From 4e56e548872a162129efee1b09987af7893b0e0c Mon Sep 17 00:00:00 2001 From: Adam Malczewski Date: Tue, 25 Aug 2026 23:54:36 +0900 Subject: fix(tabs): never crash on duplicate toolCallIds in tool batches --- src/core/chunks/groups.test.ts | 73 ++++++++++++++++++++++++++++++++++++++++++ src/core/chunks/groups.ts | 8 +++++ 2 files changed, 81 insertions(+) (limited to 'src/core') diff --git a/src/core/chunks/groups.test.ts b/src/core/chunks/groups.test.ts index c8b0fa2..db632a3 100644 --- a/src/core/chunks/groups.test.ts +++ b/src/core/chunks/groups.test.ts @@ -122,4 +122,77 @@ describe("groupRenderedChunks", () => { expect(singles[0]?.kind === "single" && singles[0].chunk.chunk.type).toBe("tool-call"); expect(singles[1]?.kind === "single" && singles[1].chunk.chunk.type).toBe("tool-result"); }); + + // --- Duplicate toolCallId hardening (fix-toolcall-duplicate-keys) --- + // A provider double-emission (e.g. openrouter/stealth/ox-alpha) can persist the + // SAME toolCallId twice within one step. The renderer must NEVER crash on that + // (Svelte `each_key_duplicate` aborts the reactive flush — the tab appears dead). + + it("dedupes duplicate toolCallIds within a batch", () => { + // The PII_REDACTED_CONVERSATION_SHORT_01 pattern: every id emitted twice, in order (seq 1,2 then 3,4). + const groups = groupRenderedChunks([ + call(1, "a", "s1"), + call(2, "b", "s1"), + call(3, "a", "s1"), + call(4, "b", "s1"), + ]); + expect(groups).toHaveLength(1); + const g = groups[0]; + if (g?.kind !== "tool-batch") throw new Error("expected a tool-batch group"); + expect(g.entries).toHaveLength(2); + expect(g.entries.map((e) => e.call.toolCallId)).toEqual(["a", "b"]); + }); + + it("dedupes duplicate toolCallIds and pairs each unique call with its result", () => { + const groups = groupRenderedChunks([ + call(1, "a", "s1"), + call(2, "b", "s1"), + call(3, "a", "s1"), + call(4, "b", "s1"), + result(5, "a", "s1"), + result(6, "b", "s1"), + result(7, "a", "s1"), + result(8, "b", "s1"), + ]); + expect(groups).toHaveLength(1); + const g = groups[0]; + if (g?.kind !== "tool-batch") throw new Error("expected a tool-batch group"); + expect(g.entries).toHaveLength(2); + expect(g.entries[0]?.call.toolCallId).toBe("a"); + expect(g.entries[0]?.result?.content).toBe("result-a"); + expect(g.entries[1]?.call.toolCallId).toBe("b"); + expect(g.entries[1]?.result?.content).toBe("result-b"); + }); + + it("keeps the first occurrence of a duplicate toolCallId and skips the second", () => { + const firstA = call(1, "a", "s1"); // first occurrence of a + const firstB = call(2, "b", "s1"); // first occurrence of b + const groups = groupRenderedChunks([ + firstA, + firstB, + call(3, "a", "s1"), // duplicate of a — skipped + call(4, "b", "s1"), // duplicate of b — skipped + ]); + const g = groups[0]; + if (g?.kind !== "tool-batch") throw new Error("expected a tool-batch group"); + expect(g.entries).toHaveLength(2); + // The FIRST occurrence (by stream order) is the one kept — referential + // equality against the seq-1 / seq-2 chunks (the entry stores the inner + // chunk; the outer `seq` lives on the RenderedChunk wrapper). + expect(g.entries[0]?.call).toBe(firstA.chunk); + expect(g.entries[1]?.call).toBe(firstB.chunk); + }); + + it("keeps the first (provisional) occurrence of a duplicate and marks the batch provisional", () => { + // One provisional + one committed chunk with the SAME id: the first + // (provisional) occurrence wins, and the batch stays provisional. + const prov = call(1, "a", "s1", true); + const groups = groupRenderedChunks([prov, call(2, "a", "s1")]); + const g = groups[0]; + if (g?.kind !== "tool-batch") throw new Error("expected a tool-batch group"); + expect(g.provisional).toBe(true); + expect(g.entries).toHaveLength(1); + expect(g.entries[0]?.call).toBe(prov.chunk); + expect(g.entries[0]?.result).toBeNull(); + }); }); diff --git a/src/core/chunks/groups.ts b/src/core/chunks/groups.ts index 53a2873..c1c80c9 100644 --- a/src/core/chunks/groups.ts +++ b/src/core/chunks/groups.ts @@ -72,9 +72,17 @@ export function groupRenderedChunks(rendered: readonly RenderedChunk[]): readonl emittedSteps.add(stepId); const entries: ToolBatchEntry[] = []; + // Dedupe by toolCallId within the batch: a provider double-emission (or any + // corrupted transcript) can carry the SAME call id on multiple tool-call + // chunks in one step. The first occurrence (stream order) wins; later ones + // are skipped so the keyed `{#each}` in the UI never sees duplicate keys + // (Svelte `each_key_duplicate` aborts the reactive flush — tab appears dead). + const seenCallIds = new Set(); let provisional = false; for (const inner of rendered) { if (inner.chunk.type === "tool-call" && inner.chunk.stepId === stepId) { + if (seenCallIds.has(inner.chunk.toolCallId)) continue; + seenCallIds.add(inner.chunk.toolCallId); const result = resultByCallId.get(inner.chunk.toolCallId) ?? null; entries.push({ call: inner.chunk, result }); if (inner.provisional) provisional = true; -- cgit v1.2.3