A platform for community discussion. Free, open, simple.
import { render } from "@ember/test-helpers";
import { module, test } from "qunit";
import { setupRenderingTest } from "discourse/tests/helpers/component-test";
import AiDecodedTranscript from "discourse/plugins/discourse-ai/discourse/components/ai-decoded-transcript";
module("Integration | Component | AiDecodedTranscript", function (hooks) {
setupRenderingTest(hooks);
test("renders thinking, tool activity, and the response", async function (assert) {
const transcript = {
thinking: "Check <img src=x onerror=alert(1)>",
tool_calls: [
{
id: "call-1",
name: "search",
arguments: { query: "decoder" },
},
],
tool_results: [
{
call_id: "call-1",
type: "search_result",
result: { matches: ["Found it"] },
is_error: true,
},
],
response: "Done",
};
await render(
<template><AiDecodedTranscript @response={{transcript}} /></template>
);
assert
.dom(".ai-decoded-transcript__thinking[open]")
.exists("thinking is visible by default");
assert
.dom(".ai-decoded-transcript__thinking")
.includesText(
"Check <img src=x onerror=alert(1)>",
"thinking is rendered as text"
);
assert
.dom(".ai-decoded-transcript img")
.doesNotExist("model-generated HTML is escaped");
assert
.dom(".ai-decoded-transcript__item")
.exists({ count: 2 }, "the tool call and result are listed");
assert
.dom(".ai-decoded-transcript__item-heading")
.includesText("search", "the tool name is visible");
assert
.dom(".ai-decoded-transcript__code")
.includesText('"query": "decoder"', "arguments are formatted");
assert
.dom(".ai-decoded-transcript__item.--error")
.includesText('"matches": [', "structured error results are formatted");
assert
.dom(".ai-decoded-transcript__list[role='list']")
.exists({ count: 2 }, "tool collections retain list semantics");
assert
.dom(".ai-decoded-transcript__code[dir='ltr']")
.exists({ count: 2 }, "structured payloads use isolated text direction");
assert
.dom(".ai-decoded-transcript__code[tabindex]")
.doesNotExist("non-scrollable code blocks are not tab stops");
assert
.dom(".ai-decoded-transcript__section.--response")
.includesText("Done", "the final response is visible");
});
test("renders answer-only decoded responses without inferring transcript fields", async function (assert) {
const response = { response: '{"thinking":"about this"}' };
await render(
<template><AiDecodedTranscript @response={{response}} /></template>
);
assert
.dom(".ai-decoded-transcript__thinking")
.doesNotExist("model-authored JSON is not presented as thinking");
assert
.dom(".ai-decoded-transcript__section.--response")
.includesText(
'{"thinking":"about this"}',
"the model response remains intact"
);
});
test("ignores invalid and empty decoded responses", async function (assert) {
for (const response of [
null,
"response",
[],
{},
{ tool_calls: [null, {}], tool_results: [null, {}] },
]) {
await render(
<template><AiDecodedTranscript @response={{response}} /></template>
);
assert
.dom(".ai-decoded-transcript")
.doesNotExist("invalid decoded responses render nothing");
}
});
test("tolerates missing and unexpected tool data", async function (assert) {
const transcript = {
thinking: "Still checking",
tool_calls: [null, { name: "calculator", arguments: false }],
tool_results: "not an array",
};
await render(
<template><AiDecodedTranscript @response={{transcript}} /></template>
);
assert
.dom(".ai-decoded-transcript__item")
.exists({ count: 1 }, "invalid calls are omitted from the transcript");
assert
.dom(".ai-decoded-transcript__code")
.hasText("false", "non-object arguments remain readable");
assert
.dom(".ai-decoded-transcript__section")
.exists({ count: 1 }, "invalid result collections are omitted");
});
});