From 076cb844fc0b13e2797dfd309cdb756bdca6dc66 Mon Sep 17 00:00:00 2001
From: timothycarambat <rambat1010@gmail.com>
Date: Tue, 13 May 2025 16:23:23 -0700
Subject: [PATCH 1/5] Add context window finder from litellm maintained list
 apply to all cloud providers, have client cache for 3 days

---
 server/storage/models/.gitignore              |   1 +
 server/utils/AiProviders/anthropic/index.js   |   4 +-
 server/utils/AiProviders/cohere/index.js      |   4 +-
 server/utils/AiProviders/deepseek/index.js    |   6 +-
 .../utils/AiProviders/gemini/defaultModels.js |  15 +-
 server/utils/AiProviders/gemini/index.js      |  12 +-
 .../AiProviders/gemini/syncStaticLists.mjs    |  31 ----
 server/utils/AiProviders/groq/index.js        |   4 +-
 server/utils/AiProviders/modelMap/index.js    | 140 ++++++++++++++++++
 .../{modelMap.js => modelMap/legacy.js}       |  13 +-
 server/utils/AiProviders/openAi/index.js      |   9 +-
 server/utils/AiProviders/xai/index.js         |   6 +-
 12 files changed, 177 insertions(+), 68 deletions(-)
 create mode 100644 server/utils/AiProviders/modelMap/index.js
 rename server/utils/AiProviders/{modelMap.js => modelMap/legacy.js} (90%)

diff --git a/server/storage/models/.gitignore b/server/storage/models/.gitignore
index f5c78ac877c..5e83df7bcc4 100644
--- a/server/storage/models/.gitignore
+++ b/server/storage/models/.gitignore
@@ -9,3 +9,4 @@ gemini
 togetherAi
 tesseract
 ppio
+context-windows/*
\ No newline at end of file
diff --git a/server/utils/AiProviders/anthropic/index.js b/server/utils/AiProviders/anthropic/index.js
index c4f53acf639..450b376b3e0 100644
--- a/server/utils/AiProviders/anthropic/index.js
+++ b/server/utils/AiProviders/anthropic/index.js
@@ -45,11 +45,11 @@ class AnthropicLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.anthropic[modelName] ?? 100_000;
+    return MODEL_MAP.get("anthropic", modelName) ?? 100_000;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.anthropic[this.model] ?? 100_000;
+    return MODEL_MAP.get("anthropic", this.model) ?? 100_000;
   }
 
   isValidChatCompletionModel(_modelName = "") {
diff --git a/server/utils/AiProviders/cohere/index.js b/server/utils/AiProviders/cohere/index.js
index 33b65df21c0..a6647f5b658 100644
--- a/server/utils/AiProviders/cohere/index.js
+++ b/server/utils/AiProviders/cohere/index.js
@@ -63,11 +63,11 @@ class CohereLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.cohere[modelName] ?? 4_096;
+    return MODEL_MAP.get("cohere", modelName) ?? 4_096;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.cohere[this.model] ?? 4_096;
+    return MODEL_MAP.get("cohere", this.model) ?? 4_096;
   }
 
   async isValidChatCompletionModel(model = "") {
diff --git a/server/utils/AiProviders/deepseek/index.js b/server/utils/AiProviders/deepseek/index.js
index b91332a84a1..ddb2de317c7 100644
--- a/server/utils/AiProviders/deepseek/index.js
+++ b/server/utils/AiProviders/deepseek/index.js
@@ -29,7 +29,7 @@ class DeepSeekLLM {
 
     this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
-    this.log("Initialized with model:", this.model);
+    this.log(`Initialized ${this.model} with context window ${this.promptWindowLimit()}`);
   }
 
   log(text, ...args) {
@@ -53,11 +53,11 @@ class DeepSeekLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.deepseek[modelName] ?? 8192;
+    return MODEL_MAP.get("deepseek", modelName) ?? 8192;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.deepseek[this.model] ?? 8192;
+    return MODEL_MAP.get("deepseek", this.model) ?? 8192;
   }
 
   async isValidChatCompletionModel(modelName = "") {
diff --git a/server/utils/AiProviders/gemini/defaultModels.js b/server/utils/AiProviders/gemini/defaultModels.js
index fde63341d2a..4a52dc99c65 100644
--- a/server/utils/AiProviders/gemini/defaultModels.js
+++ b/server/utils/AiProviders/gemini/defaultModels.js
@@ -1,7 +1,7 @@
 const { MODEL_MAP } = require("../modelMap");
 
 const stableModels = [
-  // %STABLE_MODELS% - updated 2025-04-07T20:29:49.276Z
+  // %STABLE_MODELS% - updated 2025-05-13T23:13:58.920Z
   "gemini-1.5-pro-001",
   "gemini-1.5-pro-002",
   "gemini-1.5-pro",
@@ -14,6 +14,7 @@ const stableModels = [
   "gemini-2.0-flash-001",
   "gemini-2.0-flash-lite-001",
   "gemini-2.0-flash-lite",
+  "gemini-2.0-flash-preview-image-generation",
   // %EOC_STABLE_MODELS%
 ];
 
