admin/Hermes-ui
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(sessions): harden compressed session lineage projection (#226)

Browse Source 
- Project compressed roots to their continuation tip in session lists.
- Search title/content candidates through logical compression lineage.
- Hydrate detail views along the requested continuation branch while preserving requested ids.
- Scope model-context cache lookup by provider to avoid same-name cross-provider matches.
- Add regression coverage for lineage and provider lookup behavior.
This commit is contained in:
Zhicheng Han
2026-04-26 04:10:01 +02:00
committed by GitHub
parent d0bd09ae57
commit f1a6d97c8b
5 changed files with 747 additions and 230 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/server/model-context.test.ts
+124 -51
View File
@@ -1,87 +1,160 @@
import { mkdirSync, writeFileSync } from 'fs'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { mkdirSync, mkdtempSync, rmSync, writeFileSync } from 'fs'
import { join } from 'path'
import { tmpdir } from 'os'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
function makeHome() {
const root = join(tmpdir(), `wui-model-context-${Date.now()}-${Math.random().toString(36).slice(2)}`)
const hermes = join(root, '.hermes')
mkdirSync(hermes, { recursive: true })
return { root, hermes }
let homeDir = ''
function hermesPath(...parts: string[]) {
return join(homeDir, '.hermes', ...parts)
}
function writeConfig(hermes: string, yaml: string) {
writeFileSync(join(hermes, 'config.yaml'), yaml)
function writeConfig(content: string) {
mkdirSync(hermesPath(), { recursive: true })
writeFileSync(hermesPath('config.yaml'), content)
}
function writeModelsCache(hermes: string) {
writeFileSync(join(hermes, 'models_dev_cache.json'), JSON.stringify({
openai: {
models: {
'gpt-5.5': { limit: { context: 1_050_000 } },
'gpt-5.4': { limit: { context: 1_050_000 } },
},
},
google: {
models: {
'gemini-3.1-pro-preview': { limit: { context: 1_000_000 } },
},
},
}))
function writeModelsCache(data: Record<string, unknown>) {
mkdirSync(hermesPath(), { recursive: true })
writeFileSync(hermesPath('models_dev_cache.json'), JSON.stringify(data))
}
async function importContextService(home: string) {
async function loadModelContext() {
vi.resetModules()
vi.stubEnv('HOME', home)
return await import('../../packages/server/src/services/hermes/model-context')
vi.doMock('os', async () => ({
...(await vi.importActual<typeof import('os')>('os')),
homedir: () => homeDir,
}))
return import('../../packages/server/src/services/hermes/model-context')
}
describe('model context length resolution', () => {
describe('getModelContextLength', () => {
beforeEach(() => {
vi.unstubAllEnvs()
homeDir = mkdtempSync(join(tmpdir(), 'hwui-model-context-'))
})
afterEach(() => {
vi.unstubAllEnvs()
vi.resetModules()
vi.doUnmock('os')
if (homeDir) rmSync(homeDir, { recursive: true, force: true })
homeDir = ''
})
it('does not borrow OpenAI context metadata for an openai-codex model with the same name', async () => {
const { root, hermes } = makeHome()
writeConfig(hermes, 'model:\n provider: openai-codex\n default: gpt-5.5\n')
writeModelsCache(hermes)
it('does not borrow a same-named model context from another provider when the configured provider is uncached', async () => {
writeConfig(`model:\n default: gpt-5.5\n provider: openai-codex\n`)
writeModelsCache({
openai: {
models: {
'gpt-5.5': { limit: { context: 1_050_000 } },
},
},
})
const { getModelContextLength } = await importContextService(root)
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(200_000)
})
