kimi-code/packages/agent-core/test/session/init.test.ts
7Sageer 59dde734f3
feat(agent-core): unify the v1 MCP management plane (#2858)
* feat(agent-core): unify the v1 MCP management plane

- McpServerRegistry: one config view over global (layered mcp.json),
  plugin (manifests, read-only, final effective config), and caller
  (SDK-injected) servers; name collisions keep both entries.
- Write plane: add/update/removeGlobalMcpServer mutate the user-level
  file and push into live sessions; getGlobalMcpServer returns the
  effective config; mutations of read-only entries are rejected.
- testGlobalMcpServer accepts an inline config; addSessionMcpServer
  connects a server in one live session with an optional persist flag;
  reconnect accepts a replacement config and re-resolves via the registry.
- One process-wide McpOAuthService shared with every session: obtained_at
  stamps, offline token state, single-flight and proactive refresh, and
  credential events. Sessions self-subscribe in the constructor, so even
  initializing sessions see every event; token writes serialize through
  the process-local OAuthTokenTransaction per credential identity.
- inspectAppMcpServers + locator-addressed begin/complete/cancel/reset
  cover plugin servers; inspection output redacts env/headers to sorted
  key lists; locator OAuth ops reject ambiguous shared runtime names.
- The legacy auth-status surface reads the registry (offline by default,
  verify=true probes) and never mutates credentials.
- VS Code panel receives source/origin/mutable and hides mutating
  actions on read-only entries.
- v2 client facade in node-sdk mirrors the surface over agent-core-v2
  (plugin inventory stays v1-only for now).

* fix(agent-core): close the v1 MCP live-session reconciliation gaps

Recompute each live session's MCP target from the registry's runtime
resolution (enabled plugin > project layer > user file; caller injection
shadows everything) behind every config mutation, instead of per-path
patching: shadowed file layers recover when a plugin winner is disabled or
removed, removing a user-level entry resurrects its project-layer shadow,
disabled plugin descriptors no longer block removals, persisted session
adds validate against the session's project layer and broadcast to other
live sessions, and per-session sync failures are logged with context.

Session status entries and read-only management entries now report
redacted config views (envKeys/headerKeys instead of literal env/headers
values); core-internal reconciliation compares full configs via the
connection manager's raw-entry accessor.

OAuth: interactive flows are serialized per credential (concurrent
begins join the in-flight flow instead of clobbering its PKCE/state), a
malformed credential meta sidecar no longer aborts core start, grants
inside the refresh-ahead window refresh immediately while far-future
grants re-arm through a max-length timer, and the service shuts its
timers and flows down with KimiCore/SDKRpcClient close.

* fix(agent-core): route the proactive MCP OAuth refresh through the token transaction

 refreshNow ran its /token request with the SDK default fetch, outside the
 credential-serializing OAuthTokenTransaction that every other token write
 uses; a slower response carrying an older rotating refresh token could
 overwrite a newer grant written by a concurrent transport-side refresh.

* fix(agent-core): keep disabled MCP servers out of auth-state classification

The unified mcpServerAuthState dropped the previous enabled short-circuit,
so a disabled oauth-flagged server reported oauth-required — or was even
probed over the network — instead of not-applicable.

* fix(kimi-code-sdk): short-circuit disabled MCP servers in the v2 auth-status classifier

The v2 parity copy of v1's mcpServerAuthState missed the same enabled
guard v1 just regained; a disabled oauth-flagged entry would report
oauth-required (or be probed). The parity suite now pins the disabled
case on both engines.

* fix(kimi-code): refresh the VS Code MCP list with the workspace cwd after mutations

The add/update/remove RPCs return a cwd-less management list, so the
webview broadcast dropped project-layer entries until the next full load;
re-list with the workspace cwd after every mutation instead.

* fix(agent-core): keep SDK token saves matched to the OAuth token transaction

saveTokens stamped obtained_at onto a fresh object before calling
tokenTransaction.save, so it never matched the exact payload the
transaction recorded for a grant fetch; the consume path was dead and
every save re-wrote. Between the fetch and the SDK callback an intervening
clear could then be overwritten — the resurrected grant came back after a
reset. The write callback stamps the durable record instead.

* fix(agent-core): reject ambiguous legacy name-based MCP auth lookups

The legacy begin/reset auth RPCs took the registry's first name match,
silently starting OAuth for one entry of a runtime-name collision while
the locator path refused the same ambiguity; align them on the shared
uniqueness rule and point callers at the locator-addressed variants.

* fix(agent-core): propagate registry errors during live-session MCP sync

resolveMcpRuntimeTarget collapsed every registry failure into "no target":
a project config file that turned malformed mid-session made sync treat a
still-configured server as gone (tearing down the live connection) and
made config-aware reconnects report "no longer configured" instead of the
actionable config error. Absence still resolves to undefined; malformed
config now propagates — per-session sync logs and keeps the entry, and
reconnect surfaces config.invalid.