@@ -22,7 +23,7 @@ const stableModels = [
 // generally, v1beta models have `exp` in the name, but not always
 // so we check for both against a static list as well via API.
 const v1BetaModels = [
-  // %V1BETA_MODELS% - updated 2025-04-07T20:29:49.276Z
+  // %V1BETA_MODELS% - updated 2025-05-13T23:13:58.920Z
   "gemini-1.5-pro-latest",
   "gemini-1.5-flash-latest",
   "gemini-1.5-flash-8b-latest",
@@ -30,6 +31,9 @@ const v1BetaModels = [
   "gemini-1.5-flash-8b-exp-0924",
   "gemini-2.5-pro-exp-03-25",
   "gemini-2.5-pro-preview-03-25",
+  "gemini-2.5-flash-preview-04-17",
+  "gemini-2.5-flash-preview-04-17-thinking",
+  "gemini-2.5-pro-preview-05-06",
   "gemini-2.0-flash-exp",
   "gemini-2.0-flash-exp-image-generation",
   "gemini-2.0-flash-lite-preview-02-05",
@@ -41,6 +45,7 @@ const v1BetaModels = [
   "gemini-2.0-flash-thinking-exp",
   "gemini-2.0-flash-thinking-exp-1219",
   "learnlm-1.5-pro-experimental",
+  "learnlm-2.0-flash-experimental",
   "gemma-3-1b-it",
   "gemma-3-4b-it",
   "gemma-3-12b-it",
@@ -48,17 +53,17 @@ const v1BetaModels = [
   // %EOC_V1BETA_MODELS%
 ];
 
-const defaultGeminiModels = [
+const defaultGeminiModels = () => [
   ...stableModels.map((model) => ({
     id: model,
     name: model,
-    contextWindow: MODEL_MAP.gemini[model],
+    contextWindow: MODEL_MAP.get("gemini", model),
     experimental: false,
   })),
   ...v1BetaModels.map((model) => ({
     id: model,
     name: model,
-    contextWindow: MODEL_MAP.gemini[model],
+    contextWindow: MODEL_MAP.get("gemini", model),
     experimental: true,
   })),
 ];
diff --git a/server/utils/AiProviders/gemini/index.js b/server/utils/AiProviders/gemini/index.js
index bd2268e5174..393f0cd5227 100644
--- a/server/utils/AiProviders/gemini/index.js
+++ b/server/utils/AiProviders/gemini/index.js
@@ -107,7 +107,7 @@ class GeminiLLM {
     try {
       const cacheModelPath = path.resolve(cacheFolder, "models.json");
       if (!fs.existsSync(cacheModelPath))
-        return MODEL_MAP.gemini[modelName] ?? 30_720;
+        return MODEL_MAP.get("gemini", modelName) ?? 30_720;
 
       const models = safeJsonParse(fs.readFileSync(cacheModelPath));
       const model = models.find((model) => model.id === modelName);
@@ -118,15 +118,13 @@ class GeminiLLM {
       return model.contextWindow;
     } catch (e) {
       console.error(`GeminiLLM:promptWindowLimit`, e.message);
-      return MODEL_MAP.gemini[modelName] ?? 30_720;
+      return MODEL_MAP.get("gemini", modelName) ?? 30_720;
     }
   }
 
   promptWindowLimit() {
     try {
-      if (!fs.existsSync(this.cacheModelPath))
-        return MODEL_MAP.gemini[this.model] ?? 30_720;
-
+      if (!fs.existsSync(this.cacheModelPath)) return MODEL_MAP.get("gemini", this.model) ?? 30_720;
       const models = safeJsonParse(fs.readFileSync(this.cacheModelPath));
       const model = models.find((model) => model.id === this.model);
       if (!model)
@@ -136,7 +134,7 @@ class GeminiLLM {
       return model.contextWindow;
     } catch (e) {
       console.error(`GeminiLLM:promptWindowLimit`, e.message);
-      return MODEL_MAP.gemini[this.model] ?? 30_720;
+      return MODEL_MAP.get("gemini", this.model) ?? 30_720;
     }
   }
 
@@ -281,7 +279,7 @@ class GeminiLLM {
 
     if (allModels.length === 0) {
       console.error(`Gemini:getGeminiModels - No models found`);
-      return defaultGeminiModels;
+      return defaultGeminiModels();
     }
 
     console.log(
diff --git a/server/utils/AiProviders/gemini/syncStaticLists.mjs b/server/utils/AiProviders/gemini/syncStaticLists.mjs
index 42382fae388..b276995acdf 100644
--- a/server/utils/AiProviders/gemini/syncStaticLists.mjs
+++ b/server/utils/AiProviders/gemini/syncStaticLists.mjs
@@ -9,7 +9,6 @@
 import fs from "fs";
 import path from "path";
 import dotenv from "dotenv";
-import { MODEL_MAP } from "../modelMap.js";
 
 dotenv.config({ path: `../../../.env.development` });
 const existingCachePath = path.resolve('../../../storage/models/gemini')
@@ -46,34 +45,4 @@ function updateDefaultModelsFile(models) {
   fs.writeFileSync(path.join("./defaultModels.js"), defaultModelFileContents);
   console.log("Updated defaultModels.js. Dont forget to `yarn lint` and commit!");
 }
-
-function updateModelMap(models) {
-  const existingModelMap = MODEL_MAP;
-  console.log('Updating modelMap.js `gemini` object...')
-  console.log(`Removed existing gemini object (${Object.keys(existingModelMap.gemini).length} models) from modelMap.js`);
-  existingModelMap.gemini = {};
-
-  for (const model of models) existingModelMap.gemini[model.id] = model.contextWindow;
-  console.log(`Updated modelMap.js 'gemini' object with ${Object.keys(existingModelMap.gemini).length} models from API`);
-
-  // Update the modelMap.js file
-  const contents = `/**
- * The model name and context window for all know model windows
- * that are available through providers which has discrete model options.
- * This file is automatically generated by syncStaticLists.mjs
- * and should not be edited manually.
- * 
- * Last updated: ${new Date().toISOString()}
- */
-const MODEL_MAP = {
-  ${Object.entries(existingModelMap).map(([key, value]) => `${key}: ${JSON.stringify(value, null, 2)}`).join(',\n')}
-};
-
-module.exports = { MODEL_MAP };
-`;
-  fs.writeFileSync(path.resolve("../modelMap.js"), contents);
-  console.log('Updated modelMap.js `gemini` object. Dont forget to `yarn lint` and commit!');
-}
-
 updateDefaultModelsFile(models);
-updateModelMap(models);
diff --git a/server/utils/AiProviders/groq/index.js b/server/utils/AiProviders/groq/index.js
index 9e7e77fa16b..8eddefab0c5 100644
--- a/server/utils/AiProviders/groq/index.js
+++ b/server/utils/AiProviders/groq/index.js
@@ -49,11 +49,11 @@ class GroqLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.groq[modelName] ?? 8192;
+    return MODEL_MAP.get("groq", modelName) ?? 8192;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.groq[this.model] ?? 8192;
+    return MODEL_MAP.get("groq", this.model) ?? 8192;
   }
 
