Verity Mod API Connection Failed: Fix 401, 429 and Timeouts

What “API connection failed” means

A Verity Mod API connection failure means the game reached the AI request step but did not receive a usable response from the selected provider. On Java Edition, that provider is usually Groq or a local Ollama service. The message does not automatically mean the mod file is broken, and it does not tell you to reinstall Minecraft.

Start with the exact wording. A visible status such as 401, 403, 404, 429, 500, 502, or 503 is more useful than the general phrase “connection failed.” A timeout, connection refusal, and model-not-found response also point to different repairs. Take one screenshot with the secret hidden, note the time, and stop sending more prompts until you know which provider is active.

If Minecraft crashes before a world opens, Verity never appears, or the dialogue interface is absent, use the general troubleshooting guide. This page begins after the mod and interaction have already loaded.

Use the status code as your first route

Groq and Ollama both return standard HTTP status codes, but the likely cause depends on the provider.

Error	What it usually means	First action
400	The request format or a setting is invalid	Re-check the current Verity setup and selected model
401	Groq did not receive valid credentials	Check the active key without displaying or sharing it
403	The account or project cannot use the requested resource	Check model permissions and account restrictions
404	The endpoint or model name was not found	Verify the provider, endpoint, and exact model name
429	A request, token, or other rate limit was reached	Stop retrying, wait, then check current account limits
500	The provider encountered an internal error	Wait and retry one short prompt later
502	A gateway or upstream service failed	Treat it as a likely temporary provider path failure
503	The service is unavailable or overloaded	Wait, reduce repeated requests, and check provider status
Timeout	No complete response arrived in time	Test network reachability or local model response time
Connection refused	Nothing is listening at the configured local address	Start Ollama and verify the endpoint

Use the table to choose the first diagnostic branch, then read the message beside the code. It may name a model, endpoint, authentication problem, or limit. Preserve that message when you create a support report, but remove credentials and personal paths.

Confirm whether Verity is using Groq or Ollama

Test one provider at a time. In stable Verity JE 5.7.4, open Mods > Verity > Config. Check Local LLM > Use Ollama first: when enabled, it overrides AI Settings > AI Provider. A menu that still reads GROQ does not prove Groq is handling the request. For a different release, use its matching Java configuration instructions.

Groq is a cloud connection. It needs an active Groq API key, a model your Groq project can use, and outbound internet access. A working local Ollama installation does not help a Verity profile that is still configured for Groq.

Ollama is a local service. Its API normally runs on your own computer. It needs the Ollama application or service to be running and the configured model to be installed. Replacing a Groq key does not help a profile configured for Ollama.

After you identify the provider, write it in your notes. If you cannot tell which provider is selected, return to the Java setup guide and configure one path from the beginning. Switching settings repeatedly can hide the original error.

Fix a Groq 401 or invalid API key error

Groq documents 401 Unauthorized as a request without valid authentication credentials. For a Verity user, the practical causes are usually a missing key, a revoked key, an incomplete paste, extra characters, or a key entered in the wrong place.

Check the following without exposing the value:

1. Open your own GroqCloud console and confirm that the intended key still exists.
2. Enter the key only in the location shown by the current Verity JE setup flow.
3. Remove accidental spaces, line breaks, quotation marks, or a copied label around the key.
4. Save the setting, close the configuration screen, and restart the game if the setup requires it.
5. Send one short text prompt.

Do not paste the key into this website, a random key checker, a Discord message, or a public issue. A service can verify a credential only by receiving it, so an unknown “checker” creates a new security problem.

If the key appeared on a stream, in a screenshot, in a log, or in shell history, consider it exposed. Revoke it in GroqCloud and create a replacement. Editing the screenshot after posting does not erase copies that may already exist.

Fix a Groq 403 error

A 403 Forbidden response means the request was understood but is not allowed. This differs from 401: changing the spelling of a valid key may not solve a permissions problem.

Check which Groq project or organization owns the key. Then confirm that the requested model is available to that project and has not been blocked by an organization or project policy. Groq documents that restricted models can return a 403. If the Verity configuration names an older model that is no longer available to your account, compare it with the current maintainer setup before changing anything.

Choose the exact model supported by the current mod instructions rather than a similar-looking name. Verity may expect a specific response format or capability. If the project instructions and your account permissions disagree, record both details for support.

Fix a 404 or model-not-found error

A 404 Not Found usually points to the requested resource rather than the Minecraft world. With Groq, check the configured model and provider endpoint. With Ollama, check that the model exists locally under the exact name Verity requests.

Model names are exact identifiers. A missing tag, different punctuation, or a name copied from an old tutorial can produce a failure even when another model works on the same computer. List the models installed in Ollama, compare the complete name with the Verity setting, and test that same model outside Minecraft.

Also confirm that you did not copy an Ollama endpoint into the Groq setting or the other way around. A valid address for one provider can be a valid-looking but nonexistent route for the other.