* fix(agent-core): close the remaining registry-error and ambiguity gaps

The management guard lookup mapped every registry failure to "absent",
so a malformed project config let a persisted session add write a
user-level entry over an unknown state; only not-found is a miss now. And
the name-only connection test now shares the auth paths' uniqueness rule
instead of probing the first match of a runtime-name collision.

* fix(agent-core): probe the enabled MCP entry under a disabled-name collision

The name-only connection test counted enabled matches for its ambiguity
guard but still probed the first registry match, and the file layers list
before plugins. With a disabled file entry shadowing an enabled plugin of
the same runtime name, Test probed the disabled entry instead of the one a
live session would run. Select the sole enabled match, falling back to the
first entry only when every match is disabled so it reports as disabled.

* fix(agent-core): let session-local MCP adds shadow plugin entries

Caller injection shadows every registry source at session start, plugins
included, and reconciliation leaves caller entries untouched; the live
non-persist add path rejected plugin-owned names anyway, so SDK clients
could not apply the same per-session override without a restart. Gate the
plugin-source rejection on persist: session-local adds connect as caller,
while persisted adds stay rejected as user-level writes behind a read-only
owner.

* fix(agent-core): normalize session MCP names before connecting

The persisted store trims server names, but addSessionMcpServer used the
raw name for the live connect and cross-session reconciliation: a padded
name persisted under the trimmed key while the requesting session ran and
reconciled the raw one, and a blank name connected with no identity at
all. Normalize once up front (rejecting blank) so the store write, the
session entry, and reconciliation agree on the same server.

* fix(agent-core,node-sdk): close the collision-selection and probe-freshness gaps

The legacy name-only auth resolver started from the first registry match,
so a disabled file-layer shadow plus an enabled plugin of the same runtime
name was misread as an ambiguity conflict; select the sole enabled match
before judging ambiguity, exactly like the test probe path. On the v2
client, addSessionMcpServer connected the raw name while the store wrote
the trimmed key — normalize once for both, and route the verify-triggered
auth probes through the per-call OAuth service instead of the cached one
whose providers snapshot tokens at construction, so a grant saved after
the first probe is honored.

* fix(agent-core): normalize global MCP mutation names and guard disabled reconnect swaps

The global add/update/remove mutations guarded and reconciled with the raw
server name while the store persisted the trimmed key, so a padded name
left live sessions unreconciled and could slip past the plugin read-only
guard; normalize once before lookup, persistence, and reconciliation. And
a config-carrying reconnect assigned the replacement before the disabled
check fired, leaving a connected entry that reported the disabled config;
reject disabled replacements before mutating, keeping the same error.

* fix(agent-core): skip proactive refresh while an interactive flow owns the credential

refreshNow reset the shared provider's flow state before and after the
token request; when a proactive timer (or a manual refresh) fired while
beginAuthorization was waiting on the browser callback for the same store
key, that wiped the redirect URL, PKCE verifier, and state the in-flight
flow needed — complete() then failed the exchange even though the user
authorized. Refresh now skips when an interactive flow is active for the
credential: the flow delivers fresh tokens on completion, and the 401
transport path is the backstop if it fails.

* fix(agent-core): allow global MCP adds over disabled plugin descriptors

A disabled plugin entry is absent from the runtime target, but the
read-only guard still treated it as the owner, so a user-level fallback
could only exist if it predated the plugin disable. Relax the shared
guard: disabled plugin descriptors never block mutations (disabled
project entries still shadow the user file and keep their rejection).

* fix(node-sdk): close the v2 session-MCP parity gaps

A v2 reconnect with an explicit enabled:false replacement config used
connect()'s upsert semantics — closing the live client and reporting
success where v1's manager reconnect rejects before applying anything;
reject disabled replacements up front with the same error. And a persisted
v2 session add never consulted the workspace config, so a same-named
project-layer entry was silently shadowed: the user-level write never
takes effect while the direct workspace-manager upsert displaces the
project config for every live session. Resolve the workspace layers and
reject like v1's read-only rule.

* fix(agent-core): keep __proto__-named MCP servers through config parsing

A z.record() parse rebuilds its output via property assignment, so a
server literally named __proto__ hit the prototype setter and vanished
before validation; the layer merge then repeated the same trap with plain
object accumulators. Parse the server map entry-by-entry over the JSON own
keys and accumulate into null-prototype maps, so session startup and the
unified registry keep the declared server and its origin.

* fix(node-sdk): begin v2 MCP auth against a fresh OAuth service

The v2 begin path ran through the cached globalMcpOAuth, whose providers
snapshot tokens at construction: a grant another process saved (or reset)
after that cache materialized was invisible, so begin could open a browser
flow over a valid grant, or report already-authorized off a removed one.
Build the service per call — the read path and the verify probes already
do — and route the status list through the same helper. The test fixture
grows a real token endpoint honoring one rotating refresh token; the
regression fails against the cached-service implementation on v2.