   async isValidChatCompletionModel(modelName = "") {
diff --git a/server/utils/AiProviders/modelMap/index.js b/server/utils/AiProviders/modelMap/index.js
new file mode 100644
index 00000000000..dbf0aff6a60
--- /dev/null
+++ b/server/utils/AiProviders/modelMap/index.js
@@ -0,0 +1,140 @@
+const path = require("path");
+const fs = require("fs");
+const LEGACY_MODEL_MAP = require("./legacy");
+
+class ContextWindowFinder {
+  static instance = null;
+  static modelMap = LEGACY_MODEL_MAP;
+
+  /**
+   * Mapping for AnythingLLM provider <> LiteLLM provider
+   * @type {Record<string, string>}
+   */
+  static trackedProviders = {
+    anthropic: "anthropic",
+    openai: "openai",
+    cohere: "cohere_chat",
+    gemini: "vertex_ai-language-models",
+    groq: "groq",
+    xai: "xai",
+    deepseek: "deepseek",
+  };
+  static expiryMs = 1000 * 60 * 60 * 24 * 3; // 3 days
+  static remoteUrl =
+    "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json";
+
+  cacheLocation = path.resolve(
+    process.env.STORAGE_DIR
+      ? path.resolve(process.env.STORAGE_DIR, "models", "context-windows")
+      : path.resolve(__dirname, `../../../storage/models/context-windows`)
+  );
+  cacheFilePath = path.resolve(this.cacheLocation, "context-windows.json");
+  cacheFileExpiryPath = path.resolve(this.cacheLocation, ".cached_at");
+
+  constructor() {
+    if (ContextWindowFinder.instance) return ContextWindowFinder.instance;
+    ContextWindowFinder.instance = this;
+    if (!fs.existsSync(this.cacheLocation))
+      fs.mkdirSync(this.cacheLocation, { recursive: true });
+    this.#pullRemoteModelMap();
+  }
+
+  log(text, ...args) {
+    console.log(`\x1b[33m[ContextWindowFinder]\x1b[0m ${text}`, ...args);
+  }
+
+  /**
+   * Checks if the cache is stale by checking if the cache file exists and if the cache file is older than the expiry time.
+   * @returns {boolean}
+   */
+  get isCacheStale() {
+    if (!fs.existsSync(this.cacheFileExpiryPath)) return true;
+    const cachedAt = fs.readFileSync(this.cacheFileExpiryPath, "utf8");
+    return Date.now() - cachedAt > ContextWindowFinder.expiryMs;
+  }
+
+  get cache() {
+    if (!fs.existsSync(this.cacheFileExpiryPath)) return null;
+    if (!this.isCacheStale)
+      return JSON.parse(
+        fs.readFileSync(this.cacheFilePath, { encoding: "utf8" })
+      );
+    return null;
+  }
+
+  /**
+   * Pulls the remote model map from the remote URL, formats it and caches it.
+   * @returns {Record<string, Record<string, number>>} - The formatted model map
+   */
+  async #pullRemoteModelMap() {
+    const remoteContexWindowMap = await fetch(ContextWindowFinder.remoteUrl)
+      .then((res) => res.json())
+      .then((data) => {
+        fs.writeFileSync(this.cacheFilePath, JSON.stringify(data, null, 2));
+        fs.writeFileSync(this.cacheFileExpiryPath, Date.now().toString());
+        this.log("Remote model map synced and cached");
+        return data;
+      })
+      .catch((error) => {
+        this.log("Error syncing remote model map", error);
+        return null;
+      });
+    if (!remoteContexWindowMap) return null;
+
+    const modelMap = this.#formatModelMap(remoteContexWindowMap);
+    fs.writeFileSync(this.cacheFilePath, JSON.stringify(modelMap, null, 2));
+    fs.writeFileSync(this.cacheFileExpiryPath, Date.now().toString());
+    return modelMap;
+  }
+
+  /**
+   * Formats the remote model map to a format that is compatible with how we store the model map
+   * for all providers who use it.
+   * @param {Record<string, any>} modelMap - The remote model map
+   * @returns {Record<string, Record<string, number>>} - The formatted model map
+   */
+  #formatModelMap(modelMap = {}) {
+    const formattedModelMap = {};
+
+    for (const [provider, liteLLMProviderTag] of Object.entries(
+      ContextWindowFinder.trackedProviders
+    )) {
+      formattedModelMap[provider] = {};
+      const matches = Object.entries(modelMap).filter(
+        ([_key, config]) => config.litellm_provider === liteLLMProviderTag
+      );
+      for (const [key, config] of matches) {
+        const contextWindow = Number(config.max_input_tokens);
+        if (isNaN(contextWindow)) continue;
+
+        // Some models have a provider/model-tag format, so we need to get the last part since we dont do paths
+        // for names with the exception of some router-providers like OpenRouter or Together.
+        const modelName = key.split("/").pop();
+        formattedModelMap[provider][modelName] = contextWindow;
+      }
+    }
+    return formattedModelMap;
+  }
+
+  /**
+   * Gets the context window for a given provider and model.
+   * @param {string} provider - The provider to get the context window for
+   * @param {string} model - The model to get the context window for
+   * @returns {number} - The context window for the given provider and model
+   */
+  get(provider = null, model = null) {
+    if (!provider || !this.cache || !this.cache[provider]) return null;
+    if (!model) return this.cache[provider];
+    const modelContextWindow = this.cache[provider][model];
+    if (!modelContextWindow) {
+      this.log("Invalid access to model context window - not found in cache", {
+        provider,
+        model,
+      });
+      return null;
+    }
+    return Number(modelContextWindow);
+  }
+}
+
+module.exports = { MODEL_MAP: new ContextWindowFinder() };
diff --git a/server/utils/AiProviders/modelMap.js b/server/utils/AiProviders/modelMap/legacy.js
similarity index 90%
rename from server/utils/AiProviders/modelMap.js
rename to server/utils/AiProviders/modelMap/legacy.js
index a80400c30b4..2faf99dc237 100644
--- a/server/utils/AiProviders/modelMap.js
+++ b/server/utils/AiProviders/modelMap/legacy.js
@@ -1,12 +1,4 @@
-/**
- * The model name and context window for all know model windows
- * that are available through providers which has discrete model options.
- * This file is automatically generated by syncStaticLists.mjs
- * and should not be edited manually.
- *
- * Last updated: 2025-04-07T20:29:49.277Z
- */
-const MODEL_MAP = {
+const LEGACY_MODEL_MAP = {
   anthropic: {
     "claude-instant-1.2": 100000,
     "claude-2.0": 100000,
@@ -117,5 +109,4 @@ const MODEL_MAP = {
     "grok-beta": 131072,
   },
 };
-
-module.exports = { MODEL_MAP };
+module.exports = LEGACY_MODEL_MAP;
diff --git a/server/utils/AiProviders/openAi/index.js b/server/utils/AiProviders/openAi/index.js
index 2e5996c939a..586532ce8dd 100644
--- a/server/utils/AiProviders/openAi/index.js
+++ b/server/utils/AiProviders/openAi/index.js
@@ -25,6 +25,11 @@ class OpenAiLLM {
 