it('still honors explicit model.context_length before provider-aware cache lookup', async () => {
const { root, hermes } = makeHome()
writeConfig(hermes, 'model:\n provider: openai-codex\n default: gpt-5.5\n context_length: 272000\n')
writeModelsCache(hermes)
it('does not scan other providers when the configured provider exists without that model', async () => {
writeConfig(`model:\n default: gpt-5.5\n provider: openai-codex\n`)
writeModelsCache({
'openai-codex': {
models: {
'gpt-5.4': { limit: { context: 200_000 } },
},
},
openai: {
models: {
'gpt-5.5': { limit: { context: 1_050_000 } },
},
},
})
const { getModelContextLength } = await importContextService(root)
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(272_000)
expect(getModelContextLength()).toBe(200_000)
})
it('preserves providerless legacy lookup by model name', async () => {
const { root, hermes } = makeHome()
writeConfig(hermes, 'model:\n default: gpt-5.5\n')
writeModelsCache(hermes)
it('uses the configured provider cache entry when the provider matches', async () => {
writeConfig(`model:\n default: gpt-5.5\n provider: openai\n`)
writeModelsCache({
openai: {
models: {
'gpt-5.5': { limit: { context: 1_050_000 } },
},
},
})
const { getModelContextLength } = await importContextService(root)
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(1_050_000)
})
it('uses intentional cache provider aliases without conflating openai-codex with openai', async () => {
const { root, hermes } = makeHome()
writeConfig(hermes, 'model:\n provider: gemini\n default: gemini-3.1-pro-preview\n')
writeModelsCache(hermes)
it('keeps legacy model-name cache lookup when no provider is configured', async () => {
writeConfig(`model:\n default: gpt-5.5\n`)
writeModelsCache({
openai: {
models: {
'gpt-5.5': { limit: { context: 1_050_000 } },
},
},
})
const { getModelContextLength } = await importContextService(root)
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(1_050_000)
})
it('keeps providerless legacy lookup on global exact matches before prefixed suffix matches', async () => {
writeConfig(`model:\n default: gpt-5\n`)
writeModelsCache({
vercel: {
models: {
'openai/gpt-5': { limit: { context: 1_000_000 } },
},
},
openai: {
models: {
'gpt-5': { limit: { context: 400_000 } },
},
},
})
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(400_000)
})
it('maps WUI provider keys to model-cache provider keys before looking up limits', async () => {
writeConfig(`model:\n default: gemini-3.1-pro-preview\n provider: gemini\n`)
writeModelsCache({
google: {
models: {
'gemini-3.1-pro-preview': { limit: { context: 1_000_000 } },
},
},
})
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(1_000_000)
})
it('uses gateway provider aliases with prefixed model names inside the aliased provider only', async () => {
writeConfig(`model:\n default: openai/gpt-5\n provider: ai-gateway\n`)
writeModelsCache({
vercel: {
models: {
'openai/gpt-5': { limit: { context: 1_000_000 } },
},
},
openai: {
models: {
'gpt-5': { limit: { context: 400_000 } },
},
},
})
const { getModelContextLength } = await loadModelContext()
expect(getModelContextLength()).toBe(1_000_000)
})
tests/server/sessions-db-lineage.test.ts
+303
View File
@@ -0,0 +1,303 @@
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { mkdtempSync, rmSync } from 'fs'
import { tmpdir } from 'os'
import { join } from 'path'
import { DatabaseSync } from 'node:sqlite'
const profileDir = vi.hoisted(() => ({ value: '' }))
vi.mock('../../packages/server/src/services/hermes/hermes-profile', () => ({
getActiveProfileDir: () => profileDir.value,
}))
function createStateDb(path: string) {
const db = new DatabaseSync(path)