* fix(agent-core): broadcast SDK-driven MCP token invalidations to live sessions

* test(agent-core-v2): give the no-op reconnect test runtime plumbing

The branch added the case against a bare McpConnectionManager, but #2961
made stdio connects resolve the runtime through runtimeResolver, matching
every other case in the file.
2026-08-17 13:19:51 +08:00

1016 lines
34 KiB
TypeScript

import { access, mkdir, mkdtemp, rm, writeFile } from 'node:fs/promises';
import { tmpdir } from 'node:os';
import { join } from 'pathe';
import { testKaos } from '../fixtures/test-kaos';
import type { ProviderConfig, ToolCall } from '@moonshot-ai/kosong';
import type { Kaos, StatResult } from '@moonshot-ai/kaos';
import { afterEach, describe, expect, it, vi } from 'vitest';
import type { Agent, AgentOptions } from '../../src/agent';
import { trimTrailingOpenToolExchange } from '../../src/agent/context/projector';
import type { KimiConfig } from '../../src/config';
import { FlagResolver } from '../../src/flags';
import { McpOAuthService, type McpOAuthEvent } from '../../src/mcp';
import { ProviderManager } from '../../src/session/provider-manager';
import type { ResolvedAgentProfile } from '../../src/profile';
import type { SDKSessionRPC } from '../../src/rpc';
import { Session } from '../../src/session';
import { SessionAPIImpl } from '../../src/session/rpc';
import { estimateTokensForMessages } from '../../src/utils/tokens';
import { createScriptedGenerate } from '../agent/harness/scripted-generate';
import { recordingTelemetry, type TelemetryRecord } from '../fixtures/telemetry';
import { executeTool } from '../tools/fixtures/execute-tool';
import { createFakeKaos, toolContentString } from '../tools/fixtures/fake-kaos';
const MOCK_PROVIDER = {
type: 'kimi',
apiKey: 'test-key',
model: 'mock-model',
} as const satisfies ProviderConfig;
const here = import.meta.dirname;
const mcpStdioFixture = join(here, '..', 'mcp', 'fixtures', 'mock-stdio-server.mjs');
const tempDirs: string[] = [];
afterEach(async () => {
for (const dir of tempDirs.splice(0)) {
await rm(dir, { recursive: true, force: true });
}
});
describe('Session.init', () => {
it('subscribes to MCP credential events before app-level registration', async () => {
const mcpOAuth = new McpOAuthService({ kimiHomeDir: await makeTempDir() });
const session = new Session({
id: 'test-mcp-credentials-during-init',
kaos: testKaos.withCwd(await makeTempDir()),
homedir: await makeTempDir(),
rpc: createSessionRpc([]),
providerManager: testProviderManager(),
mcpOAuthService: mcpOAuth,
});
const handle = vi
.spyOn(
session as unknown as {
handleMcpOAuthEvent(event: McpOAuthEvent): Promise<void>;
},
'handleMcpOAuthEvent',
)
.mockResolvedValue(undefined);
await mcpOAuth.invalidate('remote', 'https://mcp.example.test/service', 'tokens');
expect(handle).toHaveBeenCalledOnce();
await session.close();
await mcpOAuth.invalidate('remote', 'https://mcp.example.test/service', 'tokens');
expect(handle).toHaveBeenCalledOnce();
});
it('runs an isolated system-trigger turn and records the latest AGENTS as a system reminder', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
await mkdir(join(workDir, '.git'));
await writeFile(join(workDir, 'AGENTS.md'), 'latest project instructions', 'utf-8');
const events: Array<Record<string, unknown>> = [];
const scripted = createScriptedGenerate();
const session = new Session({
id: 'test-init',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc(events),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
const { agent: mainAgent } = await session.createAgent(
{ type: 'main', generate: scripted.generate },
{ profile: testProfile() },
);
mainAgent.config.update({
modelAlias: 'mock-model',
thinkingEffort: 'off',
});
mainAgent.tools.setActiveTools([]);
events.length = 0;
scripted.mockNextResponse({
type: 'text',
text: 'Explored the project structure, identified the build and test commands, mapped the module layout, and wrote a comprehensive summary into AGENTS.md covering architecture, conventions, and the developer workflow for future agents.',
});
await session.generateAgentsMd();
expect(session.agents.size).toBe(2);
expect(session.agents.get('main')).toBe(mainAgent);
expect(events).toContainEqual(
expect.objectContaining({
type: 'subagent.spawned',
agentId: 'main',
subagentId: 'agent-0',
subagentName: 'coder',
parentToolCallId: 'generate-agents-md',
}),
);
expect(events).toContainEqual(
expect.objectContaining({
type: 'turn.started',
agentId: 'agent-0',
origin: { kind: 'system_trigger', name: 'subagent' },
}),
);
expect(events).toContainEqual(
expect.objectContaining({
type: 'subagent.completed',
agentId: 'main',
subagentId: 'agent-0',
contextTokens: expect.any(Number),
}),
);