     this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
+    this.log(`Initialized ${this.model} with context window ${this.promptWindowLimit()}`);
+  }
+
+  log(text, ...args) {
+    console.log(`\x1b[36m[${this.constructor.name}]\x1b[0m ${text}`, ...args);
   }
 
   /**
@@ -54,11 +59,11 @@ class OpenAiLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.openai[modelName] ?? 4_096;
+    return MODEL_MAP.get("openai", modelName) ?? 4_096;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.openai[this.model] ?? 4_096;
+    return MODEL_MAP.get("openai", this.model) ?? 4_096;
   }
 
   // Short circuit if name has 'gpt' since we now fetch models from OpenAI API
diff --git a/server/utils/AiProviders/xai/index.js b/server/utils/AiProviders/xai/index.js
index 2319e72206e..ad0a7e5eecf 100644
--- a/server/utils/AiProviders/xai/index.js
+++ b/server/utils/AiProviders/xai/index.js
@@ -28,7 +28,7 @@ class XAiLLM {
 
     this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
-    this.log("Initialized with model:", this.model);
+    this.log(`Initialized ${this.model} with context window ${this.promptWindowLimit()}`);
   }
 
   log(text, ...args) {
@@ -52,11 +52,11 @@ class XAiLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.xai[modelName] ?? 131_072;
+    return MODEL_MAP.get("xai", modelName) ?? 131_072;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.xai[this.model] ?? 131_072;
+    return MODEL_MAP.get("xai", this.model) ?? 131_072;
   }
 
   isValidChatCompletionModel(_modelName = "") {

From 36722ee9fd3dac8c41c651a8f4fd78af310e10e8 Mon Sep 17 00:00:00 2001
From: timothycarambat <rambat1010@gmail.com>
Date: Wed, 14 May 2025 10:27:53 -0700
Subject: [PATCH 2/5] docker container bootup warning

---
 .github/workflows/dev-build.yaml |  2 +-
 docker/docker-entrypoint.sh      | 15 ++++++++++++++-
 2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/dev-build.yaml b/.github/workflows/dev-build.yaml
index e4ff3689ce1..0e2e15c33b6 100644
--- a/.github/workflows/dev-build.yaml
+++ b/.github/workflows/dev-build.yaml
@@ -6,7 +6,7 @@ concurrency:
 
 on:
   push:
-    branches: ['873-pgvector-support'] # put your current branch to create a build. Core team only.
+    branches: ['warn-bad-docker-command'] # put your current branch to create a build. Core team only.
     paths-ignore:
       - '**.md'
       - 'cloud-deployments/*'
diff --git a/docker/docker-entrypoint.sh b/docker/docker-entrypoint.sh
index 1ac69e5baf4..dc2faa59173 100755
--- a/docker/docker-entrypoint.sh
+++ b/docker/docker-entrypoint.sh
@@ -1,4 +1,17 @@
 #!/bin/bash
+
+# Check if STORAGE_DIR is set
+if [ -z "$STORAGE_DIR" ]; then
+    echo "================================================================"
+    echo "⚠️  ⚠️  ⚠️  WARNING: STORAGE_DIR environment variable is not set! ⚠️  ⚠️  ⚠️"
+    echo "Not setting this will result in data loss on container restart since"
+    echo "the application will not have a persistent storage location."
+    echo "Please run the container with the official docker command at"
+    echo "https://docs.anythingllm.com/installation-docker/quickstart"
+    echo "⚠️  ⚠️  ⚠️  WARNING: STORAGE_DIR environment variable is not set! ⚠️  ⚠️  ⚠️"
+    echo "================================================================"
+fi
+
 {
   cd /app/server/ &&
     npx prisma generate --schema=./prisma/schema.prisma &&
@@ -7,4 +20,4 @@
 } &
 { node /app/collector/index.js; } &
 wait -n
-exit $?
+exit $?
\ No newline at end of file

From 5006031a2a34be5cf6a8d8f0ce24ed60ed0c824b Mon Sep 17 00:00:00 2001
From: timothycarambat <rambat1010@gmail.com>
Date: Wed, 14 May 2025 10:40:05 -0700
Subject: [PATCH 3/5] update invalid ENV warning

