toor
1e9ea89d86
Import the client, server, skills, tests, and build scripts as the starting point for the rebuild, then trim the surfaces this project does not ship: Electron desktop packaging, the marketing website, and the ESP32 firmware tree. CI keeps the web build, e2e, lockfile, release, and Docker workflows. Rename the package and CLI entry points to hermes-ui, drop the prepare hook so npm ci no longer triggers a build, and drop the removed tsconfig and nodemon references. Rewrite AGENTS.md, ARCHITECTURE.md, DEVELOPMENT.md, and the READMEs so the documented package map and hard rules match this repository. reference_code/ is ignored: it is a read-only snapshot, not an input to the build.
390 lines
15 KiB
TypeScript
390 lines
15 KiB
TypeScript
import { describe, expect, it, vi } from 'vitest'
|
|
import { countTokens } from '../../packages/server/src/lib/context-compressor'
|
|
import {
|
|
estimateGroupHistoryMessageTokens,
|
|
groupBridgeReasoningDeltaFromEvent,
|
|
groupContextTokensWithFixedOverhead,
|
|
} from '../../packages/server/src/services/hermes/group-chat/agent-clients'
|
|
import { ContextEngine } from '../../packages/server/src/services/hermes/context-engine/compressor'
|
|
import type {
|
|
GatewayCaller,
|
|
MessageFetcher,
|
|
StoredMessage,
|
|
} from '../../packages/server/src/services/hermes/context-engine/types'
|
|
import {
|
|
sliceGroupMessagesCanonical,
|
|
sliceGroupMessagesForSnapshotTail,
|
|
sortGroupMessagesCanonical,
|
|
} from '../../packages/server/src/services/hermes/group-chat/group-message-ordering'
|
|
|
|
function makeMessage(overrides: Partial<StoredMessage>): StoredMessage {
|
|
return {
|
|
id: 'm1',
|
|
roomId: 'room-1',
|
|
senderId: 'user-1',
|
|
senderName: 'Alice',
|
|
content: 'hello',
|
|
timestamp: 1,
|
|
role: 'user',
|
|
...overrides,
|
|
}
|
|
}
|
|
|
|
function makeFetcher(messages: StoredMessage[], snapshot: ReturnType<MessageFetcher['getContextSnapshot']> = null): MessageFetcher {
|
|
return {
|
|
getMessagesForContext: vi.fn((_roomId: string, cutoff) => sliceGroupMessagesCanonical(messages, cutoff).messages),
|
|
getContextSnapshot: vi.fn(() => snapshot),
|
|
saveContextSnapshot: vi.fn(),
|
|
deleteContextSnapshot: vi.fn(),
|
|
}
|
|
}
|
|
|
|
function makeEngine(fetcher: MessageFetcher, summarize = vi.fn()): { engine: ContextEngine; summarize: ReturnType<typeof vi.fn> } {
|
|
const gatewayCaller: GatewayCaller = {
|
|
summarize: summarize.mockResolvedValue({ summary: 'Updated summary', sessionId: 'summary-session' }),
|
|
}
|
|
return {
|
|
engine: new ContextEngine({
|
|
config: { triggerTokens: 100_000, maxHistoryTokens: 32_000, tailMessageCount: 10, charsPerToken: 4, summarizationTimeoutMs: 30_000 },
|
|
messageFetcher: fetcher,
|
|
gatewayCaller,
|
|
}),
|
|
summarize,
|
|
}
|
|
}
|
|
|
|
describe('group chat fixed context cache helpers', () => {
|
|
it('adds cached fixed context to group chat message tokens', () => {
|
|
const history = [
|
|
{ role: 'user', content: '[Alice]: hello' },
|
|
{ role: 'assistant', content: '[Bot]: hi there' },
|
|
]
|
|
|
|
const messageTokens = estimateGroupHistoryMessageTokens(history)
|
|
|
|
expect(messageTokens).toBe(countTokens('[Alice]: hello') + countTokens('[Bot]: hi there'))
|
|
expect(groupContextTokensWithFixedOverhead(20_000, history)).toBe(20_000 + messageTokens)
|
|
})
|
|
|
|
it('signals fallback when fixed context is unavailable', () => {
|
|
expect(groupContextTokensWithFixedOverhead(undefined, [{ content: 'hello' }])).toBeUndefined()