expect(scripted.calls[0]?.history).toMatchObject([
{
role: 'user',
content: [
expect.objectContaining({
text: expect.stringContaining('Task requirements:'),
}),
],
},
]);
const contextText = mainAgent.context.history
.flatMap((message) => message.content)
.map((part) => (part.type === 'text' ? part.text : ''))
.join('\n');
expect(contextText).toContain('<system-reminder>');
expect(contextText).toContain('Latest AGENTS.md file content:');
expect(contextText).toContain('latest project instructions');
expect(contextText).not.toContain('Task requirements:');
});
it('loads AGENTS.md via the persistence kaos when the tool kaos rejects readText (Zed ACP "Internal error" regression)', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
await mkdir(join(workDir, '.git'));
await writeFile(join(workDir, 'AGENTS.md'), 'project instructions from disk', 'utf-8');
// Simulate Zed's `fs/readTextFile` returning a generic -32603 Internal
// error: every `readText` through the tool kaos rejects. The persistence
// kaos is a real LocalKaos that can reach AGENTS.md on disk.
const toolKaos = wrapReadTextWithError(
testKaos.withCwd(workDir),
new Error('acp: readTextFile failed: Internal error'),
);
const capturedContext: { agentsMd: string | undefined } = { agentsMd: undefined };
const events: Array<Record<string, unknown>> = [];
const session = new Session({
id: 'test-bootstrap-acp-fallback',
kaos: toolKaos,
persistenceKaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc(events),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
try {
const { agent } = await session.createAgent(
{ type: 'main' },
{
profile: {
name: 'capture',
systemPrompt: (ctx) => {
capturedContext.agentsMd = ctx.agentsMd;
return '<system-prompt>';
},
tools: [],
},
},
);
expect(agent.config.systemPrompt).toBe('<system-prompt>');
expect(capturedContext.agentsMd).toContain('project instructions from disk');
} finally {
await session.close();
}
});
it('refreshes AGENTS.md from a resumed native session system prompt', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
await mkdir(join(workDir, '.git'));
await writeFile(join(workDir, 'AGENTS.md'), 'initial resume instructions', 'utf-8');
const firstSession = new Session({
id: 'test-resume-system-prompt-refresh',
kaos: testKaos.withCwd(workDir),
persistenceKaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
try {
const agent = await firstSession.createMain();
expect(agent.config.systemPrompt).toContain('initial resume instructions');
} finally {
await firstSession.closeForReload();
}
await writeFile(join(workDir, 'AGENTS.md'), 'updated resume instructions', 'utf-8');
const resumedSession = new Session({
id: 'test-resume-system-prompt-refresh',
kaos: testKaos.withCwd(workDir),
persistenceKaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
try {
await resumedSession.resume();
const resumedAgent = await resumedSession.ensureAgentResumed('main');
expect(resumedAgent.config.systemPrompt).toContain('initial resume instructions');
await resumedAgent.refreshSystemPrompt();
expect(resumedAgent.config.systemPrompt).toContain('updated resume instructions');
expect(resumedAgent.config.systemPrompt).not.toContain('initial resume instructions');
} finally {
await resumedSession.close();
}
});
it('resumes a v2 session from the standard path instead of persisted homedir', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const options = {
id: 'test-resume-without-agent-homedir',
kaos: testKaos.withCwd(workDir),
persistenceKaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
};
const firstSession = new Session(options);
try {
await firstSession.createMain();
} finally {
await firstSession.closeForReload();
}
const statePath = join(sessionDir, 'state.json');
const now = Date.now();
await writeFile(statePath, JSON.stringify({
id: options.id,
version: 2,
createdAt: now,
updatedAt: now,
archived: false,
cwd: workDir,
agents: { main: { homedir: '/stale/session/agents/main', type: 'main' } },
custom: {},
}), 'utf-8');
const resumedSession = new Session(options);
try {
await resumedSession.resume();
const main = await resumedSession.ensureAgentResumed('main');
expect(main.homedir).toBe(join(sessionDir, 'agents', 'main'));
expect(main.config.systemPrompt).not.toBe('');
} finally {
await resumedSession.close();
}
});
it('rebuilds builtin tools when rebinding the session tool kaos', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const staleKaos = createReadToolKaos(workDir, 'stale kaos\n');
const replacementKaos = createReadToolKaos(workDir, 'replacement kaos\n');
const session = new Session({
id: 'test-rebind-tool-kaos',
kaos: staleKaos,
persistenceKaos: testKaos.withCwd(sessionDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
try {
