Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

[Bug] Magpie 单个模型缺少 modalities.input 导致所有 Provider 验证失败

Chiusa
  • Qualcosa non va?
#2,007 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Una pull request collegata è già stata integrata.

  • #2110 di @wibus-wee — integrata

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
56/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
typescript
Ambito
api, backend, testing

Direzione di ricerca

Start with apps/cli/src/agent/magpie-runtime.ts, especially modelSchema, readMagpieModels(), and the prepareMagpieRuntime() path. Reproduce the reported schema failure with the minimal catalog in the issue, then inspect downstream modalities?.input.includes('image') uses before changing how missing input is handled. Done means mixed catalogs with missing modalities, output-only modalities, and explicit text or image input are handled safely, with regression coverage preserving model ID and size validation.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Affected area

Agent runtime / ACP

Installation method

Desktop release

Lody version or commit

0.108.1

Operating system

macOS 27.0.1 arm64

Agent or runtime

Magpie 0.1.1725；通过 Magpie 导入的内置 Claude、Codex、Pi、Kimi、Grok Provider。

What happened?

导入 Magpie 后，设置 → Agents 中的 Claude-magpie、Codex-magpie、Pi-magpie、Kimi Code-magpie、Grok-magpie 均显示「Provider 验证失败，请重试。」；重试仍失败。

只读检查本地网关：

  • GET /api/hello：HTTP 200，name: "magpie"。
  • GET /v1/models：HTTP 200，返回 91 个模型，没有重复 ID。
  • 第 81 个模型 commandcode-plan/zai-org/GLM-5.3 的 modalities 只有 output: ["text"]，没有 input。
  • Lody 对整份目录执行 schema 校验，因此单个模型缺少能力元数据会阻断全部 Magpie Provider 的准备流程。
What did you expect?

缺少可选能力描述时，Lody 应能采用保守的既有回退行为，或明确处理受影响的模型，避免让无关的有效模型和所有 Magpie Provider 一起无法使用。

至少应在诊断中保留具体模型和字段，而非把非空目录统一提示为 invalid or empty。

How can we reproduce it?
  1. 启动 Magpie，并选择 Command Code Plan 下的 zai-org/GLM-5.3。在上述版本的现用环境中，该模型出现在 /v1/models，但只返回 modalities.output。
  2. 通过 Magpie 导入 Lody 的多个内置 Provider。
  3. 打开 Lody → 设置 → Agents，观察所有导入的 Provider 验证失败；点击「重试」仍失败。
  4. 检查日志中的 Magpie returned an invalid or empty model catalog。

模型对象中触发问题的字段摘录：

{
  "id": "commandcode-plan/zai-org/GLM-5.3",
  "display_name": "GLM-5.3",
  "modalities": {
    "output": ["text"]
  }
}

维护者也可为 readMagpieModels() 注入 fetcher：让 /api/hello 返回 {"name":"magpie"}，让 /v1/models 返回下面的最小目录，即可触发同一 schema 条件，无需真实账号或推理请求：

{
  "object": "list",
  "data": [
    {
      "id": "example-text-model",
      "modalities": { "input": ["text"] }
    },
    {
      "id": "example-output-only-model",
      "modalities": { "output": ["text"] }
    }
  ]
}
How often does it happen?

Every time

相同目录下，现有日志中的多次准备/重试均失败；对捕获目录的隔离 schema 校验稳定复现。未跨机器复测。

Relevant log output

以下摘录移除了工作区标识和其他无关信息：

[acp-capabilities] Refresh failed (cliType=builtin agentType=claude): Magpie returned an invalid or empty model catalog
[acp-capabilities] Refresh failed (cliType=builtin agentType=codex): Magpie returned an invalid or empty model catalog
[acp-capabilities] Refresh failed (cliType=builtin agentType=pi): Magpie returned an invalid or empty model catalog
[acp-capabilities] Refresh failed (cliType=builtin agentType=kimi): Magpie returned an invalid or empty model catalog
[acp-capabilities] Refresh failed (cliType=builtin agentType=grok): Magpie returned an invalid or empty model catalog

隔离校验给出的具体错误：

{
  "expected": "array",
  "code": "invalid_type",
  "path": ["data", 80, "modalities", "input"],
  "message": "Invalid input: expected array, received undefined"
}
Additional context

源码定位

检查 0.108.1 安装包的 resources/cli/index.js，并对照检查时的公开 main（固定在 66c9d019046951b3222ef75ee9ca321d25b011c5），两者具有相同逻辑：

上述 main commit 用于定位公开源码，不作为 0.108.1 的精确构建 commit。

已执行的对照验证

从安装包提取该 schema 定义，配合安装包内的 Zod，对实际捕获的 91 项响应进行隔离校验：

输入 结果
原始响应，91 项 失败，唯一错误为 data[80].modalities.input 缺失
仅为该项补上 input: ["text"]，仍为 91 项 通过
仅移除该项，剩余 90 项 通过

这只证明模型目录校验的触发条件；没有修改实际网关、安装包或用户配置，也没有构建 main、部署修复或验证真实对话调用。

兼容性背景与修复方向

OpenAI 的模型列表接口定义没有要求 modalities.input；这是扩展元数据。这里只报告 Lody 对缺省扩展字段的处理及失败影响范围，不据此断言双方是否另有私有协议约定。

建议兼容缺少 input 的情况，并沿用明确、保守的能力回退；保留模型 ID、大小上限等边界校验。若改为允许 input 缺省，还需一起处理下游多处 m.modalities?.input.includes('image')，否则会将验证失败变成运行时异常。回归用例应覆盖「整个 modalities 缺省」「只有 output」「显式 text/image input」及多个模型混合的目录。

Before submitting
  • I searched the existing issues and did not find a duplicate.
  • This report concerns an open-source component in this repository, not a hosted service, Web or mobile app, account, or billing issue.
  • This is not a security vulnerability; security reports follow the repository's security policy.
  • I removed credentials, private source, conversations, prompts, personal data, and other sensitive information.
Lingua principale
TypeScript
Stelle
1.2k
Fork
141
Merge medio
12h 35m
PR unite (30g)
664

Preparare l'ambiente

Questo progetto non fornisce container di sviluppo, Dockerfile né guida per i contributori, quindi l'ambiente è a tuo carico: parti dal suo README e consulta la nostra guida al primo contributo per i passaggi generali.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di LodyAI/Lody

Tutte le issue di LodyAI/Lody

Issue simili

Altre issue su TypeScript

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.