|
|
expect(groupContextTokensWithFixedOverhead(null, [{ content: 'hello' }])).toBeUndefined()
|
|
})
|
|
|
|
it('keeps spinner thinking events out of persisted group-chat reasoning', () => {
|
|
expect(groupBridgeReasoningDeltaFromEvent({
|
|
event: 'thinking.delta',
|
|
text: '(◕‿◕✿) pondering...',
|
|
})).toBeNull()
|
|
expect(groupBridgeReasoningDeltaFromEvent({
|
|
event: 'reasoning.delta',
|
|
text: 'real reasoning',
|
|
})).toBe('real reasoning')
|
|
expect(groupBridgeReasoningDeltaFromEvent({
|
|
event: 'reasoning.delta',
|
|
text: '',
|
|
})).toBeNull()
|
|
})
|
|
})
|
|
|
|
describe('group chat context cursors', () => {
|
|
it('orders multipart assistant/toolcall/toolresult groups canonically before cursor slicing', () => {
|
|
const ordered = sortGroupMessagesCanonical([
|
|
makeMessage({ id: 'run-1_part_1_toolcall_weather', content: 'call-2', timestamp: 1_000, role: 'assistant' }),
|
|
makeMessage({ id: 'run-1_part_0_toolresult_weather', content: 'result-1', timestamp: 1_000, role: 'tool' }),
|
|
makeMessage({ id: 'run-1_part_1', content: 'assistant-2', timestamp: 1_000, role: 'assistant' }),
|
|
makeMessage({ id: 'run-1_part_0', content: 'assistant-1', timestamp: 1_000, role: 'assistant' }),
|
|
makeMessage({ id: 'run-1_part_1_toolresult_weather', content: 'result-2', timestamp: 1_000, role: 'tool' }),
|
|
makeMessage({ id: 'run-1_part_0_toolcall_weather', content: 'call-1', timestamp: 1_000, role: 'assistant' }),
|
|
makeMessage({ id: 'run-2', content: 'later run', timestamp: 2_000, role: 'assistant' }),
|
|
])
|
|
|
|
expect(ordered.map(message => message.id)).toEqual([
|
|
'run-1_part_0',
|
|
'run-1_part_0_toolcall_weather',
|
|
'run-1_part_0_toolresult_weather',
|
|
'run-1_part_1',
|
|
'run-1_part_1_toolcall_weather',
|
|
'run-1_part_1_toolresult_weather',
|
|
'run-2',
|
|
])
|
|
|
|
const snapshotTail = sliceGroupMessagesForSnapshotTail(ordered, 'run-1_part_0_toolresult_weather')
|
|
expect(snapshotTail.snapshotCursorFound).toBe(true)
|
|
expect(snapshotTail.messages.map(message => message.id)).toEqual([
|
|
'run-1_part_1',
|
|
'run-1_part_1_toolcall_weather',
|
|
'run-1_part_1_toolresult_weather',
|
|
'run-2',
|
|
])
|
|
})
|
|
|
|
it('uses the current message id as the same-timestamp context boundary', async () => {
|
|
const messages = sortGroupMessagesCanonical([
|
|
makeMessage({ id: 'm1', content: 'first', timestamp: 1_000 }),
|
|
makeMessage({ id: 'm2', content: 'second', timestamp: 1_000 }),
|
|
makeMessage({ id: 'm3', content: 'third', timestamp: 1_000 }),
|
|
])
|
|
const fetcher = makeFetcher(messages)
|
|
const { engine, summarize } = makeEngine(fetcher)
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[1],
|
|
})
|
|
|
|
expect(fetcher.getMessagesForContext).toHaveBeenCalledWith('room-1', { throughMessageId: 'm2' })
|
|
expect(result.conversationHistory.map(message => message.content)).toEqual([
|
|
'[Alice]: first',
|
|
'[Alice]: second',
|
|
])
|
|
expect(summarize).not.toHaveBeenCalled()
|
|
})
|
|
|
|
it('increments snapshots from lastMessageId even when timestamps are identical', async () => {
|
|
const messages = sortGroupMessagesCanonical([
|
|
makeMessage({ id: 'm1', content: 'first', timestamp: 1_000 }),
|
|
makeMessage({ id: 'm2', content: 'second', timestamp: 1_000 }),
|
|
makeMessage({ id: 'm3', content: 'third', timestamp: 1_000 }),
|
|
])
|
|
const fetcher = makeFetcher(messages, {
|
|
roomId: 'room-1',
|
|
summary: 'Earlier summary',
|
|
lastMessageId: 'm1',
|
|
lastMessageTimestamp: 1_000,