const { agent } = await session.createAgent({ type: 'main' }, { profile: testProfile() });
agent.config.update({
modelAlias: 'mock-model',
thinkingEffort: 'off',
});
agent.tools.initializeBuiltinTools();
agent.tools.setActiveTools(['Read']);
session.setToolKaos(replacementKaos);
const readTool = agent.tools.loopTools.find((candidate) => candidate.name === 'Read');
expect(readTool).toBeDefined();
const result = await executeTool(readTool!, {
args: { path: join(workDir, 'file.txt') },
turnId: '1',
toolCallId: 'call_read',
signal: new AbortController().signal,
});
expect(result.isError).not.toBe(true);
expect(toolContentString(result)).toContain('replacement kaos');
} finally {
await session.close();
}
});
it('tracks connected and failed MCP server totals after initial load', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const records: TelemetryRecord[] = [];
const session = new Session({
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
providerManager: testProviderManager(),
mcpConfig: {
servers: {
connected: {
transport: 'stdio',
command: process.execPath,
args: [mcpStdioFixture],
},
failed: {
transport: 'stdio',
command: '/this/path/does/not/exist/anywhere',
},
disabled: {
transport: 'stdio',
command: process.execPath,
args: [mcpStdioFixture],
enabled: false,
},
},
},
telemetry: recordingTelemetry(records),
});
try {
await session.mcp.waitForInitialLoad();
await expect(new SessionAPIImpl(session).getMcpStartupMetrics({})).resolves.toEqual({
durationMs: expect.any(Number),
});
expect(records).toContainEqual({
event: 'mcp_connected',
properties: {
server_count: 1,
total_count: 2,
},
});
expect(records).toContainEqual({
event: 'mcp_failed',
properties: {
failed_count: 1,
total_count: 2,
},
});
} finally {
await session.close();
}
}, 20000);
});
describe('AgentAPI.startBtw', () => {
it('runs a side subagent from a stable parent context snapshot without writing btw history', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const events: Array<Record<string, unknown>> = [];
const scripted = createScriptedGenerate();
const session = new Session({
id: 'test-btw',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc(events),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
const { agent: mainAgent } = await session.createAgent(
{ type: 'main', generate: scripted.generate },
{ profile: testProfile() },
);
mainAgent.config.update({
modelAlias: 'mock-model',
thinkingEffort: 'off',
});
mainAgent.tools.setActiveTools(['Read']);
registerLookupNoteTool(mainAgent);
mainAgent.context.appendUserMessage([{ type: 'text', text: 'Main task: implement /btw.' }]);
mainAgent.context.appendLoopEvent({
type: 'step.begin',
uuid: 'open-step',
turnId: 'main-turn',
step: 1,
});
mainAgent.context.appendLoopEvent({
type: 'tool.call',
uuid: 'open-call',
turnId: 'main-turn',
step: 1,
stepUuid: 'open-step',
toolCallId: 'call-open',
name: 'Read',
args: { path: 'src/main.ts' },
});
events.length = 0;
const summary = 'Main agent is implementing /btw.';
scripted.mockNextResponse({ type: 'text', text: summary });
try {
const api = new SessionAPIImpl(session);
const agentId = await api.startBtw({ agentId: 'main' });
expect(agentId).toBe('agent-0');
expect(scripted.calls).toHaveLength(0);
expect(session.metadata.agents[agentId]).toBeUndefined();
const childAgent = session.getReadyAgent(agentId);
if (childAgent === undefined) throw new Error('Expected /btw child agent');
const inheritedHistory = trimTrailingOpenToolExchange(
mainAgent.context.project(mainAgent.context.history),
);
expect(childAgent.context.history.slice(0, inheritedHistory.length)).toEqual(inheritedHistory);
expect(childAgent.context.tokenCount).toBe(0);
expect(childAgent.context.tokenCountWithPending).toBeGreaterThanOrEqual(
estimateTokensForMessages(inheritedHistory),
);
await api.prompt({
agentId,
input: [{ type: 'text', text: 'What are you working on right now?' }],
});
await vi.waitFor(() => {
expect(events).toContainEqual(
expect.objectContaining({
type: 'turn.ended',
agentId: 'agent-0',
reason: 'completed',
}),
);
});
expect(events.filter((event) => String(event['type']).startsWith('subagent.'))).toEqual([]);
expect(events).toContainEqual(
expect.objectContaining({
type: 'turn.started',
agentId: 'agent-0',
origin: { kind: 'user' },
}),
);
expect(scripted.calls).toHaveLength(1);
expect(scripted.calls[0]?.systemPrompt).toBe('<system-prompt>');
expect(scripted.calls[0]?.tools.map((tool) => tool.name)).toEqual([
'LookupNote',
'Read',
]);
const historyText = JSON.stringify(scripted.calls[0]?.history);
expect(historyText).toContain('Main task: implement /btw.');
expect(historyText).toContain('This is a side-channel conversation with the user.');
expect(historyText).toContain('All tool calls are disabled and will be rejected.');
