feat(cluster,api,models,cli): add cluster-aware model catalog deployments and request routing

This commit is contained in:
2026-04-20 23:00:50 +00:00
parent 83cacd0cf1
commit 4f2266e1b7
55 changed files with 3970 additions and 1630 deletions
+11 -8
View File
@@ -21,7 +21,7 @@ export class DriverManager {
constructor() {
this.gpuDetector = new GpuDetector();
this.drivers = new Map([
this.drivers = new Map<TGpuVendor, BaseDriver>([
['nvidia', new NvidiaDriver()],
['amd', new AmdDriver()],
['intel', new IntelDriver()],
@@ -197,10 +197,15 @@ export class DriverManager {
// Print status for each vendor
for (const [vendor, gpuList] of vendorGpus) {
if (vendor === 'unknown') {
logger.logBox('Unknown GPUs', [
`${gpuList.length} GPU(s) with unknown vendor`,
'Manual driver installation may be required',
], 50, 'warning');
logger.logBox(
'Unknown GPUs',
[
`${gpuList.length} GPU(s) with unknown vendor`,
'Manual driver installation may be required',
],
50,
'warning',
);
continue;
}
@@ -219,9 +224,7 @@ export class DriverManager {
const args: string[] = [];
// Filter to specific GPUs if provided
const targetGpus = gpuIds
? gpus.filter((g) => gpuIds.includes(g.id))
: gpus;
const targetGpus = gpuIds ? gpus.filter((g) => gpuIds.includes(g.id)) : gpus;
if (targetGpus.length === 0) {
return args;