|
|
updatedAt: Date.now(),
|
|
})
|
|
const { engine, summarize } = makeEngine(fetcher)
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[2],
|
|
})
|
|
|
|
expect(result.conversationHistory.map(message => message.content)).toEqual([
|
|
'[Previous conversation summary]\nEarlier summary',
|
|
'I have reviewed the conversation history and understand the context.',
|
|
'[Alice]: second',
|
|
'[Alice]: third',
|
|
])
|
|
expect(summarize).not.toHaveBeenCalled()
|
|
})
|
|
|
|
it('preserves snapshot summaries when the snapshot anchor was pruned from retained history', async () => {
|
|
const messages = sortGroupMessagesCanonical([
|
|
makeMessage({ id: 'm2', content: 'second', timestamp: 1_000 }),
|
|
makeMessage({ id: 'm3', content: 'third', timestamp: 1_000 }),
|
|
])
|
|
const fetcher = makeFetcher(messages, {
|
|
roomId: 'room-1',
|
|
summary: 'Stale summary',
|
|
lastMessageId: 'm1',
|
|
lastMessageTimestamp: 1_000,
|
|
updatedAt: Date.now(),
|
|
})
|
|
const { engine, summarize } = makeEngine(fetcher)
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[1],
|
|
})
|
|
|
|
expect(result.conversationHistory.map(message => message.content)).toEqual([
|
|
'[Previous conversation summary]\nStale summary',
|
|
'I have reviewed the conversation history and understand the context.',
|
|
'[Alice]: second',
|
|
'[Alice]: third',
|
|
])
|
|
expect(result.meta.hadSnapshot).toBe(true)
|
|
expect(result.meta.verbatimCount).toBe(2)
|
|
expect(summarize).not.toHaveBeenCalled()
|
|
})
|
|
})
|
|
|
|
|
|
describe('group chat fallback trimming', () => {
|
|
it('drops oldest verbatim turns first when full compression falls back to trimming', async () => {
|
|
const messages = Array.from({ length: 6 }, (_value, index) => makeMessage({
|
|
id: `m${index + 1}`,
|
|
timestamp: index + 1,
|
|
content: `message-${index + 1} `.repeat(10),
|
|
}))
|
|
const fetcher = makeFetcher(messages)
|
|
const summarize = vi.fn().mockRejectedValue(new Error('summary unavailable'))
|
|
const gatewayCaller: GatewayCaller = { summarize }
|
|
const engine = new ContextEngine({
|
|
config: { triggerTokens: 1, maxHistoryTokens: 60, tailMessageCount: 2, charsPerToken: 4, summarizationTimeoutMs: 30_000 },
|
|
messageFetcher: fetcher,
|
|
gatewayCaller,
|
|
})
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[messages.length - 1],
|
|
contextTokenEstimator: vi.fn().mockResolvedValue(999),
|
|
})
|
|
|
|
expect(result.conversationHistory.some(message => message.content.includes('message-1'))).toBe(false)
|
|
expect(result.conversationHistory.some(message => message.content.includes('message-2'))).toBe(false)
|
|
expect(result.conversationHistory).toHaveLength(2)
|
|
expect(result.conversationHistory[0]?.content).toContain('message-5')
|
|
expect(result.conversationHistory[1]?.content).toContain('message-6')
|
|
})
|
|
|
|
it('preserves the summary prefix while trimming oldest post-summary turns first', async () => {
|
|
const messages = Array.from({ length: 5 }, (_value, index) => makeMessage({
|
|
id: `m${index + 1}`,
|
|
timestamp: index + 1,
|
|
content: `message-${index + 1} `.repeat(10),
|
|
}))
|
|
const fetcher = makeFetcher(messages, {
|
|
roomId: 'room-1',
|
|
summary: 'Earlier summary',
|
|
lastMessageId: 'm1',
|
|
lastMessageTimestamp: 1,
|
|
updatedAt: Date.now(),
|
|
})
|
|
const summarize = vi.fn().mockRejectedValue(new Error('summary unavailable'))
|
|
const gatewayCaller: GatewayCaller = { summarize }
|
|
const engine = new ContextEngine({
|
|