db.exec(`
CREATE TABLE sessions (
id TEXT PRIMARY KEY,
source TEXT NOT NULL,
user_id TEXT,
model TEXT,
title TEXT,
started_at REAL,
ended_at REAL,
end_reason TEXT,
message_count INTEGER,
tool_call_count INTEGER,
input_tokens INTEGER,
output_tokens INTEGER,
cache_read_tokens INTEGER,
cache_write_tokens INTEGER,
reasoning_tokens INTEGER,
billing_provider TEXT,
estimated_cost_usd REAL,
actual_cost_usd REAL,
cost_status TEXT,
parent_session_id TEXT
);
CREATE TABLE messages (
id INTEGER PRIMARY KEY,
session_id TEXT NOT NULL,
role TEXT NOT NULL,
content TEXT,
tool_call_id TEXT,
tool_calls TEXT,
tool_name TEXT,
timestamp REAL,
token_count INTEGER,
finish_reason TEXT,
reasoning TEXT,
reasoning_details TEXT,
codex_reasoning_items TEXT,
reasoning_content TEXT
);
CREATE VIRTUAL TABLE messages_fts USING fts5(content);
`)
return db
}
function insertSession(
db: DatabaseSync,
row: {
id: string
source?: string
parent_session_id?: string | null
title?: string
started_at: number
ended_at?: number | null
end_reason?: string | null
message_count?: number
model?: string
},
) {
db.prepare(`
INSERT INTO sessions (
id, source, user_id, model, title, started_at, ended_at, end_reason,
message_count, tool_call_count, input_tokens, output_tokens,
cache_read_tokens, cache_write_tokens, reasoning_tokens, billing_provider,
estimated_cost_usd, actual_cost_usd, cost_status, parent_session_id
) VALUES (?, ?, '', ?, ?, ?, ?, ?, ?, 0, 0, 0, 0, 0, 0, '', 0, NULL, '', ?)
`).run(
row.id,
row.source || 'api_server',
row.model || 'gpt-5.5',
row.title || '',
row.started_at,
row.ended_at ?? null,
row.end_reason ?? null,
row.message_count ?? 1,
row.parent_session_id ?? null,
)
}
function insertMessage(
db: DatabaseSync,
row: {
id: number
session_id: string
role?: string
content: string
timestamp: number
},
) {
db.prepare(`
INSERT INTO messages (
id, session_id, role, content, tool_call_id, tool_calls, tool_name,
timestamp, token_count, finish_reason, reasoning, reasoning_details,
codex_reasoning_items, reasoning_content
) VALUES (?, ?, ?, ?, NULL, NULL, NULL, ?, NULL, NULL, NULL, NULL, NULL, NULL)
`).run(row.id, row.session_id, row.role || 'user', row.content, row.timestamp)
db.prepare('INSERT INTO messages_fts(rowid, content) VALUES (?, ?)').run(row.id, row.content)
}
function seedCompressionChain(db: DatabaseSync) {
insertSession(db, {
id: 'root',
source: 'api_server',
title: 'Mermaid fix',
started_at: 100,
ended_at: 200,
end_reason: 'compression',
message_count: 2,
})
insertSession(db, {
id: 'middle',
source: 'cli',
parent_session_id: 'root',
title: 'Mermaid fix #2',
started_at: 201,
ended_at: 300,
end_reason: 'compression',
message_count: 3,
})
insertSession(db, {
id: 'tip',
source: 'cli',
parent_session_id: 'middle',
title: 'Mermaid fix #3',
started_at: 301,
ended_at: null,
end_reason: null,
message_count: 4,
})
insertMessage(db, { id: 1, session_id: 'root', content: 'root turn', timestamp: 101 })
insertMessage(db, { id: 2, session_id: 'middle', content: 'middle turn', timestamp: 202 })
insertMessage(db, { id: 3, session_id: 'tip', content: 'tip lineageunique turn', timestamp: 302 })
}
describe('session DB compression lineage', () => {
let tempDir = ''
let db: DatabaseSync | null = null
beforeEach(() => {
vi.resetModules()
tempDir = mkdtempSync(join(tmpdir(), 'wui-session-lineage-'))
profileDir.value = tempDir
db = createStateDb(join(tempDir, 'state.db'))
})
afterEach(() => {
db?.close()
db = null
if (tempDir) rmSync(tempDir, { recursive: true, force: true })
})
it('projects compressed root summaries to the latest continuation tip', async () => {
seedCompressionChain(db!)
