exs --seed 0
16 tests, 0 failures
mix test test/memory/intake_test.exs --seed 0
4 tests, 0 failures
mix test test/pipeline/multimodal_adapter_runner_test.exs --seed 0
8 tests, 0 failures
mix optimal.reality_check
126 probes, 126 ok, 0 warn, 0 failThe full legacy suite still contains older optional/backend warnings. The focused slice above is the current verified build path.
For the detailed build audit, see docs/reference/build-goal-alignment.md. It maps each intended layer to the code, tests, and remaining gaps.
Human And Agent Usage
Humans can use the engine through:
markdown
CLI
web/app UI
reports
workspace exportsAgents can use the engine through:
API
MCP/tool surface
Context Packages
Active Memory Pools
Skill Packages
workspace filesHumans and agents should not have separate memory systems. They use different interfaces into the same governed workspace runtime.
Task context refresh now follows the same governed recall path:
flowchart LR
Package[Loaded Context Package] --> Invalidate[Fact or Memory change marks it stale]
Invalidate --> Pool[Active Memory Pool refresh]
Pool --> Recall[Retrieval Coordinator replays original request scope]
Recall --> Fresh[Fresh Context Package]
Fresh --> Pool2[Pool links refreshed context]
Quick Start
Requirements:
- Elixir
~> 1.17 - Erlang/OTP 26+
- Node 20+ for app/site surfaces
- a local C toolchain for optional native dependencies
Local engine:
mix deps.get
mix compile
mix optimal.reality_checkFocused verification for the current build:
mix test test/memory_core/spine_test.exs \
test/workspace_export_test.exs \
test/topology/workspace_topology_test.exs \
test/topology/node_member_test.exs \
test/topology/node_test.exs \
test/topology/workspace_surface_spine_test.exs \
test/signal/dispatcher_test.exs \
--seed 0Focused multimodal asset verification:
mix test test/pipeline/multimodal_tool_registry_test.exs \
test/pipeline/multimodal_adapter_runner_test.exs \
test/memory_core/asset_store_test.exs \
test/pipeline/pipeline_asset_store_test.exs \
test/pipeline/indexer_asset_store_test.exs \
--seed 0Useful commands:
mix optimal.reality_check
mix optimal.search "project"
mix optimal.rag "what changed this week?"
🎯 aiskill88 AI 点评
A 级
2026-06-03
优化引擎，企业工作流自动化利器
📚 实用指南（长尾问题）
适合谁
- 需要 OptimalEngine 解决具体问题的开发者与运营人员
最佳实践
- 先在测试环境跑通最小用例，再接入生产数据
常见错误
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
部署方案
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
👥 适合谁
- 需要 OptimalEngine 解决具体问题的开发者与运营人员
⭐ 最佳实践
- 先在测试环境跑通最小用例，再接入生产数据
⚠️ 常见错误
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
👥 适合人群
🎯 使用场景
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
⚖️ 优点与不足
✅ 优点
- +大幅减少重复性人工操作
- +可视化流程，清晰直观
- +可扩展性强，支持复杂场景
⚠️ 不足
- −未明确开源协议，商用场景需谨慎评估
- −初始配置和调试需投入一定时间
- −强依赖外部服务的稳定性
- −复杂场景需具备一定技术基础
⚠️ 使用须知
该工具未明确声明开源协议，商业使用前请联系原作者确认授权范围，避免侵权风险。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
🔗 相关工具推荐
📰 相关 AI 新闻
🍿 AI 圈相关吃瓜
🗺️ 相关解决方案
🧩 你可能还需要
基于当前 Skill 的能力图谱，自动补全的工具组合
❓ 常见问题 FAQ
开源AI工作流平台
💡 AI Skill Hub 点评
AI Skill Hub 点评：优化引擎 的核心功能完整，质量良好。对于自动化工程师和运维人员来说，这是一个值得纳入个人工具库的选择。建议先在非生产环境试用，再逐步推广。
⬇️ 获取与下载
⚠️ 该工具未声明开源协议，不提供直接下载。请访问原项目了解使用条款。
🌐 原始信息
| 原始名称 | OptimalEngine |
| 原始描述 | 开源AI工作流：The second brain of a company. Model-agnostic Elixir/OTP data-storage substrate:。⭐6 · Elixir |
| Topics | elixiraiworkflow |
| GitHub | https://github.com/Miosa-osa/OptimalEngine |
| 语言 | Elixir |
🔗 原始来源
🐙 GitHub 仓库 https://github.com/Miosa-osa/OptimalEngine
收录时间：2026-06-03 · 更新时间：2026-06-03 · License：未公布 · AI Skill Hub 不对第三方内容的准确性作法律背书。