Add Node.js low-level tool-definition E2E test [3/6]

[edburns]

Summary

Adds Node.js low-level tool-definition E2E coverage and includes the session-lifecycle replay mismatch fix needed for stable cross-platform replay matching.

This PR is related to issue #1682 but does not fix #1682.

What changed

• Adds low_level_tool_definition test coverage in:
  • nodejs/test/e2e/tools.e2e.test.ts
• Includes the session-lifecycle fix from the source branch:
  • nodejs/test/e2e/session_lifecycle.e2e.test.ts
  • test/snapshots/session_lifecycle/should_list_created_sessions_after_sending_a_message.yaml
• Reuses the shared snapshot from PR Add Java low-level tool definition E2E test and skill [1/6] #1721:
  • test/snapshots/tools/low_level_tool_definition.yaml
• Keeps low-level test tool definitions aligned with what the shared snapshot actually exercises (no unexercised grep override in this scenario).

Dependency / sequencing

• Depends on PR Add Java low-level tool definition E2E test and skill [1/6] #1721 being merged first (already merged), because it introduces the shared low-level snapshot.
• Part of the Start iterating on #1682 On branch edburns/1682-java-tool-ergonomics #1692 breakout plan (PR 3 of 6).

Related

• Issue: Java: Ergonomics: Defining tools #1682 (not fixed by this PR)
• PR Add Java low-level tool definition E2E test and skill [1/6] #1721
• Original combined PR: Start iterating on #1682 On branch edburns/1682-java-tool-ergonomics #1692

[Copilot]

Pull request overview

Adds Node.js E2E coverage for the shared “low-level tool definition” snapshot scenario, and updates the session-lifecycle replay snapshot + Node test prompt to address a replay-matching mismatch.

Changes:

• Add Node.js E2E test low_level_tool_definition that exercises explicit tool definitions + available-tools filtering using the shared replay snapshot.
• Update the session-lifecycle E2E prompt from “Say world” → “Say hi” and adjust the stored replay snapshot content accordingly.
• Import and use ToolSet in Node E2E tools tests to express source-qualified available-tools patterns.

Show a summary per file

File	Description
nodejs/test/e2e/tools.e2e.test.ts	Adds the Node E2E test that replays tools/low_level_tool_definition.yaml and validates tool execution + handler side effects.
nodejs/test/e2e/session_lifecycle.e2e.test.ts	Changes the second session activity prompt to “Say hi” to match the updated shared replay snapshot.
test/snapshots/session_lifecycle/should_list_created_sessions_after_sending_a_message.yaml	Updates the recorded replay conversation for the session-lifecycle scenario (prompt + assistant response).

Copilot's findings

• Files reviewed: 3/3 changed files
• Comments generated: 2

[github-actions]

Cross-SDK Consistency Review

Summary

This PR is part of a planned 6-PR series adding low_level_tool_definition E2E coverage across all SDKs. The Node.js implementation is well-structured and the ToolSet API usage (new ToolSet().addCustom("*").addBuiltIn("web_fetch")) is consistent with the Java implementation added in #1721. ✅

One Consistency Observation: Java vs Node.js Test Scope

The Java test (LowLevelToolDefinitionIT.java) registers three tools in the session config — including a grep override via ToolDefinition.createOverride — while this Node.js test registers only two tools. Both tests replay the same snapshot (test/snapshots/tools/low_level_tool_definition.yaml), which only exercises the first two tools.

The PR description explicitly notes this choice: "no unexercised grep override in this scenario." That reasoning is sound for Node.js — defining tools that the LLM never calls doesn't add replay value here. However, it creates a minor divergence between the two test implementations:

SDK	Tools defined	Covers override-tool coexistence?
Java	3 (incl. grep override)	✅ incidentally
Node.js (this PR)	2	❌

Suggestion for the remaining PRs in this series (Python, Go, .NET, Rust): consider aligning on one approach — either consistently include the grep override tool to test that override tools can coexist in a session alongside regular tools, or consistently omit it across all SDKs. The Java test covers override-tool coexistence incidentally; the Node.js test relies on its separate overrides built-in tool with custom tool test for that coverage. Either way, it would be good for the series to converge on a single pattern so the low_level_tool_definition test means the same thing in every SDK.

Everything Else Looks Good

• ToolSet().addCustom("*").addBuiltIn("web_fetch") pattern is parallel across Java and Node.js ✅
• Raw invocation.arguments access in search_items handler mirrors invocation.getArguments() in Java ✅
• Reusing the shared snapshot from Add Java low-level tool definition E2E test and skill [1/6] #1721 is the correct approach ✅

Generated by SDK Consistency Review Agent for issue #1725 · sonnet46 1.7M · ◷