config: { triggerTokens: 1, maxHistoryTokens: 90, tailMessageCount: 2, charsPerToken: 4, summarizationTimeoutMs: 30_000 },
|
|
messageFetcher: fetcher,
|
|
gatewayCaller,
|
|
})
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[messages.length - 1],
|
|
contextTokenEstimator: vi.fn().mockResolvedValue(999),
|
|
})
|
|
|
|
expect(result.conversationHistory[0]).toEqual({ role: 'user', content: '[Previous conversation summary]\nEarlier summary' })
|
|
expect(result.conversationHistory[1]).toEqual({ role: 'assistant', content: 'I have reviewed the conversation history and understand the context.' })
|
|
expect(result.conversationHistory.some(message => message.content.includes('message-2'))).toBe(false)
|
|
expect(result.conversationHistory[2]?.content).toContain('message-4')
|
|
expect(result.conversationHistory[3]?.content).toContain('message-5')
|
|
})
|
|
|
|
it('keeps the newest verbatim turn even when full fallback budget is smaller than a single message', async () => {
|
|
const messages = Array.from({ length: 4 }, (_value, index) => makeMessage({
|
|
id: `m${index + 1}`,
|
|
timestamp: index + 1,
|
|
content: `message-${index + 1} `.repeat(10),
|
|
}))
|
|
const fetcher = makeFetcher(messages)
|
|
const summarize = vi.fn().mockRejectedValue(new Error('summary unavailable'))
|
|
const gatewayCaller: GatewayCaller = { summarize }
|
|
const engine = new ContextEngine({
|
|
config: { triggerTokens: 1, maxHistoryTokens: 1, tailMessageCount: 2, charsPerToken: 4, summarizationTimeoutMs: 30_000 },
|
|
messageFetcher: fetcher,
|
|
gatewayCaller,
|
|
})
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[messages.length - 1],
|
|
contextTokenEstimator: vi.fn().mockResolvedValue(999),
|
|
})
|
|
|
|
expect(result.conversationHistory).toHaveLength(1)
|
|
expect(result.conversationHistory[0]?.content).toContain('message-4')
|
|
})
|
|
|
|
it('keeps the summary prefix and newest verbatim turn when snapshot fallback budget is tiny', async () => {
|
|
const messages = Array.from({ length: 5 }, (_value, index) => makeMessage({
|
|
id: `m${index + 1}`,
|
|
timestamp: index + 1,
|
|
content: `message-${index + 1} `.repeat(10),
|
|
}))
|
|
const fetcher = makeFetcher(messages, {
|
|
roomId: 'room-1',
|
|
summary: 'Earlier summary',
|
|
lastMessageId: 'm1',
|
|
lastMessageTimestamp: 1,
|
|
updatedAt: Date.now(),
|
|
})
|
|
const summarize = vi.fn().mockRejectedValue(new Error('summary unavailable'))
|
|
const gatewayCaller: GatewayCaller = { summarize }
|
|
const engine = new ContextEngine({
|
|
config: { triggerTokens: 1, maxHistoryTokens: 1, tailMessageCount: 2, charsPerToken: 4, summarizationTimeoutMs: 30_000 },
|
|
messageFetcher: fetcher,
|
|
gatewayCaller,
|
|
})
|
|
|
|
const result = await engine.buildContext({
|
|
roomId: 'room-1',
|
|
agentId: 'agent-1',
|
|
agentName: 'Worker',
|
|
agentDescription: '',
|
|
agentSocketId: 'agent-socket',
|
|
roomName: 'general',
|
|
memberNames: ['Alice'],
|
|
members: [{ userId: 'user-1', name: 'Alice', description: '' }],
|
|
upstream: '',
|
|
apiKey: null,
|
|
currentMessage: messages[messages.length - 1],
|
|
contextTokenEstimator: vi.fn().mockResolvedValue(999),
|
|
})
|
|
|
|
expect(result.conversationHistory).toEqual([
|
|
{ role: 'user', content: '[Previous conversation summary]\nEarlier summary' },
|
|
{ role: 'assistant', content: 'I have reviewed the conversation history and understand the context.' },
|
|
expect.objectContaining({ role: 'user', content: expect.stringContaining('message-5') }),
|
|
])
|
|
})
|
|
})
|