Verity Mod Groq API Key Setup

What the Groq API key does

Verity JE can use Groq to send a prompt to a cloud language model and receive a response. The API key authorizes those requests under your Groq account. It is not a Verity download code, a Minecraft login, or a value that this website should ever receive.

Treat it like a password. Anyone who obtains it may be able to make requests under your account until it is revoked or limited. Create it in GroqCloud, enter it only in the Verity JE settings on your computer, and keep it out of screenshots, logs, chat, streams, and repositories.

Groq or Ollama: which should you pick?

	Groq (cloud)	Ollama (local)
Needs an API key	Yes	No
Needs a strong computer	No	Yes, especially RAM or VRAM
Response speed	Usually fast	Depends on your hardware and model
Where prompts go	Groq’s servers	Stay on your computer
Extra software	None beyond the mod	Ollama, plus LiteLLM for 5.7.4
Typical problems	Invalid key, rate limits, model access	Model not loaded, wrong tag, gateway errors

Most players start with Groq because it involves fewer moving parts: one key, one menu, one test. Choose Ollama if you prefer not to send prompts to a cloud service or want to play without depending on an online provider.

Before creating an API key

Finish the non-AI part of the Java setup first. Minecraft should launch with the exact loader and game version. Verity JE and its current dependencies should load in a clean profile. The companion or documented interaction should appear in a new test world.

If Minecraft crashes before the title screen, a Groq API key will not fix it. Resolve Java runtime, loader, dependency, duplicate-file, or mod-conflict errors before adding a cloud provider. This also reduces the amount of configuration you need to expose while asking for help.

Use Verity JE 5.7.4 with Forge 1.20.1 for the stable setup. Its changelog records a missing Groq model fix. If an older 5.7.x file reports a missing model, check the installed mod version before repeatedly replacing a valid key. The older NeoForge 3.4.1 release and the 6.0.0 betas are separate setup targets.

Create the API key in GroqCloud

1. Open GroqCloud’s API Keys page and sign in to your own account.
2. Choose Create API Key and give the key a recognizable name such as “Verity test profile.” Keep the secret out of its name.
3. Copy the generated value into your trusted local configuration or password manager. The expected result is a new entry in your Groq project’s API Keys list, not a Minecraft login or download code.

The console generally shows the full secret only when it is created. If you close the dialog without copying it, do not search old screenshots; create a new key and delete the unused one.

The console says only a team owner or a user with the developer role can create or manage keys. If that control is unavailable, check your Groq project permissions; a third-party key generator cannot resolve them. Groq’s quickstart links to this same official key-creation page.

Do not create an API key through a “free key generator,” paste site, support bot, video description mirror, or form claiming to configure Verity remotely. The key comes from GroqCloud. Verity Mod Guide has no setup form and cannot validate it for you.

Enter it in stable Verity JE 5.7.4

The following labels are verified in the maintainer’s published 5.7.4 Forge artifact. They are not instructions for legacy 3.4.1 or the reworked 6.0.0 beta menu.

1. Launch the Forge 1.20.1 profile containing Verity 5.7.4. At the title screen, choose Mods, select Verity, and open Config. Expect the Verity Configuration screen.
2. Open AI Settings. Set AI Provider to GROQ and paste the generated secret into API Key, without extra spaces or quotation marks.
3. Leave AI Intelligence Level at its default, FAST, for the first connection test. This is a preset control, not a field for pasting a Groq model ID.
4. Open Local LLM and turn Use Ollama off. In 5.7.4, that switch overrides AI Provider, so leaving it on sends conversation requests to the local gateway instead of Groq.
5. Apply the changes, return to the title screen, and reopen Config to check the selected provider and switch. Keep the key field out of screenshots.

For reading an existing configuration, the corresponding names are AISettings.apiKey, AISettings.aiProvider, AISettings.aiModel, and AISettings.use_ollama. Use the in-game screen to edit them; do not copy a complete configuration from another profile or invent a beta field name.

Never paste the value into Minecraft chat; chat can be logged, shown to other players, or captured by a stream.

If the setup stores a local configuration file, protect that file as a secret. Do not include it when sharing the whole profile, uploading logs, making a modpack archive, or committing launcher files to Git. A .gitignore entry is useful for repositories, but it does not erase a secret already committed to history.

Choose a model without mixing up the settings

In 5.7.4, FAST is Verity’s AI Intelligence Level preset. It is not a Groq model ID, and the API-key field is not a place to enter either value. Start with the shipped preset, then read the exact model named in any error. The 5.7.4 changelog’s model fix describes that release; it does not guarantee permanent access to that model.

Compare a rejected model with Groq’s supported models and your project’s model permissions. Availability and permission are different checks. A model can exist in Groq’s catalog but be unavailable to your project. Replacing a valid key will not remove that restriction. Do not substitute a speech-recognition model for a chat model simply because both appear in the catalog.

For 6.0.0-beta.9, follow the maintainer’s 6.0.0+ tutorial and compare it with your installed menu. The beta.9 notes confirm that AI configuration changed again. Do not paste a 5.7.4 configuration file into the beta or assume that the FAST control has the same role there.

If JE Extra’s Chat override is enabled, it supplies separate endpoint, credential, and model settings. Diagnose those active settings, not just Verity’s base provider menu. Leave the override disabled while establishing a working base-mod setup.

