Skip to content

#1 Persistent memory for AI coding agents based on real-world benchmarks

fix(config): apply the premium-model cost warning to OPENAI_MODEL (#1501)

Nanami45b556e56497
diff --git a/src/config.ts b/src/config.ts
index 34205d98..649917f9 100644
--- a/src/config.ts
+++ b/src/config.ts
@@ -90,14 +90,40 @@ export function hydrateProcessEnvFromFile(): void {
}
}
+// Warn once when a configured model is in the premium tier and likely to burn
+// money on background compression. Captured workload data shows ~$5/35h on
+// claude-sonnet-4 vs ~$0.46/35h on deepseek-v4-pro for the same compression
+// mix. Heuristic match avoids hard-coding a pricing table.
+function warnIfPremiumModel(env: Record<string, string>, envName: string, model: string): void {
+ if (
+ warnPremiumModelShown ||
+ !/sonnet|opus|gpt-5\.\d+-sol|gpt-4o(?!.*mini)|gpt-4-turbo/i.test(model) ||
+ env["AGENTMEMORY_SUPPRESS_COST_WARNING"] === "1" ||
+ env["AGENTMEMORY_SUPPRESS_COST_WARNING"] === "true"
+ ) {
+ return;
+ }
+ warnPremiumModelShown = true;
+ process.stderr.write(
+ `[agentmemory] ${envName}=${model} is in the premium tier. ` +
+ `Background compression on this model can cost $5+/day under active use. ` +
+ `Cheaper alternatives with comparable quality for memory compression: ` +
+ `deepseek/deepseek-v4-flash-0731, deepseek/deepseek-v4-pro, qwen/qwen3-coder. ` +
+ `See README "Cost-aware model selection" for the full table. ` +
+ `Set AGENTMEMORY_SUPPRESS_COST_WARNING=1 to silence.\n`,
+ );
+}
+
function detectProvider(env: Record<string, string>): ProviderConfig {
const maxTokens = parseInt(env["MAX_TOKENS"] || "4096", 10);
// OpenAI-compatible: supports OpenAI, DeepSeek, SiliconFlow, Azure, vLLM, LM Studio
if (hasRealValue(env["OPENAI_API_KEY"]) && env["OPENAI_API_KEY_FOR_LLM"] !== "false") {
+ const model = env["OPENAI_MODEL"] || "gpt-5.6-luna";
+ warnIfPremiumModel(env, "OPENAI_MODEL", model);
return {
provider: "openai",
- model: env["OPENAI_MODEL"] || "gpt-5.6-luna",
+ model,
maxTokens,
baseURL: env["OPENAI_BASE_URL"],
};
@@ -135,27 +161,7 @@ function detectProvider(env: Record<string, string>): ProviderConfig {
}
if (hasRealValue(env["OPENROUTER_API_KEY"])) {
const model = env["OPENROUTER_MODEL"] || "anthropic/claude-sonnet-5";
- // warn when the configured OpenRouter model is in the
- // premium tier and likely to burn money on background compression.
- // Captured workload data shows ~$5/35h on claude-sonnet-4 vs
- // ~$0.46/35h on deepseek-v4-pro for the same compression mix.
- // Heuristic match avoids hard-coding a pricing table.
- if (
- !warnPremiumModelShown &&
- /sonnet|opus|gpt-5\.\d+-sol|gpt-4o(?!.*mini)|gpt-4-turbo/i.test(model) &&
- env["AGENTMEMORY_SUPPRESS_COST_WARNING"] !== "1" &&
- env["AGENTMEMORY_SUPPRESS_COST_WARNING"] !== "true"
- ) {
- warnPremiumModelShown = true;
- process.stderr.write(
- `[agentmemory] OPENROUTER_MODEL=${model} is in the premium tier. ` +
- `Background compression on this model can cost $5+/day under active use. ` +
- `Cheaper alternatives with comparable quality for memory compression: ` +
- `deepseek/deepseek-v4-flash-0731, deepseek/deepseek-v4-pro, qwen/qwen3-coder. ` +
- `See README "Cost-aware model selection" for the full table. ` +
- `Set AGENTMEMORY_SUPPRESS_COST_WARNING=1 to silence.\n`,
- );
- }
+ warnIfPremiumModel(env, "OPENROUTER_MODEL", model);
return {
provider: "openrouter",
model,
diff --git a/test/premium-model-warning.test.ts b/test/premium-model-warning.test.ts
new file mode 100644
index 00000000..2e09e6a6
--- /dev/null
+++ b/test/premium-model-warning.test.ts
@@ -0,0 +1,67 @@
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+
+const KEYS = [
+ "OPENAI_API_KEY",
+ "OPENAI_MODEL",
+ "OPENAI_BASE_URL",
+ "OPENROUTER_API_KEY",
+ "OPENROUTER_MODEL",
+ "OPENAI_API_KEY_FOR_LLM",
+ "AGENTMEMORY_SUPPRESS_COST_WARNING",
+];
+
+async function warningsFor(env: Record<string, string>, loads = 1): Promise<string[]> {
+ for (const [k, v] of Object.entries(env)) process.env[k] = v;
+ const writes: string[] = [];
+ vi.spyOn(process.stderr, "write").mockImplementation((chunk: string | Uint8Array) => {
+ writes.push(String(chunk));
+ return true;
+ });
+ vi.resetModules();
+ const { loadConfig } = await import("../src/config.js");
+ for (let i = 0; i < loads; i++) loadConfig();
+ return writes.filter((w) => w.includes("premium tier"));
+}
+
+describe("premium-model cost warning", () => {
+ beforeEach(() => {
+ for (const k of KEYS) delete process.env[k];
+ });
+ afterEach(() => {
+ for (const k of KEYS) delete process.env[k];
+ vi.restoreAllMocks();
+ });
+
+ it("warns once for a premium OPENAI_MODEL, naming that variable", async () => {
+ const warnings = await warningsFor(
+ { OPENAI_API_KEY: "k", OPENAI_BASE_URL: "https://openrouter.ai/api/v1", OPENAI_MODEL: "anthropic/claude-sonnet-5" },
+ 2,
+ );
+ expect(warnings).toHaveLength(1);
+ expect(warnings[0]).toContain("OPENAI_MODEL=anthropic/claude-sonnet-5 is in the premium tier");
+ });
+
+ it("stays quiet for the default and for cheap OpenAI-compatible models", async () => {
+ expect(await warningsFor({ OPENAI_API_KEY: "k" })).toHaveLength(0);
+ expect(await warningsFor({ OPENAI_API_KEY: "k", OPENAI_MODEL: "deepseek/deepseek-v4.1-flash:floor" })).toHaveLength(0);
+ });
+
+ it("is silenced by AGENTMEMORY_SUPPRESS_COST_WARNING", async () => {
+ const warnings = await warningsFor({
+ OPENAI_API_KEY: "k",
+ OPENAI_MODEL: "gpt-4o",
+ AGENTMEMORY_SUPPRESS_COST_WARNING: "1",
+ });
+ expect(warnings).toHaveLength(0);
+ });
+
+ it("keeps the OpenRouter warning text", async () => {
+ const warnings = await warningsFor({
+ OPENROUTER_API_KEY: "k",
+ OPENROUTER_MODEL: "anthropic/claude-opus-5",
+ OPENAI_API_KEY_FOR_LLM: "false",
+ });
+ expect(warnings).toHaveLength(1);
+ expect(warnings[0]).toContain("OPENROUTER_MODEL=anthropic/claude-opus-5 is in the premium tier");
+ });
+});