Fix Verity Mod error 429

Groq documents 429 Too Many Requests as a rate-limit response. Its limits can be measured by requests, tokens, audio time, or separate input and output token allowances. The first limit reached can stop the request even when another allowance still has room.

The worst response is to click or send the prompt repeatedly. Those retries can add more requests while the current window is still limited.

Use this sequence:

1. Stop sending prompts.
2. Wait for the active rate-limit window to reset.
3. Open the Groq Limits page for the organization and project that owns the key.
4. Check the selected model and the available request and token limits.
5. Restart with one short text prompt, not voice and a long conversation history.

Groq may include a retry-after value in a direct 429 response, but a Minecraft mod may not display that header. If Verity shows only the code, waiting before one controlled retry is safer than estimating how many rapid attempts will work.

A 429 does not prove that someone stole your key. If the usage looks unfamiliar, revoke the key and review account activity. If the usage matches your own repeated tests, reduce the retry rate and conversation size.

Handle 500, 502, and 503 errors

Groq classifies 500, 502, and 503 as server-side failures. Ollama can also return a 500, a 502 when a cloud model cannot be reached, or a 503 when the service is overloaded. These codes are not instructions to recreate your world.

First, wait and try one short prompt later. If every user is affected by a provider incident, local file changes will not restore the service. If only your setup fails, test the same provider independently and compare the result.

For local Ollama, too many queued requests can produce a 503. Close duplicate clients, stop repeated prompts, and let the current model finish loading. A large model may need more memory than the computer has available, which can look like a long pause before an eventual failure. Use the Ollama setup guide to check model size and local resources.

Diagnose a timeout

A timeout means Verity waited but did not receive a complete answer in the allowed period. It does not identify the cause by itself.

For Groq, confirm that normal internet access works, then consider firewall, VPN, filtering, DNS, or a temporary provider delay. Test without changing several network tools at once. If a school, workplace, or managed device blocks the provider, use only changes allowed by that network’s policy.

For Ollama, the first response can be slow while a model loads into memory. Test the exact model directly. If it responds outside Minecraft after a long delay, the service works but may not finish within Verity’s timeout. A smaller supported model or a machine with more available memory may be necessary, but follow the current maintainer guidance before changing models.

A “connection refused” error usually means the local service is not listening at the configured address. Start Ollama, confirm its API is available at the expected local endpoint, and retry once instead of waiting for a timeout.

Test Ollama without changing Verity

Ollama’s official documentation states that its local API is served by default at http://localhost:11434/api. Stable Verity JE 5.7.4 instead configures a LiteLLM gateway at http://127.0.0.1:4000/v1/, which forwards requests to Ollama. Both processes must be running. Follow the standalone service and gateway checks before changing Verity’s LiteLLM URL. Do not replace the gateway address with Ollama’s native /api URL or apply a 6.0.0 beta tutorial to this stable release.

Use Ollama’s own interface or commands to answer a short prompt with the exact model configured in Verity. This separates three cases:

• If Ollama is not reachable, repair or start the local service.
• If Ollama is reachable but the model is missing, install or select the documented model.
• If the model answers directly but Verity fails, compare Verity’s model name and endpoint with the successful local test.

Check Ollama’s logs only after reproducing the failure once. Logs are more useful when you know the time and model. Before sharing an excerpt, remove your computer username, home-directory paths, network addresses, prompt content you consider private, and any unrelated account details.

Separate connection errors from voice errors

Test text before voice. A microphone problem, speech-to-text failure, or text-to-speech failure can make Verity seem silent even when the language model connection works.

Send a short typed prompt. If typed text receives an answer, the Groq or Ollama response path is working. Move to the talking and voice guide for microphone permission, input selection, speech recognition, language behavior, and audio output.

If typed text also fails with a status code, remain on this page. Changing microphone settings cannot repair a 401 or 429.

Use a clean diagnostic test

Once the provider can answer independently, test Verity in a clean Java profile:

1. Use the exact Minecraft version and loader listed for your Verity release.
2. Keep only Verity and its required dependencies.
3. Select one provider.
4. Open a new test world.
5. Send one short typed prompt.
6. Record the exact response and time.

This test removes unrelated mods, long conversation history, voice processing, and valuable world data from the diagnosis. If it succeeds, add your normal mods back in small groups. If it fails, the resulting report is much easier for a maintainer to reproduce.

What to include in a support report

Include:

• Verity project name and Project ID.
• Verity release, Minecraft version, and loader.
• Groq or Ollama as the selected provider.
• Exact status code and error message.
• Time and time zone of the test.
• Whether the provider worked outside Minecraft.
• Whether a clean profile and new world produced the same result.
• The smallest relevant log excerpt with secrets removed.

Leave out the API key, authorization header, account password, full log archive, IP address, private server address, and unedited configuration screenshots. A useful report proves the failing layer without giving anyone access to your account.

After the error is fixed, keep a note of the working release, loader, provider, model, and date. That gives you a known-good baseline when a future Minecraft, mod, model, or provider update changes the connection.