AgentLedger

English · 简体中文

An append-only reputation ledger for AI agents.

---

English

AgentLedger records reputation events and computes transparent trust scores for AI agents. It answers: how has this agent performed over time?

AgentLedger is open, append-only infrastructure. Any platform may submit signed events; any verifier may recompute scores from the published formula.

Use it with AgentID and AgentCert for a complete trust stack.

Features

• Append-only — events are never mutated in place
• Transparent scoring — published formula, auditable output
• Persistent — SQLite storage in v0.1
• Privacy-aware — stores comment hashes instead of raw text by default

Quick Start

pip install -r requirements.txt

python cli.py add-event \
  --agent-id did:agent:demo123456 \
  --type task_completed \
  --rating 5 \
  --comment "Booked restaurant successfully"

python cli.py get-score --agent-id did:agent:demo123456
python cli.py list-events --agent-id did:agent:demo123456 --limit 10

Event Types

Type	Meaning
task_completed	Successful execution
task_failed	Failed execution
user_rating	User-provided rating
complaint	Negative report
complaint_resolved	Complaint closed
verification_passed	Passed verification

Ecosystem

Project	Role
AgentID	Decentralized agent identity
AgentCert	Verifiable trust credentials

License

Apache-2.0 — see LICENSE.

---

简体中文

AgentLedger 记录 Agent 信誉事件并计算透明的信任评分，回答：这个 Agent 长期表现如何？

AgentLedger 是开放的只追加基础设施。任何平台都可提交签名事件；任何验证方都可依据公开公式重算评分。

与 AgentID、AgentCert 组合，可形成完整信任栈。

特性

• 只追加 — 事件写入后不在原处修改
• 评分透明 — 公开公式，结果可审计
• 持久化 — v0.1 使用 SQLite 存储
• 隐私友好 — 默认仅存评论哈希，不存原文

快速开始

pip install -r requirements.txt

python cli.py add-event \
  --agent-id did:agent:demo123456 \
  --type task_completed \
  --rating 5 \
  --comment "餐厅预订成功"

python cli.py get-score --agent-id did:agent:demo123456
python cli.py list-events --agent-id did:agent:demo123456 --limit 10

事件类型

类型	含义
task_completed	任务成功
task_failed	任务失败
user_rating	用户评分
complaint	投诉
complaint_resolved	投诉已解决
verification_passed	通过核验

生态

项目	角色
AgentID	去中心化 Agent 身份
AgentCert	可验证信任凭证

许可证

Apache-2.0 — 见 LICENSE。