---
 docker/docker-entrypoint.sh | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/docker/docker-entrypoint.sh b/docker/docker-entrypoint.sh
index dc2faa59173..9c47de25c4e 100755
--- a/docker/docker-entrypoint.sh
+++ b/docker/docker-entrypoint.sh
@@ -6,6 +6,8 @@ if [ -z "$STORAGE_DIR" ]; then
     echo "⚠️  ⚠️  ⚠️  WARNING: STORAGE_DIR environment variable is not set! ⚠️  ⚠️  ⚠️"
     echo "Not setting this will result in data loss on container restart since"
     echo "the application will not have a persistent storage location."
+    echo "It can also result in weird errors in various parts of the application."
+    echo "\n\n"
     echo "Please run the container with the official docker command at"
     echo "https://docs.anythingllm.com/installation-docker/quickstart"
     echo "⚠️  ⚠️  ⚠️  WARNING: STORAGE_DIR environment variable is not set! ⚠️  ⚠️  ⚠️"

From da3fd0990f8bdd1d17744db48de843459f630bb6 Mon Sep 17 00:00:00 2001
From: timothycarambat <rambat1010@gmail.com>
Date: Wed, 14 May 2025 10:43:13 -0700
Subject: [PATCH 4/5] rebased with current HEAD

---
 server/storage/models/.gitignore              |   1 -
 server/utils/AiProviders/anthropic/index.js   |   4 +-
 server/utils/AiProviders/cohere/index.js      |   4 +-
 server/utils/AiProviders/deepseek/index.js    |   6 +-
 .../utils/AiProviders/gemini/defaultModels.js |  15 +-
 server/utils/AiProviders/gemini/index.js      |  12 +-
 .../AiProviders/gemini/syncStaticLists.mjs    |  31 ++++
 server/utils/AiProviders/groq/index.js        |   4 +-
 .../{modelMap/legacy.js => modelMap.js}       |  13 +-
 server/utils/AiProviders/modelMap/index.js    | 140 ------------------
 server/utils/AiProviders/openAi/index.js      |   9 +-
 server/utils/AiProviders/xai/index.js         |   6 +-
 12 files changed, 68 insertions(+), 177 deletions(-)
 rename server/utils/AiProviders/{modelMap/legacy.js => modelMap.js} (90%)
 delete mode 100644 server/utils/AiProviders/modelMap/index.js

diff --git a/server/storage/models/.gitignore b/server/storage/models/.gitignore
index 5e83df7bcc4..f5c78ac877c 100644
--- a/server/storage/models/.gitignore
+++ b/server/storage/models/.gitignore
@@ -9,4 +9,3 @@ gemini
 togetherAi
 tesseract
 ppio
-context-windows/*
\ No newline at end of file
diff --git a/server/utils/AiProviders/anthropic/index.js b/server/utils/AiProviders/anthropic/index.js
index 450b376b3e0..c4f53acf639 100644
--- a/server/utils/AiProviders/anthropic/index.js
+++ b/server/utils/AiProviders/anthropic/index.js
@@ -45,11 +45,11 @@ class AnthropicLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.get("anthropic", modelName) ?? 100_000;
+    return MODEL_MAP.anthropic[modelName] ?? 100_000;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.get("anthropic", this.model) ?? 100_000;
+    return MODEL_MAP.anthropic[this.model] ?? 100_000;
   }
 
   isValidChatCompletionModel(_modelName = "") {
diff --git a/server/utils/AiProviders/cohere/index.js b/server/utils/AiProviders/cohere/index.js
index a6647f5b658..33b65df21c0 100644
--- a/server/utils/AiProviders/cohere/index.js
+++ b/server/utils/AiProviders/cohere/index.js
@@ -63,11 +63,11 @@ class CohereLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.get("cohere", modelName) ?? 4_096;
+    return MODEL_MAP.cohere[modelName] ?? 4_096;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.get("cohere", this.model) ?? 4_096;
+    return MODEL_MAP.cohere[this.model] ?? 4_096;
   }
 
   async isValidChatCompletionModel(model = "") {
diff --git a/server/utils/AiProviders/deepseek/index.js b/server/utils/AiProviders/deepseek/index.js
index ddb2de317c7..b91332a84a1 100644
--- a/server/utils/AiProviders/deepseek/index.js
+++ b/server/utils/AiProviders/deepseek/index.js
@@ -29,7 +29,7 @@ class DeepSeekLLM {
 
     this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
-    this.log(`Initialized ${this.model} with context window ${this.promptWindowLimit()}`);
+    this.log("Initialized with model:", this.model);
   }
 
   log(text, ...args) {
@@ -53,11 +53,11 @@ class DeepSeekLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.get("deepseek", modelName) ?? 8192;
+    return MODEL_MAP.deepseek[modelName] ?? 8192;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.get("deepseek", this.model) ?? 8192;
+    return MODEL_MAP.deepseek[this.model] ?? 8192;
   }
 
   async isValidChatCompletionModel(modelName = "") {
diff --git a/server/utils/AiProviders/gemini/defaultModels.js b/server/utils/AiProviders/gemini/defaultModels.js
index 4a52dc99c65..fde63341d2a 100644
--- a/server/utils/AiProviders/gemini/defaultModels.js
+++ b/server/utils/AiProviders/gemini/defaultModels.js
@@ -1,7 +1,7 @@
 const { MODEL_MAP } = require("../modelMap");
 
 const stableModels = [
-  // %STABLE_MODELS% - updated 2025-05-13T23:13:58.920Z
+  // %STABLE_MODELS% - updated 2025-04-07T20:29:49.276Z
   "gemini-1.5-pro-001",
   "gemini-1.5-pro-002",
   "gemini-1.5-pro",
@@ -14,7 +14,6 @@ const stableModels = [
   "gemini-2.0-flash-001",
   "gemini-2.0-flash-lite-001",
   "gemini-2.0-flash-lite",
-  "gemini-2.0-flash-preview-image-generation",
   // %EOC_STABLE_MODELS%
 ];
 
