Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it } from 'vitest'
import type { AgentChatMessage } from '../../uiTypes'
import { findLoadedChatMessageMatches, wrapMatchIndex } from '../chatLocalSearch'
import { buildLoadedChatSemanticModels } from '../chatMessageSemantics'
function message(id: string, content: string, role: AgentChatMessage['role'] = 'assistant') {
return { id, content, role, timestamp: 1 } satisfies AgentChatMessage
}
describe('loaded chat local search', () => {
it('matches case-insensitively across renderer-owned loaded message data', () => {
expect(
findLoadedChatMessageMatches(
buildLoadedChatSemanticModels([
message('one', 'Alpha alpha'),
message('two', 'BETA Alpha'),
]),
'ALPHA'
)
).toEqual([
{ messageId: 'one', occurrence: 0 },
{ messageId: 'one', occurrence: 1 },
{ messageId: 'two', occurrence: 0 },
])
})
it('uses visible user text rather than hidden attachment prompt content', () => {
const content =
'Visible question\n\nUSER-ATTACHED CONTEXT: The user selected these capabilities/files for this message.\nPlugins: hidden-search-plugin.'
expect(
findLoadedChatMessageMatches(
buildLoadedChatSemanticModels([message('user', content, 'user')]),
'visible'
)
).toHaveLength(1)
expect(
findLoadedChatMessageMatches(
buildLoadedChatSemanticModels([message('user', content, 'user')]),
'hidden-search'
)
).toEqual([])
})
it('counts visible Markdown text without hidden destinations or syntax metadata', () => {
const content = [
'# Needle heading',
'[needle link](https://needle.example/path "needle title")',
'```needle-language',
'needle in code',
'```',
'[hidden-ref]: https://example.test/needle "needle title"',
].join('\n')
expect(
findLoadedChatMessageMatches(
buildLoadedChatSemanticModels([message('assistant', content)]),
'needle'
)
).toEqual([
{ messageId: 'assistant', occurrence: 0 },
{ messageId: 'assistant', occurrence: 1 },
{ messageId: 'assistant', occurrence: 2 },
])
})
it('keeps code lines that are not valid closers for a longer fence', () => {
const content = ['````typescript', '```needle', 'visible needle', '````'].join('\n')
expect(
findLoadedChatMessageMatches(
buildLoadedChatSemanticModels([message('assistant', content)]),
'needle'
)
).toEqual([
{ messageId: 'assistant', occurrence: 0 },
{ messageId: 'assistant', occurrence: 1 },
])
})
it('wraps forward and backward through loaded matches', () => {
expect(wrapMatchIndex(2, 3, 1)).toBe(0)
expect(wrapMatchIndex(0, 3, -1)).toBe(2)
expect(wrapMatchIndex(0, 0, 1)).toBe(0)
})
})