Enforce free plan at generation level (#2531)

* Enforce free plan at generation level

* fix lint

front/lib/api/assistant/generation.ts

@@ -13,9 +13,11 @@ import {
   getSupportedModelConfig,
   GPT_4_32K_MODEL_ID,
   GPT_4_MODEL_CONFIG,
+  isLargeModel,
 } from "@app/lib/assistant";
 import { Authenticator } from "@app/lib/auth";
 import { CoreAPI } from "@app/lib/core_api";
+import { FREE_TEST_PLAN_CODE } from "@app/lib/plans/plan_codes";
 import { redisClient } from "@app/lib/redis";
 import { Err, Ok, Result } from "@app/lib/result";
 import logger from "@app/logger/logger";
@@ -311,7 +313,8 @@ export async function* runGeneration(
   void
 > {
   const owner = auth.workspace();
-  if (!owner) {
+  const plan = auth.plan();
+  if (!owner || !plan) {
     throw new Error("Unexpected unauthenticated call to `runGeneration`");
   }
 
@@ -334,6 +337,20 @@ export async function* runGeneration(
 
   let model = c.model;
 
+  if (isLargeModel(model) && plan.code === FREE_TEST_PLAN_CODE) {
+    yield {
+      type: "generation_error",
+      created: Date.now(),
+      configurationId: configuration.sId,
+      messageId: agentMessage.sId,
+      error: {
+        code: "free_plan_error",
+        message: `Free plan does not support large models. Please upgrade to a paid plan to use this model.`,
+      },
+    };
+    return;
+  }
+
   const contextSize = getSupportedModelConfig(c.model).contextSize;
 
   const MIN_GENERATION_TOKENS = 2048;

front/lib/assistant.ts

@@ -15,6 +15,7 @@ export const GPT_4_32K_MODEL_CONFIG = {
   displayName: "GPT 4",
   contextSize: 32768,
   recommendedTopK: 32,
+  largeModel: true,
 } as const;
 
 export const GPT_4_MODEL_CONFIG = {
@@ -23,6 +24,7 @@ export const GPT_4_MODEL_CONFIG = {
   displayName: "GPT 4",
   contextSize: 8192,
   recommendedTopK: 16,
+  largeModel: true,
 };
 
 export const GPT_4_TURBO_MODEL_CONFIG = {
@@ -31,6 +33,7 @@ export const GPT_4_TURBO_MODEL_CONFIG = {
   displayName: "GPT 4",
   contextSize: 128000,
   recommendedTopK: 32,
+  largeModel: true,
 } as const;
 
 export const GPT_3_5_TURBO_16K_MODEL_CONFIG = {
@@ -39,6 +42,7 @@ export const GPT_3_5_TURBO_16K_MODEL_CONFIG = {
   displayName: "GPT 3.5 Turbo",
   contextSize: 16384,
   recommendedTopK: 16,
+  largeModel: false,
 } as const;
 
 export const GPT_3_5_TURBO_MODEL_CONFIG = {
@@ -47,6 +51,7 @@ export const GPT_3_5_TURBO_MODEL_CONFIG = {
   displayName: "GPT 3.5 Turbo",
   contextSize: 4096,
   recommendedTopK: 16,
+  largeModel: false,
 } as const;
 
 export const CLAUDE_DEFAULT_MODEL_CONFIG = {
@@ -55,6 +60,7 @@ export const CLAUDE_DEFAULT_MODEL_CONFIG = {
   displayName: "Claude 2",
   contextSize: 100000,
   recommendedTopK: 32,
+  largeModel: true,
 } as const;
 
 export const CLAUDE_INSTANT_DEFAULT_MODEL_CONFIG = {
@@ -63,6 +69,7 @@ export const CLAUDE_INSTANT_DEFAULT_MODEL_CONFIG = {
   displayName: "Claude Instant 1.2",
   contextSize: 100000,
   recommendedTopK: 32,
+  largeModel: false,
 } as const;
 
 export const MISTRAL_7B_DEFAULT_MODEL_CONFIG = {
@@ -71,6 +78,7 @@ export const MISTRAL_7B_DEFAULT_MODEL_CONFIG = {
   displayName: "Mistral 7B",
   contextSize: 8192,
   recommendedTopK: 16,
+  largeModel: false,
 } as const;
 
 export const SUPPORTED_MODEL_CONFIGS = [
@@ -100,6 +108,19 @@ export function isSupportedModel(model: unknown): model is SupportedModel {
   );
 }
 
+export function isLargeModel(model: unknown): model is SupportedModel {
+  const maybeSupportedModel = model as SupportedModel;
+  const m = SUPPORTED_MODEL_CONFIGS.find(
+    (m) =>
+      m.modelId === maybeSupportedModel.modelId &&
+      m.providerId === maybeSupportedModel.providerId
+  );
+  if (m) {
+    return m.largeModel;
+  }
+  return false;
+}
+
 export function getSupportedModelConfig(supportedModel: SupportedModel) {
   // here it is safe to cast the result to non-nullable because SupportedModel
   // is derived from the const array of configs above