@@ -23,7 +22,7 @@ const stableModels = [
 // generally, v1beta models have `exp` in the name, but not always
 // so we check for both against a static list as well via API.
 const v1BetaModels = [
-  // %V1BETA_MODELS% - updated 2025-05-13T23:13:58.920Z
+  // %V1BETA_MODELS% - updated 2025-04-07T20:29:49.276Z
   "gemini-1.5-pro-latest",
   "gemini-1.5-flash-latest",
   "gemini-1.5-flash-8b-latest",
@@ -31,9 +30,6 @@ const v1BetaModels = [
   "gemini-1.5-flash-8b-exp-0924",
   "gemini-2.5-pro-exp-03-25",
   "gemini-2.5-pro-preview-03-25",
-  "gemini-2.5-flash-preview-04-17",
-  "gemini-2.5-flash-preview-04-17-thinking",
-  "gemini-2.5-pro-preview-05-06",
   "gemini-2.0-flash-exp",
   "gemini-2.0-flash-exp-image-generation",
   "gemini-2.0-flash-lite-preview-02-05",
@@ -45,7 +41,6 @@ const v1BetaModels = [
   "gemini-2.0-flash-thinking-exp",
   "gemini-2.0-flash-thinking-exp-1219",
   "learnlm-1.5-pro-experimental",
-  "learnlm-2.0-flash-experimental",
   "gemma-3-1b-it",
   "gemma-3-4b-it",
   "gemma-3-12b-it",
@@ -53,17 +48,17 @@ const v1BetaModels = [
   // %EOC_V1BETA_MODELS%
 ];
 
-const defaultGeminiModels = () => [
+const defaultGeminiModels = [
   ...stableModels.map((model) => ({
     id: model,
     name: model,
-    contextWindow: MODEL_MAP.get("gemini", model),
+    contextWindow: MODEL_MAP.gemini[model],
     experimental: false,
   })),
   ...v1BetaModels.map((model) => ({
     id: model,
     name: model,
-    contextWindow: MODEL_MAP.get("gemini", model),
+    contextWindow: MODEL_MAP.gemini[model],
     experimental: true,
   })),
 ];
diff --git a/server/utils/AiProviders/gemini/index.js b/server/utils/AiProviders/gemini/index.js
index 393f0cd5227..bd2268e5174 100644
--- a/server/utils/AiProviders/gemini/index.js
+++ b/server/utils/AiProviders/gemini/index.js
@@ -107,7 +107,7 @@ class GeminiLLM {
     try {
       const cacheModelPath = path.resolve(cacheFolder, "models.json");
       if (!fs.existsSync(cacheModelPath))
-        return MODEL_MAP.get("gemini", modelName) ?? 30_720;
+        return MODEL_MAP.gemini[modelName] ?? 30_720;
 
       const models = safeJsonParse(fs.readFileSync(cacheModelPath));
       const model = models.find((model) => model.id === modelName);
@@ -118,13 +118,15 @@ class GeminiLLM {
       return model.contextWindow;
     } catch (e) {
       console.error(`GeminiLLM:promptWindowLimit`, e.message);
-      return MODEL_MAP.get("gemini", modelName) ?? 30_720;
+      return MODEL_MAP.gemini[modelName] ?? 30_720;
     }
   }
 
   promptWindowLimit() {
     try {
-      if (!fs.existsSync(this.cacheModelPath)) return MODEL_MAP.get("gemini", this.model) ?? 30_720;
+      if (!fs.existsSync(this.cacheModelPath))
+        return MODEL_MAP.gemini[this.model] ?? 30_720;
+
       const models = safeJsonParse(fs.readFileSync(this.cacheModelPath));
       const model = models.find((model) => model.id === this.model);
       if (!model)
@@ -134,7 +136,7 @@ class GeminiLLM {
       return model.contextWindow;
     } catch (e) {
       console.error(`GeminiLLM:promptWindowLimit`, e.message);
-      return MODEL_MAP.get("gemini", this.model) ?? 30_720;
+      return MODEL_MAP.gemini[this.model] ?? 30_720;
     }
   }
 
@@ -279,7 +281,7 @@ class GeminiLLM {
 
     if (allModels.length === 0) {
       console.error(`Gemini:getGeminiModels - No models found`);
-      return defaultGeminiModels();
+      return defaultGeminiModels;
     }
 
     console.log(
diff --git a/server/utils/AiProviders/gemini/syncStaticLists.mjs b/server/utils/AiProviders/gemini/syncStaticLists.mjs
index b276995acdf..42382fae388 100644
--- a/server/utils/AiProviders/gemini/syncStaticLists.mjs
+++ b/server/utils/AiProviders/gemini/syncStaticLists.mjs
@@ -9,6 +9,7 @@
 import fs from "fs";
 import path from "path";
 import dotenv from "dotenv";
+import { MODEL_MAP } from "../modelMap.js";
 
