fix(cli): resolve embedding provider explicitly and surface lane status in sl search (#192)

* feat(cli): add tryUseManagedLocalEmbeddingsDaemon for read-only callers

* feat(cli): add resolveProjectEmbeddingProvider helper

* fix(cli): wire sl search through resolveProjectEmbeddingProvider so semantic lane works

* fix(cli): wire wiki/knowledge search through resolveProjectEmbeddingProvider

* feat(cli): surface embeddings-unavailable status when sl search returns empty

* refactor(cli): route admin reindex through resolveProjectEmbeddingProvider

* refactor: pass embeddingProvider into ingest/scan instead of resolving inside @ktx/context

* refactor(mcp): resolve embedding provider in CLI factory, pass into context ports

* refactor(context): delete MANAGED_SENTENCE_TRANSFORMERS_BASE_URL sentinel

* refactor(cli): delete sentinel-based managed-embeddings indirection

* chore: scrub stale managed-embeddings sentinel references from tests and smoke script

* chore: unexport unused EmbeddingResolutionMode alias

* fix(cli): force pathPrefix="" when targeting the managed embeddings daemon

The managed daemon serves /embeddings/compute directly. The default
pathPrefix in @ktx/llm is /api, so omitting sentenceTransformers from
ktx.yaml produced /api/embeddings/compute -> 404. The resolver now
sets pathPrefix='' explicitly when wiring the managed daemon URL,
matching what the daemon actually exposes.
This commit is contained in:
Andrey Avtomonov 2026-05-21 02:21:22 +02:00 committed by GitHub
parent 56a967278a
commit 9d92c79988
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
36 changed files with 750 additions and 442 deletions

View file

@ -1,12 +1,12 @@
import { describe, expect, it, vi } from 'vitest';
import { MANAGED_SENTENCE_TRANSFORMERS_BASE_URL } from '@ktx/context';
import {
ensureManagedLocalEmbeddingsDaemon,
managedLocalEmbeddingHealthConfig,
managedLocalEmbeddingProjectConfig,
tryUseManagedLocalEmbeddingsDaemon,
} from './managed-local-embeddings.js';
import type { ManagedPythonCommandRuntime } from './managed-python-command.js';
import type { ManagedPythonDaemonStartResult } from './managed-python-daemon.js';
import type { ManagedPythonDaemonLayout } from './managed-python-runtime.js';
function makeIo() {
let stdout = '';
@ -94,25 +94,6 @@ function daemonResult(status: 'started' | 'reused' = 'reused'): ManagedPythonDae
};
}
describe('managedLocalEmbeddingProjectConfig', () => {
it('uses a stable managed runtime marker instead of a random daemon port', () => {
expect(
managedLocalEmbeddingProjectConfig({
model: 'all-MiniLM-L6-v2',
dimensions: 384,
}),
).toEqual({
backend: 'sentence-transformers',
model: 'all-MiniLM-L6-v2',
dimensions: 384,
sentenceTransformers: {
base_url: MANAGED_SENTENCE_TRANSFORMERS_BASE_URL,
pathPrefix: '',
},
});
});
});
describe('managedLocalEmbeddingHealthConfig', () => {
it('uses the active KTX daemon URL for the immediate health check', () => {
expect(
@ -181,3 +162,93 @@ describe('ensureManagedLocalEmbeddingsDaemon', () => {
expect(io.stderr()).toContain('Using KTX daemon: http://127.0.0.1:61234');
});
});
describe('tryUseManagedLocalEmbeddingsDaemon', () => {
it('returns the daemon when one is running and healthy', async () => {
const readStatus = vi.fn(async () => ({
kind: 'running' as const,
detail: 'ok',
layout: {} as ManagedPythonDaemonLayout,
state: {
schemaVersion: 1 as const,
pid: 123,
host: '127.0.0.1' as const,
port: 4321,
version: '0.5.0',
features: ['local-embeddings' as const],
startedAt: '2026-05-21T00:00:00Z',
stdoutLog: '/tmp/stdout.log',
stderrLog: '/tmp/stderr.log',
},
baseUrl: 'http://127.0.0.1:4321',
}));
const result = await tryUseManagedLocalEmbeddingsDaemon({
cliVersion: '0.5.0',
projectDir: '/work/proj',
readStatus,
});
expect(result).toEqual({
baseUrl: 'http://127.0.0.1:4321',
stdoutLog: '/tmp/stdout.log',
stderrLog: '/tmp/stderr.log',
});
expect(readStatus).toHaveBeenCalledWith({
cliVersion: '0.5.0',
projectDir: '/work/proj',
});
});
it('returns null when no daemon state exists', async () => {
const readStatus = vi.fn(async () => ({
kind: 'stopped' as const,
detail: 'no state',
layout: {} as ManagedPythonDaemonLayout,
}));
const result = await tryUseManagedLocalEmbeddingsDaemon({
cliVersion: '0.5.0',
projectDir: '/work/proj',
readStatus,
});
expect(result).toBeNull();
});
it('returns null when daemon is stale', async () => {
const readStatus = vi.fn(async () => ({
kind: 'stale' as const,
detail: 'process gone',
layout: {} as ManagedPythonDaemonLayout,
}));
const result = await tryUseManagedLocalEmbeddingsDaemon({
cliVersion: '0.5.0',
projectDir: '/work/proj',
readStatus,
});
expect(result).toBeNull();
});
it('rejects daemons that do not advertise local-embeddings', async () => {
const readStatus = vi.fn(async () => ({
kind: 'running' as const,
detail: 'ok',
layout: {} as ManagedPythonDaemonLayout,
state: {
schemaVersion: 1 as const,
pid: 123,
host: '127.0.0.1' as const,
port: 4321,
version: '0.5.0',
features: ['core' as const],
startedAt: '2026-05-21T00:00:00Z',
stdoutLog: '/tmp/stdout.log',
stderrLog: '/tmp/stderr.log',
},
baseUrl: 'http://127.0.0.1:4321',
}));
const result = await tryUseManagedLocalEmbeddingsDaemon({
cliVersion: '0.5.0',
projectDir: '/work/proj',
readStatus,
});
expect(result).toBeNull();
});
});