server: refresh model catalog daily
This commit is contained in:
@@ -5,7 +5,7 @@ import swaggerUI from "@fastify/swagger-ui";
|
||||
import sensible from "@fastify/sensible";
|
||||
import { env } from "./env.js";
|
||||
import { ensureDatabaseReady } from "./db-init.js";
|
||||
import { warmModelCatalog } from "./llm/model-catalog.js";
|
||||
import { startModelCatalogRefreshLoop, warmModelCatalog } from "./llm/model-catalog.js";
|
||||
import { registerRoutes } from "./routes.js";
|
||||
|
||||
const app = Fastify({
|
||||
@@ -21,6 +21,7 @@ const app = Fastify({
|
||||
|
||||
await ensureDatabaseReady(app.log);
|
||||
await warmModelCatalog(app.log);
|
||||
const stopModelCatalogRefreshLoop = startModelCatalogRefreshLoop(app.log);
|
||||
|
||||
await app.register(cors, {
|
||||
origin: true,
|
||||
@@ -80,6 +81,10 @@ app.setErrorHandler((err, req, reply) => {
|
||||
});
|
||||
});
|
||||
|
||||
app.addHook("onClose", async () => {
|
||||
stopModelCatalogRefreshLoop();
|
||||
});
|
||||
|
||||
await registerRoutes(app);
|
||||
|
||||
await app.listen({ port: env.PORT, host: env.HOST });
|
||||
|
||||
@@ -13,6 +13,7 @@ export type ModelCatalogSnapshot = Partial<Record<Provider, ProviderModelSnapsho
|
||||
|
||||
const baseProviders: Provider[] = ["openai", "anthropic", "xai"];
|
||||
const MODEL_FETCH_TIMEOUT_MS = 15000;
|
||||
const MODEL_CATALOG_REFRESH_INTERVAL_MS = 24 * 60 * 60 * 1000;
|
||||
|
||||
const modelCatalog: ModelCatalogSnapshot = {
|
||||
openai: { models: [], loadedAt: null, error: null },
|
||||
@@ -20,6 +21,8 @@ const modelCatalog: ModelCatalogSnapshot = {
|
||||
xai: { models: [], loadedAt: null, error: null },
|
||||
};
|
||||
|
||||
let catalogRefreshPromise: Promise<void> | null = null;
|
||||
|
||||
function getCatalogProviders(): Provider[] {
|
||||
return isHermesAgentConfigured() ? [...baseProviders, "hermes-agent"] : baseProviders;
|
||||
}
|
||||
@@ -86,17 +89,42 @@ async function refreshProviderModels(provider: Provider, logger?: FastifyBaseLog
|
||||
logger?.info({ provider, modelCount: models.length }, "model catalog loaded");
|
||||
} catch (err: any) {
|
||||
const message = err?.message ?? String(err);
|
||||
const previous = modelCatalog[provider];
|
||||
const fallbackModels = provider === "hermes-agent" && env.HERMES_AGENT_MODEL ? [env.HERMES_AGENT_MODEL] : [];
|
||||
modelCatalog[provider] = {
|
||||
models: provider === "hermes-agent" && env.HERMES_AGENT_MODEL ? [env.HERMES_AGENT_MODEL] : [],
|
||||
loadedAt: new Date().toISOString(),
|
||||
models: previous?.models.length ? previous.models : fallbackModels,
|
||||
loadedAt: previous?.loadedAt ?? null,
|
||||
error: message,
|
||||
};
|
||||
logger?.warn({ provider, err: message }, "failed to load provider model catalog");
|
||||
}
|
||||
}
|
||||
|
||||
export async function refreshModelCatalog(logger?: FastifyBaseLogger) {
|
||||
if (catalogRefreshPromise) return catalogRefreshPromise;
|
||||
|
||||
catalogRefreshPromise = Promise.all(getCatalogProviders().map((provider) => refreshProviderModels(provider, logger)))
|
||||
.then(() => undefined)
|
||||
.finally(() => {
|
||||
catalogRefreshPromise = null;
|
||||
});
|
||||
|
||||
return catalogRefreshPromise;
|
||||
}
|
||||
|
||||
export async function warmModelCatalog(logger?: FastifyBaseLogger) {
|
||||
await Promise.all(getCatalogProviders().map((provider) => refreshProviderModels(provider, logger)));
|
||||
await refreshModelCatalog(logger);
|
||||
}
|
||||
|
||||
export function startModelCatalogRefreshLoop(logger?: FastifyBaseLogger) {
|
||||
const timer = setInterval(() => {
|
||||
void refreshModelCatalog(logger);
|
||||
}, MODEL_CATALOG_REFRESH_INTERVAL_MS);
|
||||
timer.unref?.();
|
||||
|
||||
return () => {
|
||||
clearInterval(timer);
|
||||
};
|
||||
}
|
||||
|
||||
export function getModelCatalogSnapshot(): ModelCatalogSnapshot {
|
||||
|
||||
Reference in New Issue
Block a user