fix(venice): retry model discovery on transient fetch failures

This commit is contained in:
Peter Steinberger
2026-03-03 03:20:46 +00:00
parent 85a320de54
commit 17bb87f432
3 changed files with 225 additions and 10 deletions

View File

@@ -0,0 +1,110 @@
import { afterEach, describe, expect, it, vi } from "vitest";
import {
buildVeniceModelDefinition,
discoverVeniceModels,
VENICE_MODEL_CATALOG,
} from "./venice-models.js";
const ORIGINAL_NODE_ENV = process.env.NODE_ENV;
const ORIGINAL_VITEST = process.env.VITEST;
function restoreDiscoveryEnv(): void {
if (ORIGINAL_NODE_ENV === undefined) {
delete process.env.NODE_ENV;
} else {
process.env.NODE_ENV = ORIGINAL_NODE_ENV;
}
if (ORIGINAL_VITEST === undefined) {
delete process.env.VITEST;
} else {
process.env.VITEST = ORIGINAL_VITEST;
}
}
async function runWithDiscoveryEnabled<T>(operation: () => Promise<T>): Promise<T> {
process.env.NODE_ENV = "development";
delete process.env.VITEST;
try {
return await operation();
} finally {
restoreDiscoveryEnv();
}
}
function makeModelsResponse(id: string): Response {
return new Response(
JSON.stringify({
data: [
{
id,
model_spec: {
name: id,
privacy: "private",
availableContextTokens: 131072,
capabilities: {
supportsReasoning: false,
supportsVision: false,
supportsFunctionCalling: true,
},
},
},
],
}),
{
status: 200,
headers: { "Content-Type": "application/json" },
},
);
}
describe("venice-models", () => {
afterEach(() => {
vi.unstubAllGlobals();
restoreDiscoveryEnv();
});
it("buildVeniceModelDefinition returns config with required fields", () => {
const entry = VENICE_MODEL_CATALOG[0];
const def = buildVeniceModelDefinition(entry);
expect(def.id).toBe(entry.id);
expect(def.name).toBe(entry.name);
expect(def.reasoning).toBe(entry.reasoning);
expect(def.input).toEqual(entry.input);
expect(def.cost).toEqual({ input: 0, output: 0, cacheRead: 0, cacheWrite: 0 });
expect(def.contextWindow).toBe(entry.contextWindow);
expect(def.maxTokens).toBe(entry.maxTokens);
});
it("retries transient fetch failures before succeeding", async () => {
let attempts = 0;
const fetchMock = vi.fn(async () => {
attempts += 1;
if (attempts < 3) {
throw Object.assign(new TypeError("fetch failed"), {
cause: { code: "ECONNRESET", message: "socket hang up" },
});
}
return makeModelsResponse("llama-3.3-70b");
});
vi.stubGlobal("fetch", fetchMock as unknown as typeof fetch);
const models = await runWithDiscoveryEnabled(() => discoverVeniceModels());
expect(attempts).toBe(3);
expect(models.map((m) => m.id)).toContain("llama-3.3-70b");
});
it("falls back to static catalog after retry budget is exhausted", async () => {
const fetchMock = vi.fn(async () => {
throw Object.assign(new TypeError("fetch failed"), {
cause: { code: "ENOTFOUND", message: "getaddrinfo ENOTFOUND api.venice.ai" },
});
});
vi.stubGlobal("fetch", fetchMock as unknown as typeof fetch);
const models = await runWithDiscoveryEnabled(() => discoverVeniceModels());
expect(fetchMock).toHaveBeenCalledTimes(3);
expect(models).toHaveLength(VENICE_MODEL_CATALOG.length);
expect(models.map((m) => m.id)).toEqual(VENICE_MODEL_CATALOG.map((m) => m.id));
});
});