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.listSessionSummaries(undefined, 20)
expect(rows).toHaveLength(1)
expect(rows[0]).toMatchObject({
id: 'tip',
title: 'Mermaid fix #3',
message_count: 4,
end_reason: null,
preview: 'tip lineageunique turn',
started_at: 100,
})
})
it('returns the projected logical session when search matches continuation content', async () => {
seedCompressionChain(db!)
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('lineageunique', undefined, 20)
expect(rows).toHaveLength(1)
expect(rows[0]).toMatchObject({
id: 'tip',
title: 'Mermaid fix #3',
matched_message_id: 3,
})
expect(rows[0].snippet).toContain('lineageunique')
})
it('hydrates the full compression chain when detail is requested by projected tip id', async () => {
seedCompressionChain(db!)
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const detail = await mod.getSessionDetailFromDb('tip')
expect(detail).toMatchObject({
id: 'tip',
title: 'Mermaid fix #3',
message_count: 9,
thread_session_count: 3,
})
expect(detail?.messages.map(message => message.session_id)).toEqual(['root', 'middle', 'tip'])
})
it('follows only the latest compression continuation child when a parent has multiple children', async () => {
insertSession(db!, {
id: 'root',
started_at: 100,
ended_at: 200,
end_reason: 'compression',
message_count: 1,
})
insertSession(db!, {
id: 'older-child',
parent_session_id: 'root',
title: 'Older branch',
started_at: 201,
ended_at: null,
end_reason: null,
message_count: 1,
})
insertSession(db!, {
id: 'latest-child',
parent_session_id: 'root',
title: 'Latest branch',
started_at: 205,
ended_at: null,
end_reason: null,
message_count: 1,
})
insertMessage(db!, { id: 11, session_id: 'root', content: 'root', timestamp: 101 })
insertMessage(db!, { id: 12, session_id: 'older-child', content: 'older should not merge', timestamp: 202 })
insertMessage(db!, { id: 13, session_id: 'latest-child', content: 'latest should merge', timestamp: 206 })
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const detail = await mod.getSessionDetailFromDb('root')
expect(detail).toMatchObject({
id: 'root',
title: 'Latest branch',
message_count: 2,
thread_session_count: 2,
})
expect(detail?.messages.map(message => message.session_id)).toEqual(['root', 'latest-child'])
const olderDetail = await mod.getSessionDetailFromDb('older-child')
expect(olderDetail).toMatchObject({
id: 'older-child',
title: 'Older branch',
message_count: 2,
thread_session_count: 2,
})
expect(olderDetail?.messages.map(message => message.session_id)).toEqual(['root', 'older-child'])
const olderSearch = await mod.searchSessionSummaries('older should', undefined, 20)
expect(olderSearch[0]).toMatchObject({
id: 'older-child',
title: 'Older branch',
matched_message_id: 12,
})
})
it('applies source filters before search candidate limiting', async () => {
for (let index = 0; index < 105; index += 1) {
insertSession(db!, {
id: `cli-${index}`,
source: 'cli',
title: `needle cli ${index}`,
started_at: 1000 + index,
ended_at: null,
end_reason: null,
})
}
insertSession(db!, {
id: 'telegram-match',
source: 'telegram',
title: 'needle telegram target',
started_at: 10,
ended_at: null,
end_reason: null,
})
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('needle', 'telegram', 1)
expect(rows).toHaveLength(1)
expect(rows[0]).toMatchObject({
id: 'telegram-match',
source: 'telegram',
title: 'needle telegram target',
})
})
})
tests/server/sessions-db.test.ts
+12 -12
View File
@@ -67,8 +67,8 @@ describe('session DB summaries', () => {
const rows = await mod.listSessionSummaries(undefined, 50)
expect(databaseSyncMock).toHaveBeenCalledWith('/tmp/hermes-profile/state.db', { open: true, readOnly: true })
expect(prepareMock).toHaveBeenCalledWith(expect.stringContaining("AND s.source != 'tool'"))
expect(allMock).toHaveBeenCalledWith(50)
expect(prepareMock).toHaveBeenCalledWith(expect.stringContaining("s.source != 'tool'"))
expect(allMock).toHaveBeenCalledWith(200)
expect(closeMock).toHaveBeenCalled()
expect(rows).toEqual([
{
@@ -127,8 +127,8 @@ describe('session DB summaries', () => {
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.listSessionSummaries('telegram', 2)
expect(prepareMock).toHaveBeenCalledWith(expect.stringContaining('AND s.source = ?'))