 dotenv.config({ path: `../../../.env.development` });
 const existingCachePath = path.resolve('../../../storage/models/gemini')
@@ -45,4 +46,34 @@ function updateDefaultModelsFile(models) {
   fs.writeFileSync(path.join("./defaultModels.js"), defaultModelFileContents);
   console.log("Updated defaultModels.js. Dont forget to `yarn lint` and commit!");
 }
+
+function updateModelMap(models) {
+  const existingModelMap = MODEL_MAP;
+  console.log('Updating modelMap.js `gemini` object...')
+  console.log(`Removed existing gemini object (${Object.keys(existingModelMap.gemini).length} models) from modelMap.js`);
+  existingModelMap.gemini = {};
+
+  for (const model of models) existingModelMap.gemini[model.id] = model.contextWindow;
+  console.log(`Updated modelMap.js 'gemini' object with ${Object.keys(existingModelMap.gemini).length} models from API`);
+
+  // Update the modelMap.js file
+  const contents = `/**
+ * The model name and context window for all know model windows
+ * that are available through providers which has discrete model options.
+ * This file is automatically generated by syncStaticLists.mjs
+ * and should not be edited manually.
+ * 
+ * Last updated: ${new Date().toISOString()}
+ */
+const MODEL_MAP = {
+  ${Object.entries(existingModelMap).map(([key, value]) => `${key}: ${JSON.stringify(value, null, 2)}`).join(',\n')}
+};
+
+module.exports = { MODEL_MAP };
+`;
+  fs.writeFileSync(path.resolve("../modelMap.js"), contents);
+  console.log('Updated modelMap.js `gemini` object. Dont forget to `yarn lint` and commit!');
+}
+
 updateDefaultModelsFile(models);
+updateModelMap(models);
diff --git a/server/utils/AiProviders/groq/index.js b/server/utils/AiProviders/groq/index.js
index 8eddefab0c5..9e7e77fa16b 100644
--- a/server/utils/AiProviders/groq/index.js
+++ b/server/utils/AiProviders/groq/index.js
@@ -49,11 +49,11 @@ class GroqLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.get("groq", modelName) ?? 8192;
+    return MODEL_MAP.groq[modelName] ?? 8192;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.get("groq", this.model) ?? 8192;
+    return MODEL_MAP.groq[this.model] ?? 8192;
   }
 
   async isValidChatCompletionModel(modelName = "") {
diff --git a/server/utils/AiProviders/modelMap/legacy.js b/server/utils/AiProviders/modelMap.js
similarity index 90%
rename from server/utils/AiProviders/modelMap/legacy.js
rename to server/utils/AiProviders/modelMap.js
index 2faf99dc237..a80400c30b4 100644
--- a/server/utils/AiProviders/modelMap/legacy.js
+++ b/server/utils/AiProviders/modelMap.js
@@ -1,4 +1,12 @@
-const LEGACY_MODEL_MAP = {
+/**
+ * The model name and context window for all know model windows
+ * that are available through providers which has discrete model options.
+ * This file is automatically generated by syncStaticLists.mjs
+ * and should not be edited manually.
+ *
+ * Last updated: 2025-04-07T20:29:49.277Z
+ */
+const MODEL_MAP = {
   anthropic: {
     "claude-instant-1.2": 100000,
     "claude-2.0": 100000,
@@ -109,4 +117,5 @@ const LEGACY_MODEL_MAP = {
     "grok-beta": 131072,
   },
 };
-module.exports = LEGACY_MODEL_MAP;
+
+module.exports = { MODEL_MAP };
diff --git a/server/utils/AiProviders/modelMap/index.js b/server/utils/AiProviders/modelMap/index.js
deleted file mode 100644
index dbf0aff6a60..00000000000
--- a/server/utils/AiProviders/modelMap/index.js
+++ /dev/null
@@ -1,140 +0,0 @@
-const path = require("path");
-const fs = require("fs");
-const LEGACY_MODEL_MAP = require("./legacy");
-
-class ContextWindowFinder {
-  static instance = null;
-  static modelMap = LEGACY_MODEL_MAP;
-
-  /**
-   * Mapping for AnythingLLM provider <> LiteLLM provider
-   * @type {Record<string, string>}
-   */
-  static trackedProviders = {
-    anthropic: "anthropic",
-    openai: "openai",
-    cohere: "cohere_chat",
-    gemini: "vertex_ai-language-models",
-    groq: "groq",
-    xai: "xai",
-    deepseek: "deepseek",
-  };
-  static expiryMs = 1000 * 60 * 60 * 24 * 3; // 3 days
-  static remoteUrl =
-    "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json";
-
-  cacheLocation = path.resolve(
-    process.env.STORAGE_DIR
-      ? path.resolve(process.env.STORAGE_DIR, "models", "context-windows")
-      : path.resolve(__dirname, `../../../storage/models/context-windows`)
-  );
-  cacheFilePath = path.resolve(this.cacheLocation, "context-windows.json");
-  cacheFileExpiryPath = path.resolve(this.cacheLocation, ".cached_at");
-
-  constructor() {
-    if (ContextWindowFinder.instance) return ContextWindowFinder.instance;
-    ContextWindowFinder.instance = this;
-    if (!fs.existsSync(this.cacheLocation))
-      fs.mkdirSync(this.cacheLocation, { recursive: true });
-    this.#pullRemoteModelMap();
-  }
-
-  log(text, ...args) {
-    console.log(`\x1b[33m[ContextWindowFinder]\x1b[0m ${text}`, ...args);
-  }
-
-  /**
-   * Checks if the cache is stale by checking if the cache file exists and if the cache file is older than the expiry time.
-   * @returns {boolean}
-   */
-  get isCacheStale() {
-    if (!fs.existsSync(this.cacheFileExpiryPath)) return true;
-    const cachedAt = fs.readFileSync(this.cacheFileExpiryPath, "utf8");
-    return Date.now() - cachedAt > ContextWindowFinder.expiryMs;
-  }
-
-  get cache() {
-    if (!fs.existsSync(this.cacheFileExpiryPath)) return null;
-    if (!this.isCacheStale)
-      return JSON.parse(
-        fs.readFileSync(this.cacheFilePath, { encoding: "utf8" })
-      );
-    return null;
-  }
-
-  /**
-   * Pulls the remote model map from the remote URL, formats it and caches it.
-   * @returns {Record<string, Record<string, number>>} - The formatted model map
-   */
-  async #pullRemoteModelMap() {
-    const remoteContexWindowMap = await fetch(ContextWindowFinder.remoteUrl)
-      .then((res) => res.json())
-      .then((data) => {
-        fs.writeFileSync(this.cacheFilePath, JSON.stringify(data, null, 2));
-        fs.writeFileSync(this.cacheFileExpiryPath, Date.now().toString());
-        this.log("Remote model map synced and cached");
-        return data;
-      })
-      .catch((error) => {
-        this.log("Error syncing remote model map", error);
-        return null;
-      });
-    if (!remoteContexWindowMap) return null;
-
-    const modelMap = this.#formatModelMap(remoteContexWindowMap);
-    fs.writeFileSync(this.cacheFilePath, JSON.stringify(modelMap, null, 2));
-    fs.writeFileSync(this.cacheFileExpiryPath, Date.now().toString());
-    return modelMap;
-  }
-
-  /**
-   * Formats the remote model map to a format that is compatible with how we store the model map
-   * for all providers who use it.
-   * @param {Record<string, any>} modelMap - The remote model map
-   * @returns {Record<string, Record<string, number>>} - The formatted model map
-   */
-  #formatModelMap(modelMap = {}) {
-    const formattedModelMap = {};
-
-    for (const [provider, liteLLMProviderTag] of Object.entries(
-      ContextWindowFinder.trackedProviders
-    )) {
-      formattedModelMap[provider] = {};
-      const matches = Object.entries(modelMap).filter(
-        ([_key, config]) => config.litellm_provider === liteLLMProviderTag
-      );
-      for (const [key, config] of matches) {
-        const contextWindow = Number(config.max_input_tokens);
-        if (isNaN(contextWindow)) continue;
-
-        // Some models have a provider/model-tag format, so we need to get the last part since we dont do paths
-        // for names with the exception of some router-providers like OpenRouter or Together.
-        const modelName = key.split("/").pop();
-        formattedModelMap[provider][modelName] = contextWindow;
-      }
-    }
-    return formattedModelMap;
-  }
-
-  /**
-   * Gets the context window for a given provider and model.
-   * @param {string} provider - The provider to get the context window for
-   * @param {string} model - The model to get the context window for
-   * @returns {number} - The context window for the given provider and model
-   */
-  get(provider = null, model = null) {
-    if (!provider || !this.cache || !this.cache[provider]) return null;
-    if (!model) return this.cache[provider];
-    const modelContextWindow = this.cache[provider][model];
-    if (!modelContextWindow) {
-      this.log("Invalid access to model context window - not found in cache", {
-        provider,
-        model,
-      });
-      return null;
-    }
-    return Number(modelContextWindow);
-  }
-}
-
-module.exports = { MODEL_MAP: new ContextWindowFinder() };
diff --git a/server/utils/AiProviders/openAi/index.js b/server/utils/AiProviders/openAi/index.js
index 586532ce8dd..2e5996c939a 100644
--- a/server/utils/AiProviders/openAi/index.js
+++ b/server/utils/AiProviders/openAi/index.js
@@ -25,11 +25,6 @@ class OpenAiLLM {
 