View File

@@ -1,4 +1,5 @@
import type { ModelDefinitionConfig } from "../config/types.js";
import { retryAsync } from "../infra/retry.js";
import { createSubsystemLogger } from "../logging/subsystem.js";
const log = createSubsystemLogger("venice-models");
@@ -16,6 +17,24 @@ export const VENICE_DEFAULT_COST = {
cacheWrite: 0,
};
const VENICE_DISCOVERY_TIMEOUT_MS = 10_000;
const VENICE_DISCOVERY_RETRYABLE_HTTP_STATUS = new Set([408, 425, 429, 500, 502, 503, 504]);
const VENICE_DISCOVERY_RETRYABLE_NETWORK_CODES = new Set([
"ECONNABORTED",
"ECONNREFUSED",
"ECONNRESET",
"EAI_AGAIN",
"ENETDOWN",
"ENETUNREACH",
"ENOTFOUND",
"ETIMEDOUT",
"UND_ERR_BODY_TIMEOUT",
"UND_ERR_CONNECT_TIMEOUT",
"UND_ERR_CONNECT_ERROR",
"UND_ERR_HEADERS_TIMEOUT",
"UND_ERR_SOCKET",
]);
/**
* Complete catalog of Venice AI models.
*
@@ -332,6 +351,67 @@ interface VeniceModelsResponse {
data: VeniceModel[];
}
class VeniceDiscoveryHttpError extends Error {
readonly status: number;
constructor(status: number) {
super(`HTTP ${status}`);
this.name = "VeniceDiscoveryHttpError";
this.status = status;
}
}
function staticVeniceModelDefinitions(): ModelDefinitionConfig[] {
return VENICE_MODEL_CATALOG.map(buildVeniceModelDefinition);
}
function hasRetryableNetworkCode(err: unknown): boolean {
const queue: unknown[] = [err];
const seen = new Set<unknown>();
while (queue.length > 0) {
const current = queue.shift();
if (!current || typeof current !== "object" || seen.has(current)) {
continue;
}
seen.add(current);
const candidate = current as {
cause?: unknown;
errors?: unknown;
code?: unknown;
errno?: unknown;
};
const code =
typeof candidate.code === "string"
? candidate.code
: typeof candidate.errno === "string"
? candidate.errno
: undefined;
if (code && VENICE_DISCOVERY_RETRYABLE_NETWORK_CODES.has(code)) {
return true;
}
if (candidate.cause) {
queue.push(candidate.cause);
}
if (Array.isArray(candidate.errors)) {
queue.push(...candidate.errors);
}
}
return false;
}
function isRetryableVeniceDiscoveryError(err: unknown): boolean {
if (err instanceof VeniceDiscoveryHttpError) {
return true;
}
if (err instanceof Error && err.name === "AbortError") {
return true;
}
if (err instanceof TypeError && err.message.toLowerCase() === "fetch failed") {
return true;
}
return hasRetryableNetworkCode(err);
}
/**
* Discover models from Venice API with fallback to static catalog.
* The /models endpoint is public and doesn't require authentication.
@@ -339,23 +419,45 @@ interface VeniceModelsResponse {
export async function discoverVeniceModels(): Promise<ModelDefinitionConfig[]> {
// Skip API discovery in test environment
if (process.env.NODE_ENV === "test" || process.env.VITEST) {
return VENICE_MODEL_CATALOG.map(buildVeniceModelDefinition);
return staticVeniceModelDefinitions();
}
try {
const response = await fetch(`${VENICE_BASE_URL}/models`, {
signal: AbortSignal.timeout(5000),
});
const response = await retryAsync(
async () => {
const currentResponse = await fetch(`${VENICE_BASE_URL}/models`, {
signal: AbortSignal.timeout(VENICE_DISCOVERY_TIMEOUT_MS),
headers: {
Accept: "application/json",
},
});
if (
!currentResponse.ok &&
VENICE_DISCOVERY_RETRYABLE_HTTP_STATUS.has(currentResponse.status)
) {
throw new VeniceDiscoveryHttpError(currentResponse.status);
}
return currentResponse;
},
{
attempts: 3,
minDelayMs: 300,
maxDelayMs: 2000,
jitter: 0.2,
label: "venice-model-discovery",
shouldRetry: isRetryableVeniceDiscoveryError,
},
);
if (!response.ok) {
log.warn(`Failed to discover models: HTTP ${response.status}, using static catalog`);
return VENICE_MODEL_CATALOG.map(buildVeniceModelDefinition);
return staticVeniceModelDefinitions();
}
const data = (await response.json()) as VeniceModelsResponse;
if (!Array.isArray(data.data) || data.data.length === 0) {
log.warn("No models found from API, using static catalog");
return VENICE_MODEL_CATALOG.map(buildVeniceModelDefinition);
return staticVeniceModelDefinitions();
}
// Merge discovered models with catalog metadata
@@ -395,9 +497,13 @@ export async function discoverVeniceModels(): Promise<ModelDefinitionConfig[]> {
}
}
return models.length > 0 ? models : VENICE_MODEL_CATALOG.map(buildVeniceModelDefinition);
return models.length > 0 ? models : staticVeniceModelDefinitions();
} catch (error) {
if (error instanceof VeniceDiscoveryHttpError) {
log.warn(`Failed to discover models: HTTP ${error.status}, using static catalog`);
return staticVeniceModelDefinitions();
}
log.warn(`Discovery failed: ${String(error)}, using static catalog`);
return VENICE_MODEL_CATALOG.map(buildVeniceModelDefinition);
return staticVeniceModelDefinitions();
}
}

View File

@@ -1,8 +1,7 @@
import { createRequire } from "node:module";
import { describe, expect, it } from "vitest";
const require = createRequire(import.meta.url);
const { detectChangedScope } = require("../../scripts/ci-changed-scope.mjs") as {
const { detectChangedScope } = (await import("../../scripts/ci-changed-scope.mjs")) as unknown as {
detectChangedScope: (paths: string[]) => {
runNode: boolean;
runMacos: boolean;