Skip to content

USER_GUIDE.md

Latest commit

 

History

History
228 lines (146 loc) · 4.53 KB

USER_GUIDE.md

File metadata and controls

228 lines (146 loc) · 4.53 KB
Raw

Tilo User Guide / 使用指南

This guide explains how to run and use the current Tilo Console.

本文档说明如何启动和使用当前版本的 Tilo Console。

1. Start Tilo / 启动 Tilo

git clone https://github.com/adam2go/tilo-framework.git
cd tilo-framework
cp .env.example .env

docker compose up --build

Open:

http://localhost:3000

Backend health check:

curl http://localhost:8000/api/health

2. What You See / 当前页面结构

The current Console has three main areas:

当前 Console 主要分为三个区域：

Left   : Agent Task input and demo prompts
Center : Generated Artifact result
Right  : Context panel: Trace / Memory / Skills / Files + Inbox

Left Panel: Agent Task

Use this panel to send a task to the agent.

左侧用于输入任务或点击内置 Demo。

You can:

  • click a demo prompt
  • edit the prompt manually
  • click Send Message

Center Panel: Artifact

This is where Tilo renders the final result.

中间区域是结果交付区，也就是 Tilo 的 Artifact 页面。

Depending on the prompt, it may render:

  • contract review artifact
  • dashboard artifact
  • competitive analysis table
  • document artifact

Right Panel: Context

The right panel helps you understand and control the run.

右侧区域用于理解和控制 Agent 运行过程。

Tabs:

  • Trace: visible execution steps
  • Memory: confirmed memories and memory candidates
  • Skills: skills and skill candidates
  • Files: reserved for future file-backed runs

The Inbox section shows pending human decisions.

3. First Run / 第一次运行

Try this prompt:

Review this contract and flag risky clauses around liability, termination, and payment terms.

Click Send Message.

Expected result:

  1. Tilo creates a Task.
  2. Tilo creates a Run.
  3. Tilo recalls memory.
  4. Tilo selects skills.
  5. Tilo generates an artifact.
  6. Tilo creates confirmations if needed.
  7. Tilo creates memory candidates.
  8. The UI shows artifact, trace, inbox, and memory.

4. Confirm Memory / 确认长期记忆

Open the Memory tab in the right panel.

You may see memory candidates.

Click Confirm to turn a candidate into confirmed long-term memory.

Why this matters:

  • Candidate memory is not trusted by default.
  • Confirmed memory can be recalled in future tasks.
  • This is how Tilo starts to become personalized and project-aware.

5. Approve Confirmations / 审批确认项

The Inbox section shows pending decisions.

Click Approve to approve a confirmation.

Examples:

  • approve a high-risk contract revision suggestion
  • approve a recommended follow-up action
  • approve a tool action in future versions

Tilo's philosophy is: agents execute, humans decide.

6. Try Built-in Demos / 内置 Demo

Contract Review

Review this contract and flag risky clauses around liability, termination, and payment terms.

Expected output:

  • contract review artifact
  • risk items
  • suggested revisions
  • confirmation action

Sales Follow-up

Which customers should sales follow up with this week?

Expected output:

  • dashboard artifact
  • follow-up recommendations
  • confirmation items

Competitive Analysis

Create a competitive analysis for memory-native AI agent frameworks.

Expected output:

  • comparison table
  • structured insight artifact
  • memory candidate

7. Current Limitations / 当前限制

Tilo is still early.

Current limitations:

  • UI is functional but not polished.
  • Demo data is partly mocked.
  • File upload is not fully implemented.
  • Browser/MCP/real external tools are not production-ready yet.
  • Memory recall is still evolving.
  • Artifact pages are early but already schema-driven.

8. Recommended Next Framework Improvements / 推荐下一步框架优化

For framework maturity, the next improvements should be:

  1. Keep app scaffolding and validation deterministic.
  2. Add stronger contract tests for new runtime actions.
  3. Improve artifact editing/versioning without moving action semantics into the frontend.
  4. Expand example fixtures only when they exercise reusable runtime contracts.
  5. Capture real screenshots from the running /demo only.

9. Mental Model / 使用心智模型

Do not use Tilo like a normal chatbot.

不要把 Tilo 当成普通聊天机器人。

Use it like this:

I give a goal.
Tilo creates a task.
The agent runs.
The result becomes an artifact.
I approve key decisions.
Useful context becomes memory.
Next time, Tilo works better.

This is the key product loop.