ktx/packages/cli/src/embedding-resolution.ts
Andrey Avtomonov 9d92c79988
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.
2026-05-21 02:21:22 +02:00

107 lines
4.2 KiB
TypeScript

import {
type KtxEmbeddingProvider,
createKtxEmbeddingProvider as defaultCreateKtxEmbeddingProvider,
} from '@ktx/llm';
import type { KtxLocalProject, KtxProjectEmbeddingConfig } from '@ktx/context/project';
import { resolveLocalKtxEmbeddingConfig } from '@ktx/context';
import type { KtxCliIo } from './cli-runtime.js';
import {
ensureManagedLocalEmbeddingsDaemon as defaultEnsureManagedDaemon,
tryUseManagedLocalEmbeddingsDaemon as defaultTryUseManagedDaemon,
} from './managed-local-embeddings.js';
import type { KtxManagedPythonInstallPolicy } from './managed-python-command.js';
type EmbeddingResolutionMode = 'ensure' | 'use-if-running';
export type EmbeddingProviderResolution =
| { kind: 'disabled' }
| { kind: 'configured'; provider: KtxEmbeddingProvider; baseUrl: string }
| { kind: 'managed-running'; provider: KtxEmbeddingProvider; baseUrl: string }
| { kind: 'managed-started'; provider: KtxEmbeddingProvider; baseUrl: string }
| { kind: 'managed-unavailable'; reason: string };
export interface ResolveProjectEmbeddingProviderOptions {
mode: EmbeddingResolutionMode;
cliVersion: string;
io: KtxCliIo;
/** Required when mode === 'ensure'. */
installPolicy?: KtxManagedPythonInstallPolicy;
tryUseManagedDaemon?: typeof defaultTryUseManagedDaemon;
ensureManagedDaemon?: typeof defaultEnsureManagedDaemon;
createKtxEmbeddingProvider?: typeof defaultCreateKtxEmbeddingProvider;
}
function usesManagedDaemon(embeddings: KtxProjectEmbeddingConfig): boolean {
if (embeddings.backend !== 'sentence-transformers') {
return false;
}
const baseUrl = embeddings.sentenceTransformers?.base_url;
return baseUrl === undefined || baseUrl === '';
}
export async function resolveProjectEmbeddingProvider(
project: KtxLocalProject,
options: ResolveProjectEmbeddingProviderOptions,
): Promise<EmbeddingProviderResolution> {
const embeddings = project.config.ingest.embeddings;
if (embeddings.backend === 'none') {
return { kind: 'disabled' };
}
const createProvider = options.createKtxEmbeddingProvider ?? defaultCreateKtxEmbeddingProvider;
if (!usesManagedDaemon(embeddings)) {
const resolved = resolveLocalKtxEmbeddingConfig(embeddings, process.env);
if (!resolved) {
return { kind: 'managed-unavailable', reason: 'embedding config missing required fields' };
}
const provider = createProvider(resolved);
const baseUrl = embeddings.sentenceTransformers?.base_url ?? '';
return { kind: 'configured', provider, baseUrl };
}
const tryUse = options.tryUseManagedDaemon ?? defaultTryUseManagedDaemon;
const running = await tryUse({ cliVersion: options.cliVersion, projectDir: project.projectDir });
if (running) {
const provider = buildManagedProvider(embeddings, running.baseUrl, createProvider);
return provider
? { kind: 'managed-running', provider, baseUrl: running.baseUrl }
: { kind: 'managed-unavailable', reason: 'failed to build embedding provider from running daemon' };
}
if (options.mode === 'use-if-running') {
return { kind: 'managed-unavailable', reason: 'managed embeddings daemon is not running' };
}
const ensure = options.ensureManagedDaemon ?? defaultEnsureManagedDaemon;
if (!options.installPolicy) {
throw new Error("installPolicy is required when mode === 'ensure'");
}
const daemon = await ensure({
cliVersion: options.cliVersion,
projectDir: project.projectDir,
installPolicy: options.installPolicy,
io: options.io,
});
const provider = buildManagedProvider(embeddings, daemon.baseUrl, createProvider);
return provider
? { kind: 'managed-started', provider, baseUrl: daemon.baseUrl }
: { kind: 'managed-unavailable', reason: 'failed to build embedding provider after starting daemon' };
}
function buildManagedProvider(
embeddings: KtxProjectEmbeddingConfig,
baseUrl: string,
createProvider: typeof defaultCreateKtxEmbeddingProvider,
): KtxEmbeddingProvider | null {
const merged: KtxProjectEmbeddingConfig = {
...embeddings,
sentenceTransformers: {
...embeddings.sentenceTransformers,
base_url: baseUrl,
pathPrefix: '',
},
};
const resolved = resolveLocalKtxEmbeddingConfig(merged, process.env);
return resolved ? createProvider(resolved) : null;
}