expect(allMock).toHaveBeenCalledWith('telegram', 2)
expect(prepareMock).toHaveBeenCalledWith(expect.stringContaining("s.source != 'tool'"))
expect(allMock).toHaveBeenCalledWith('telegram', 8)
expect(rows[0].last_active).toBe(1710000100)
expect(rows[0].source).toBe('telegram')
expect(rows[0].title).toBe('preview text')
@@ -375,8 +375,8 @@ describe('session DB summaries', () => {
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('node.js*', undefined, 10)
expect(titleAllMock).toHaveBeenCalledWith('%node.js%', 10)
expect(contentAllMock).toHaveBeenCalledWith('"node.js"*', 40)
expect(titleAllMock).toHaveBeenCalledWith('%node.js%', 200)
expect(contentAllMock).toHaveBeenCalledWith('"node.js"*', 200)
expect(likeAllMock).not.toHaveBeenCalled()
expect(rows).toHaveLength(2)
expect(rows[0].id).toBe('node-wildcard-title-1')
@@ -444,8 +444,8 @@ describe('session DB summaries', () => {
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('"node.js"*', undefined, 10)
expect(titleAllMock).toHaveBeenCalledWith('%node.js%', 10)
expect(contentAllMock).toHaveBeenCalledWith('"node.js"*', 40)
expect(titleAllMock).toHaveBeenCalledWith('%node.js%', 200)
expect(contentAllMock).toHaveBeenCalledWith('"node.js"*', 200)
expect(likeAllMock).not.toHaveBeenCalled()
expect(rows).toHaveLength(2)
expect(rows[0].id).toBe('node-quoted-title-1')
@@ -486,7 +486,7 @@ describe('session DB summaries', () => {
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('naïve.js', undefined, 10)
expect(contentAllMock).toHaveBeenCalledWith('"naïve.js"', 40)
expect(contentAllMock).toHaveBeenCalledWith('"naïve.js"', 200)
expect(likeAllMock).not.toHaveBeenCalled()
expect(rows).toHaveLength(1)
expect(rows[0].id).toBe('unicode-dot-1')
@@ -526,7 +526,7 @@ describe('session DB summaries', () => {
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('100%', undefined, 10)
expect(titleAllMock).toHaveBeenCalledWith('%100\\%%', 10)
expect(titleAllMock).toHaveBeenCalledWith('%100\\%%', 200)
expect(rows).toHaveLength(1)
expect(rows[0].id).toBe('percent-1')
})
@@ -567,7 +567,7 @@ describe('session DB summaries', () => {
const rows = await mod.searchSessionSummaries('记忆断裂', undefined, 10)
expect(contentAllMock).not.toHaveBeenCalled()
expect(likeAllMock).toHaveBeenCalledWith('记忆断裂', '%记忆断裂%', 40)
expect(likeAllMock).toHaveBeenCalledWith('记忆断裂', '%记忆断裂%', 200)
expect(rows).toHaveLength(1)
expect(rows[0].id).toBe('cjk-literal-1')
})
@@ -636,7 +636,7 @@ describe('session DB summaries', () => {
const mod = await import('../../packages/server/src/db/hermes/sessions-db')
const rows = await mod.searchSessionSummaries('记忆断裂', undefined, 10)
expect(likeAllMock).toHaveBeenCalledWith('记忆断裂', '%记忆断裂%', 40)
expect(likeAllMock).toHaveBeenCalledWith('记忆断裂', '%记忆断裂%', 200)
expect(rows).toHaveLength(2)
expect(rows[0].id).toBe('cjk-1')
expect(rows[1].id).toBe('cjk-title-1')