Mintplex-Labs/anything-llm · warning

\x1b[33m[.contextLimit warning]\x1b[0m Could not determine .

Error message

\x1b[33m[.contextLimit warning]\x1b[0m Could not determine .promptWindowLimit for provider ${provider}. This could lead to incorrect context window management by AnythingLLM since we cannot determine the context window limit for this provider/model combination.

What it means

AiProvider.contextLimit resolved the provider class via getLLMProviderClass and either found nothing or the class lacks a promptWindowLimit static, so it returns a hard fallback of 8,000 tokens. All context management (history trimming, prompt sizing) for that provider/model will assume 8k, which under- or over-fills the real window.

Source

Thrown at server/utils/agents/aibitat/providers/ai-provider.js:565

  /**
   * Get the context limit for a provider/model combination using static method in AIProvider class.
   * @param {string} provider
   * @param {string} modelName
   * @returns {number}
   */
  static contextLimit(provider = "openai", modelName) {
    if (typeof provider !== "string") {
      console.log(
        `\x1b[43m\x1b[30m[.contextLimit warning] A non-string provider for .contextLimit was given — Returning fallback context limit of 8000.\x1b[0m\n\x1b[43m\x1b[30mThis is a bug and should be reported so that context windows are properly managed by AnythingLLM.\x1b[0m`
      );
      console.trace();
      return 8_000;
    }

    const llm = getLLMProviderClass({ provider });
    if (!llm || !llm.hasOwnProperty("promptWindowLimit")) {
      console.warn(
        `\x1b[33m[.contextLimit warning]\x1b[0m Could not determine .promptWindowLimit for provider ${provider}. This could lead to incorrect context window management by AnythingLLM since we cannot determine the context window limit for this provider/model combination.`
      );
      return 8_000;
    }
    return llm.promptWindowLimit(modelName);
  }

  /**
   * Get the system prompt for a provider, with memories appended (when enabled).
   * @param {object} opts
   * @param {import("@prisma/client").workspaces | null} opts.workspace
   * @param {import("@prisma/client").users | null} opts.user
   * @param {string} [opts.prompt] - current user message, used for reranking injected memories
   * @returns {Promise<string>}
   */
  static async systemPrompt({ workspace = null, user = null, prompt = "" }) {
    const { SystemSettings } = require("../../../../models/systemSettings");
    const { promptWithMemories } = require("../../../memories");

View on GitHub (pinned to 20f6d3546c)

Solutions

1. Verify the provider string is spelled exactly as registered in getLLMProviderClass.
2. Add a static promptWindowLimit(modelName) to the provider class returning the correct token count.
3. Register the provider class in getLLMProviderClass's map so the lookup succeeds.
4. Until fixed, expect 8000 to be used — pick models whose real window is close or explicitly size prompts yourself.

Example fix

// before
class MyProvider extends Provider { /* no promptWindowLimit */ }

// after
class MyProvider extends Provider {
  static promptWindowLimit(modelName = "my-model") {
    return 128_000;
  }
}

// When registering a provider, assert it supports context limits:
const cls = getLLMProviderClass({ provider: 'myprovider' });
if (!cls || !('promptWindowLimit' in cls)) {
  throw new Error('provider class must implement static promptWindowLimit');
}

function hasPromptWindowLimit(cls) {
  return !!cls && typeof cls.promptWindowLimit === 'function';
}

When it happens

Trigger: provider string not matching a known key of the provider registry (typo like 'opeanai'); a custom/new provider class added without a static promptWindowLimit(modelName) method; a provider registered under a different name than the one passed in.

Common situations: Contributing a new LLM provider and forgetting the limit method; swapping provider identifiers after a refactor; using a fork where the registry map was extended but the class was not.

Related errors

• No LocalAI Base Path was set.
• No LocalAi token context limit was set.
• Unknown provider: ${config.provider}. Please use a valid pro
• Type "${type}" is not a valid type to sync.
• QEMU directory not found: ${dir}\nContents of ${parent}: ${c

AI-assisted analysis of Mintplex-Labs/anything-llm@20f6d3546c (2026-08-18). Data as JSON: /api/errors/8ecb39e5a971b58e. Report an issue: GitHub.