expect(historyText).toContain('What are you working on right now?');
expect(historyText).not.toContain('call-open');
expect(JSON.stringify(mainAgent.context.history)).not.toContain(
'What are you working on right now?',
);
expect(JSON.stringify(session.getReadyAgent('agent-0')?.context.history)).toContain(
'What are you working on right now?',
);
scripted.mockNextResponse({ type: 'text', text: 'Follow-up answer from the same side agent.' });
await api.prompt({
agentId,
input: [{ type: 'text', text: 'Can you say that another way?' }],
});
await vi.waitFor(() => {
expect(scripted.calls).toHaveLength(2);
});
const followUpHistoryText = JSON.stringify(scripted.calls[1]?.history);
expect(followUpHistoryText).toContain('What are you working on right now?');
expect(followUpHistoryText).toContain('Can you say that another way?');
await expect(access(join(sessionDir, 'agents', 'agent-0', 'wire.jsonl'))).rejects.toThrow();
} finally {
await session.close();
}
});
it('declares parent tools but rejects side-question tool calls before a second text turn', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const events: Array<Record<string, unknown>> = [];
const scripted = createScriptedGenerate();
const session = new Session({
id: 'test-btw-deny-tools',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc(events),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
const { agent: mainAgent } = await session.createAgent(
{ type: 'main', generate: scripted.generate },
{ profile: testProfile() },
);
mainAgent.config.update({
modelAlias: 'mock-model',
thinkingEffort: 'off',
});
mainAgent.tools.setActiveTools(['Read']);
registerLookupNoteTool(mainAgent);
mainAgent.context.appendUserMessage([{ type: 'text', text: 'Main task context.' }]);
events.length = 0;
scripted.mockNextResponse(lookupNoteCall());
scripted.mockNextResponse({
type: 'text',
text: 'Main agent is implementing /btw based on the existing context.',
});
try {
const api = new SessionAPIImpl(session);
const agentId = await api.startBtw({ agentId: 'main' });
expect(agentId).toBe('agent-0');
await api.prompt({
agentId,
input: [{ type: 'text', text: 'What are you working on right now?' }],
});
await vi.waitFor(() => {
expect(events).toContainEqual(
expect.objectContaining({
type: 'turn.ended',
agentId: 'agent-0',
reason: 'completed',
}),
);
});
expect(events.filter((event) => String(event['type']).startsWith('subagent.'))).toEqual([]);
expect(scripted.calls).toHaveLength(2);
expect(scripted.calls[0]?.systemPrompt).toBe('<system-prompt>');
expect(scripted.calls[1]?.systemPrompt).toBe('<system-prompt>');
expect(scripted.calls[0]?.tools.map((tool) => tool.name)).toEqual([
'LookupNote',
'Read',
]);
expect(scripted.calls[1]?.tools.map((tool) => tool.name)).toEqual([
'LookupNote',
'Read',
]);
expect(JSON.stringify(scripted.calls[1]?.history)).toContain(
'Tool calls are disabled for side questions. Answer with text only.',
);
expect(events).toContainEqual(
expect.objectContaining({
type: 'tool.result',
agentId: 'agent-0',
toolCallId: 'call_lookup_note',
isError: true,
output: 'Tool calls are disabled for side questions. Answer with text only.',
}),
);
expect(JSON.stringify(mainAgent.context.history)).not.toContain(
'What are you working on right now?',
);
} finally {
await session.close();
}
});
it('cancels a btw turn through the returned agent id', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const events: Array<Record<string, unknown>> = [];
const generate: NonNullable<AgentOptions['generate']> = vi.fn(
async (_chat, _systemPrompt, _tools, _history, _callbacks, options) => {
const signal = options?.signal;
if (signal === undefined) {
throw new Error('Expected generate signal');
}
return new Promise<never>((_resolve, reject) => {
if (signal.aborted) {
reject(signal.reason);
return;
}
signal.addEventListener(
'abort',
() => {
reject(signal.reason);
},
{ once: true },
);
});
},
);
const session = new Session({
id: 'test-btw-cancel',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc(events),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
});
const { agent: mainAgent } = await session.createAgent(
{ type: 'main', generate },
{ profile: testProfile() },
);
mainAgent.config.update({
modelAlias: 'mock-model',
thinkingEffort: 'off',
});
events.length = 0;
try {
const api = new SessionAPIImpl(session);
const agentId = await api.startBtw({ agentId: 'main' });
expect(agentId).toBe('agent-0');
await api.prompt({
agentId,
input: [{ type: 'text', text: 'Where are things right now?' }],
});
await vi.waitFor(() => {
expect(events).toContainEqual(
expect.objectContaining({
type: 'turn.started',
agentId: 'agent-0',
origin: { kind: 'user' },
}),
);
});
await api.cancel({ agentId });
await vi.waitFor(() => {