First test

Open a new test world and wait until Verity appears. Open ordinary Minecraft chat with T (or your remapped chat key), type Verity, say hello., and press Enter. The name satisfies 5.7.4’s optional General > Require Verity filter. Keep General > Immersive Mode off so it does not hide the reply. Text tests the language-model connection without involving microphone permissions, speech recognition, or voice output. Wait for one request to finish before sending another.

There are three useful outcomes:

1. Text response arrives. The cloud connection works. You can test voice separately.
2. Authentication is rejected. Re-check the API key, whether it was revoked, and whether it was entered in the correct local field.
3. Timeout or service error. Check network access, service availability, account limits, and the exact error without exposing the secret.

Record the error code or message, not the API key. When a support helper asks to “see the key,” decline. Troubleshooting does not require publishing it.

Common setup failures

What you see	What to check next
401 after pasting a key	Confirm it is active, complete, and saved in the running profile without surrounding whitespace
403 naming a model	Check that model’s permissions in the key’s Groq project
Model-not-found or 404	Compare the requested model and endpoint with the current provider documentation
429	Stop retries and check the model’s current organization limits
Requests to 127.0.0.1 despite choosing Groq	In 5.7.4, check Local LLM > Use Ollama; in 6.x also check any enabled add-on override
Timeout with no HTTP code	Check authorized network access; do not disable managed security controls

Use the full API error checklist for each code’s repair and retest. Keep the successful provider and model settings unchanged when moving on to voice.

Understanding rate limits

A 429 means Groq received the request but declined it because a limit was reached. Limits can apply per minute, per day, per model, and to tokens as well as requests. Long conversations send more text with each message, so a chat that worked for an hour can start hitting limits even though nothing changed.

If you see 429 errors, wait before retrying, send shorter prompts, and avoid holding the voice key down repeatedly while a reply is pending. Check the limits page in the GroqCloud console for your account. Creating several new keys does not usually help, because limits are tied to the organization rather than the individual key.

How many messages can you send?

Groq’s limits are listed per model on the Limits page of the console, as requests and tokens per minute and per day. Whichever runs out first stops Verity. The calculator below turns those four numbers into something practical: replies per minute, replies per day, and roughly how many hours of play that covers.

Groq limit calculator. Copy the four limits for your model from your Groq Limits page. The starting values are examples only.

Requests per minute (RPM) Tokens per minute (TPM) Requests per day (RPD) Tokens per day (TPD) Tokens per Verity message Messages you send per hour of play

Replies per minute

—

Replies per day

—

Hours of play per day

—

Without JavaScript: replies per minute = the smaller of RPM and TPM ÷ tokens per message; replies per day = the smaller of RPD and TPD ÷ tokens per message.

Finding “tokens per message”. This is the one number you have to estimate. Each Verity request carries its instructions and recent conversation as well as your words, so it is much larger than what you type and grows as a session goes on. The Groq console’s usage and log pages show the tokens used by recent requests: send a few messages, then read the typical size. Until you have measured it, 1,500 is a reasonable placeholder for a short conversation.

What the result usually shows:

• Tokens, not requests, run out first. With the example values, 30 requests per minute sounds generous, but at 1,500 tokens each, 6,000 tokens per minute allows only four replies a minute. Rapid-fire chatting hits a 429 quickly.
• Long sessions get more expensive. As tokens per message rise, both per-minute and per-day replies fall. If late-session replies start failing, a fresh conversation brings the size back down.
• Daily limits matter for streams and long recordings. If “hours of play per day” is lower than your planned session, pace your messages, choose a model with a larger daily allowance, or switch to local Ollama, which has no per-request limit.
• If a single message is bigger than the per-minute token limit, Groq returns 413 on every attempt. The error lookup explains why waiting does not fix that.

Add voice only after text works

Follow the chat and voice tests to check recognized speech and spoken replies separately. A working Groq chat connection does not prove that the microphone or speech service works. Keep the working chat configuration while diagnosing sound.

If the API key leaks

Revoke it in GroqCloud immediately. Rotation is the primary containment action; editing the screenshot or deleting a message may not remove copies. Create a replacement, update only the trusted local configuration, restart the affected process, and verify that the old value no longer works.

Search likely exposure locations: chat history, support posts, screenshots, stream recordings, shell history, shared profile archives, cloud-sync folders, and Git history. Remove or restrict those artifacts where possible, but still assume the old value was copied.

Sharing a modpack or profile safely

Players often export a working instance to share with friends. Before you do, open Verity’s Config and clear the API Key field, or remove the config file from the export. Each player should create and enter their own key. Sharing one key means everyone’s usage counts against your account, and you cannot revoke access for a single person.

The same applies to modpacks: a pack published with a filled-in key exposes that key to every downloader. Pack authors should ship with the field empty and explain where players add their own.

Ongoing hygiene

Use a separate API key for Verity so it can be revoked independently. Remove keys you no longer use. Review Groq account usage for unexpected activity. Keep prompts free of passwords, private documents, personal addresses, and information you would not send to a cloud service.

When sharing a troubleshooting log, make a copy, redact secrets completely, and inspect the final text before uploading. Do not rely on partial masking or a promise that a private message will stay private.

If you prefer not to send prompts to a cloud model, use the local Ollama setup. Ollama avoids a Groq API key but requires a running local model service and enough computer resources. It is a different privacy choice, not a one-click mode.