     this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
-    this.log(`Initialized ${this.model} with context window ${this.promptWindowLimit()}`);
-  }
-
-  log(text, ...args) {
-    console.log(`\x1b[36m[${this.constructor.name}]\x1b[0m ${text}`, ...args);
   }
 
   /**
@@ -59,11 +54,11 @@ class OpenAiLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.get("openai", modelName) ?? 4_096;
+    return MODEL_MAP.openai[modelName] ?? 4_096;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.get("openai", this.model) ?? 4_096;
+    return MODEL_MAP.openai[this.model] ?? 4_096;
   }
 
   // Short circuit if name has 'gpt' since we now fetch models from OpenAI API
diff --git a/server/utils/AiProviders/xai/index.js b/server/utils/AiProviders/xai/index.js
index ad0a7e5eecf..2319e72206e 100644
--- a/server/utils/AiProviders/xai/index.js
+++ b/server/utils/AiProviders/xai/index.js
@@ -28,7 +28,7 @@ class XAiLLM {
 
     this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
-    this.log(`Initialized ${this.model} with context window ${this.promptWindowLimit()}`);
+    this.log("Initialized with model:", this.model);
   }
 
   log(text, ...args) {
@@ -52,11 +52,11 @@ class XAiLLM {
   }
 
   static promptWindowLimit(modelName) {
-    return MODEL_MAP.get("xai", modelName) ?? 131_072;
+    return MODEL_MAP.xai[modelName] ?? 131_072;
   }
 
   promptWindowLimit() {
-    return MODEL_MAP.get("xai", this.model) ?? 131_072;
+    return MODEL_MAP.xai[this.model] ?? 131_072;
   }
 
   isValidChatCompletionModel(_modelName = "") {

From 00e1c54b4b60ad79b8d8dc3ce27e8e70c4e05262 Mon Sep 17 00:00:00 2001
From: timothycarambat <rambat1010@gmail.com>
Date: Wed, 14 May 2025 11:00:50 -0700
Subject: [PATCH 5/5] update newline printing

---
 docker/docker-entrypoint.sh | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/docker/docker-entrypoint.sh b/docker/docker-entrypoint.sh
index 9c47de25c4e..80296accc2f 100755
--- a/docker/docker-entrypoint.sh
+++ b/docker/docker-entrypoint.sh
@@ -4,12 +4,14 @@
 if [ -z "$STORAGE_DIR" ]; then
     echo "================================================================"
     echo "⚠️  ⚠️  ⚠️  WARNING: STORAGE_DIR environment variable is not set! ⚠️  ⚠️  ⚠️"
+    echo ""
     echo "Not setting this will result in data loss on container restart since"
     echo "the application will not have a persistent storage location."
     echo "It can also result in weird errors in various parts of the application."
-    echo "\n\n"
+    echo ""
     echo "Please run the container with the official docker command at"
     echo "https://docs.anythingllm.com/installation-docker/quickstart"
+    echo ""
     echo "⚠️  ⚠️  ⚠️  WARNING: STORAGE_DIR environment variable is not set! ⚠️  ⚠️  ⚠️"
     echo "================================================================"
 fi