expect(events).toContainEqual(
expect.objectContaining({
type: 'turn.ended',
agentId: 'agent-0',
reason: 'cancelled',
}),
);
});
expect(events.filter((event) => String(event['type']).startsWith('subagent.'))).toEqual([]);
} finally {
await session.close();
}
});
it('discovers sub-skills and builtins', async () => {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
const skillsRoot = join(workDir, 'skills');
await mkdir(join(skillsRoot, 'outer', 'inner'), { recursive: true });
await writeFile(
join(skillsRoot, 'outer', 'SKILL.md'),
[
'---',
'name: outer',
'description: Parent skill',
'has-sub-skill: true',
'---',
'',
'Outer body.',
].join('\n'),
);
await writeFile(
join(skillsRoot, 'outer', 'inner', 'SKILL.md'),
['---', 'name: inner', 'description: Nested skill', '---', '', 'Inner body.'].join('\n'),
);
const disabledSession = new Session({
id: 'test-disabled-sub-skills',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [skillsRoot] },
});
try {
const disabledSkills = await disabledSession.listSkills();
expect(disabledSkills.map((skill) => skill.name)).toContain('outer');
expect(disabledSkills.map((skill) => skill.name)).toContain('outer.inner');
expect(disabledSkills.map((skill) => skill.name)).toContain('sub-skill.consolidate');
} finally {
await disabledSession.close();
}
const enabledSession = new Session({
id: 'test-enabled-sub-skills',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [skillsRoot] },
});
try {
const enabledSkills = await enabledSession.listSkills();
expect(enabledSkills.map((skill) => skill.name)).toContain('outer');
expect(enabledSkills.map((skill) => skill.name)).toContain('outer.inner');
expect(enabledSkills.map((skill) => skill.name)).toContain('sub-skill.consolidate');
} finally {
await enabledSession.close();
}
});
});
describe('Session secondary-model live config', () => {
const SECONDARY_BASE_CONFIG: KimiConfig = {
providers: {
test: { type: MOCK_PROVIDER.type, apiKey: MOCK_PROVIDER.apiKey },
},
models: {
[MOCK_PROVIDER.model]: {
provider: 'test',
model: MOCK_PROVIDER.model,
maxContextSize: 1_000_000,
},
},
};
const SECONDARY_POINTER_CONFIG: KimiConfig = {
...SECONDARY_BASE_CONFIG,
secondaryModel: { model: MOCK_PROVIDER.model },
};
const SECONDARY_PATCHED_CONFIG: KimiConfig = {
...SECONDARY_BASE_CONFIG,
models: {
...SECONDARY_BASE_CONFIG.models,
__secondary__: {
...SECONDARY_BASE_CONFIG.models![MOCK_PROVIDER.model]!,
overrides: { defaultEffort: 'low' },
},
},
secondaryModel: { model: MOCK_PROVIDER.model, defaultEffort: 'low' },
};
async function makeSession(config?: KimiConfig): Promise<Session> {
const workDir = await makeTempDir();
const sessionDir = await makeTempDir();
return new Session({
id: 'test-secondary-model',
kaos: testKaos.withCwd(workDir),
homedir: sessionDir,
rpc: createSessionRpc([]),
skills: { explicitDirs: [join(workDir, 'missing-skills')] },
providerManager: testProviderManager(),
experimentalFlags: new FlagResolver({
KIMI_CODE_EXPERIMENTAL_SECONDARY_MODEL: '1',
}),
config,
});
}
it('live-applies the recipe to the session snapshot and live agents', async () => {
const session = await makeSession(SECONDARY_BASE_CONFIG);
try {
const { agent } = await session.createAgent(
{ type: 'main', generate: createScriptedGenerate().generate },
{ profile: testProfile() },
);
session.setSecondaryModelConfig(SECONDARY_POINTER_CONFIG);
expect(session.kimiConfig?.secondaryModel).toEqual({ model: MOCK_PROVIDER.model });
expect(agent.kimiConfig?.secondaryModel).toEqual({ model: MOCK_PROVIDER.model });
// A pointer-only recipe synthesizes no derived entry.
expect(session.kimiConfig?.models?.['__secondary__']).toBeUndefined();
// Agents created after the switch read the updated snapshot too.
const { agent: second } = await session.createAgent(
{ type: 'sub', generate: createScriptedGenerate().generate },
{ profile: testProfile(), parentAgentId: 'main' },
);
expect(second.kimiConfig?.secondaryModel).toEqual({ model: MOCK_PROVIDER.model });
} finally {
await session.close();
}
});
it('refreshes collaboration tool descriptions when live-applying a recipe', async () => {
const session = await makeSession(SECONDARY_BASE_CONFIG);
try {
const { agent } = await session.createAgent(
{ type: 'main', generate: createScriptedGenerate().generate },
{ profile: { ...testProfile(), tools: ['Agent', 'AgentSwarm'] } },
);
agent.config.update({ modelAlias: MOCK_PROVIDER.model, thinkingEffort: 'off' });
const descriptionsBefore = Object.fromEntries(
agent.tools.loopTools.map((tool) => [tool.name, tool.description]),
);
expect(descriptionsBefore['Agent']).not.toContain('Available models');
expect(descriptionsBefore['AgentSwarm']).not.toContain('Available models');
session.setSecondaryModelConfig(SECONDARY_POINTER_CONFIG);
