Add a deterministic Task Processing test provider for LibreSign AI flows

未关闭
#9,055 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
35/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
php
领域
backend, testing

调研方向

Start by reading the AI foundation capability contract referenced in the issue and identify the task types currently consumed by LibreSign; no specific files or tests are named here. Then inspect how the app integrates with Nextcloud Task Processing and how existing tests provide service doubles. Done means a production-separated, network-free deterministic provider with selectable fixtures and focused tests for the supported success and failure cases.

由索引模型根据 Issue 内容生成。

描述

backend php

Goal

Provide a fast, deterministic Task Processing provider/test double that lets LibreSign exercise AI-assisted flows in CI without paid APIs, external network access or probabilistic output.

This issue is part of the AI foundation epic and must follow the capability contract defined by the foundation work.

Why

AI-assisted features need normal regression coverage. Unit mocks alone are not enough to prove that LibreSign integrates correctly with the Nextcloud Task Processing path, while real LLM calls are too slow, costly and nondeterministic for ordinary pull-request CI.

The test provider should make AI behavior as testable as other LibreSign application services.

Scope

Implement the minimum provider/task types needed by the currently defined LibreSign AI capability contract.

The provider should support deterministic scenarios such as:

  • successful task with known output;
  • malformed/invalid output;
  • provider/task unavailable;
  • task failure;
  • cancellation where the supported contract requires it;
  • delayed/asynchronous completion where relevant;
  • intermediate/progress behavior only when an implemented LibreSign feature depends on it.

Do not implement fake capabilities that have no current LibreSign consumer.

Test-fixture design

Outputs must be fixture-driven and stable.

Tests should be able to request a named behavior instead of depending on prompt wording or model-like heuristics.

Examples:

  • summary fixture returns a known summary payload;
  • structured suggestion fixture returns a known valid payload;
  • hostile fixture returns a known invalid/unsafe payload;
  • failure fixture returns a controlled Task Processing failure.

Keep the test provider clearly separated from production providers and impossible to activate accidentally in a normal production installation.

Security

The deterministic provider exists to test the same trust boundaries as a real provider.

It must make it possible to verify that LibreSign:

  • validates output rather than trusting it;
  • does not use prompts as authorization;
  • handles provider failure safely;
  • rejects hostile generated markup/structured output at the correct LibreSign boundary;
  • does not expose one user's task result to another actor.

CI expectations

The deterministic provider should be suitable for normal pull-request CI and focused local development.

Keep it:

  • network-free;
  • free of paid APIs;
  • fast enough for normal CI;
  • deterministic across runs;
  • compatible with the supported Nextcloud test environment.

Done when

  • A deterministic Task Processing provider/test double can expose the required foundation task types.
  • Tests can select stable success and failure fixtures.
  • It cannot be confused with or enabled as a production provider.
  • It supports backend/integration and browser tests without external services.
  • Security/error cases can be reproduced deterministically.
  • Focused automated tests cover the provider itself.
主要语言
PHP
星标
844
派生
176
平均合并
14 小时 3 分钟
30 天内合并 PR
344

环境准备

在 Codespaces 中打开

在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

LibreSign/libresign 的其他 Issue

查看 LibreSign/libresign 的全部 Issue

相似的 Issue

更多 PHP Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。