const descriptionsAfter = Object.fromEntries(
agent.tools.loopTools.map((tool) => [tool.name, tool.description]),
);
expect(descriptionsAfter['Agent']).toContain('- secondary: mock-model');
expect(descriptionsAfter['AgentSwarm']).toContain('- secondary: mock-model');
} finally {
await session.close();
}
});
it('replaces a patched runtime snapshot with a pointer-only runtime snapshot', async () => {
const session = await makeSession(SECONDARY_BASE_CONFIG);
try {
session.setSecondaryModelConfig(SECONDARY_PATCHED_CONFIG);
const derived = session.kimiConfig?.models?.['__secondary__'];
expect(derived).toBeDefined();
expect(derived?.overrides?.defaultEffort).toBe('low');
session.setSecondaryModelConfig(SECONDARY_POINTER_CONFIG);
expect(session.kimiConfig?.models?.['__secondary__']).toBeUndefined();
} finally {
await session.close();
}
});
it('keeps unrelated session settings when applying a secondary-model config', async () => {
const session = await makeSession({
...SECONDARY_BASE_CONFIG,
loopControl: { maxStepsPerTurn: 7 },
});
try {
session.setSecondaryModelConfig({
...SECONDARY_POINTER_CONFIG,
loopControl: { maxStepsPerTurn: 99 },
});
expect(session.kimiConfig?.loopControl?.maxStepsPerTurn).toBe(7);
} finally {
await session.close();
}
});
it('rejects a dangling pointer with the wrapped secondary-model error', async () => {
const session = await makeSession(SECONDARY_BASE_CONFIG);
try {
expect(() =>
session.setSecondaryModelConfig({
...SECONDARY_BASE_CONFIG,
secondaryModel: { model: 'missing-model' },
}),
).toThrow(/\[secondary_model\]\.model/);
expect(session.kimiConfig?.secondaryModel).toBeUndefined();
} finally {
await session.close();
}
});
it('rejects when the complete config has no persisted secondary recipe', async () => {
const session = await makeSession(SECONDARY_BASE_CONFIG);
try {
expect(() => session.setSecondaryModelConfig(SECONDARY_BASE_CONFIG)).toThrow(/persist/);
} finally {
await session.close();
}
});
it('rejects when the session has no config', async () => {
const session = await makeSession();
try {
expect(() => session.setSecondaryModelConfig(SECONDARY_POINTER_CONFIG)).toThrow(/no config/);
} finally {
await session.close();
}
});
});
async function makeTempDir(): Promise<string> {
const dir = await mkdtemp(join(tmpdir(), 'kimi-core-init-'));
tempDirs.push(dir);
return dir;
}
function testProviderManager(): ProviderManager {
return new ProviderManager({
config: {
providers: {
test: {
type: MOCK_PROVIDER.type,
apiKey: MOCK_PROVIDER.apiKey,
},
},
models: {
[MOCK_PROVIDER.model]: {
provider: 'test',
model: MOCK_PROVIDER.model,
maxContextSize: 1_000_000,
},
},
},
});
}
function testProfile(): ResolvedAgentProfile {
return {
name: 'test',
systemPrompt: () => '<system-prompt>',
tools: [],
};
}
function createReadToolKaos(cwd: string, content: string): Kaos {
return createFakeKaos({
getcwd: () => cwd,
stat: async () =>
({
stMode: 0o100644,
stIno: 1,
stDev: 1,
stNlink: 1,
stUid: 0,
stGid: 0,
stSize: content.length,
stAtime: 0,
stMtime: 0,
stCtime: 0,
}) satisfies StatResult,
readBytes: async () => Buffer.from(content),
readLines: async function* () {
yield content;
},
});
}
function registerLookupNoteTool(agent: Agent): void {
agent.tools.registerUserTool({
name: 'LookupNote',
description: 'Look up a note from the host application.',
parameters: {
type: 'object',
properties: {
query: { type: 'string' },
},
required: ['query'],
additionalProperties: false,
},
});
}
function lookupNoteCall(): ToolCall {
return {
type: 'function',
id: 'call_lookup_note',
name: 'LookupNote',
arguments: JSON.stringify({ query: 'status' }),
};
}
function createSessionRpc(events: Array<Record<string, unknown>>): SDKSessionRPC {
return {
emitEvent: vi.fn(async (event) => {
events.push(event);
}),
requestApproval: vi.fn(async () => ({ decision: 'cancelled' })),
requestQuestion: vi.fn(async () => null),
toolCall: vi.fn(async () => ({
output: 'custom tools are not supported in this test',
isError: true,
})),
} as SDKSessionRPC;
}
/**
* Wrap a {@link Kaos} so every `readText` (and `readLines`, which reads via
* `readText` in the ACP bridge) rejects with `cause`. Used to simulate the
* Zed ACP `fs/readTextFile` "Internal error" path that broke session bootstrap
* before AGENTS.md loading was rerouted onto the persistence kaos.
*/
function wrapReadTextWithError(inner: Kaos, cause: Error): Kaos {
return new Proxy(inner, {
get(target, prop, receiver) {
if (prop === 'readText') {
return async () => {
throw cause;
};
}
if (prop === 'readLines') {
return async function* () {
yield* [];
throw cause;
};
}
if (prop === 'withCwd') {
return (cwd: string) => wrapReadTextWithError(target.withCwd(cwd), cause);
}
return Reflect.get(target, prop, receiver);
},
});
}