Verity Mod Guide

Live problem route checked July 20, 2026

Verity Mod API Connection Failed

Do not reinstall the jar until you read the status code. An API connection failed message usually means Verity JE loaded and reached its AI layer, but Groq, OpenRouter, or local Ollama did not return a usable answer. Match the code below, fix one layer, and retest with one short text prompt.

401 key check 429 rate limit Ollama local check

Error map

What each connection error means

Message or code Most likely layer First useful action
401 Unauthorized Credential Confirm the selected provider, revoke exposed keys, create one valid key, and paste it without quotes or spaces.
403 Forbidden Account permission Check model access, account restrictions, and provider permissions instead of changing Minecraft files.
404 Not Found Model or endpoint Return to the current provider model list and select a model supported by the installed build.
429 Too Many Requests Rate limit Stop retrying, inspect the exact limit and reset window, then test after the reset.
500, 502, or 503 Provider service Wait and retry once later. Check provider status before rotating keys or reinstalling.
Connection refused Local Ollama Confirm Ollama is running, the expected local address is reachable, and the configured model is installed.
Silent reply with no code Config, voice, or logs Test text first, open the latest log, and separate AI output from text-to-speech.

One clean test

Confirm that the mod loaded before touching the API

An API error is later than a loader error

Start by naming the layer that already works. If Minecraft reaches the main menu, the Verity entry appears in the mod list, the world opens, and the box or entity appears, the base installation has passed several checks. If a prompt then produces an API connection failed line, the immediate problem is probably not Forge, the Verity JE jar, or the world save. It is the provider path used after the character tries to answer.

If the game crashes before those steps, use the Java version route instead. The current checked Java file is verity-5.7.3.jar for Minecraft 1.20.1 Forge. The project page still lists the 1.21.1 NeoForge branch but warns that it is deprecated and has many bugs. Do not combine a 1.20.1 Forge file with a 1.21.1 NeoForge profile and then diagnose the result as an API problem.

Record five details before changing anything

Write down the provider selected in the mod, the exact status code, the selected model name, the time of the failure, and whether one short text prompt ever worked. Also note whether text appears but voice does not. Those facts distinguish a bad key, a rate limit, an unavailable model, a provider outage, a local Ollama service problem, and a separate audio failure.

Keep the log line but remove credentials. A useful support example is "Groq, status 429, model name, 14:20 UTC, text worked for five minutes." A screenshot containing the key is not useful and creates a second problem. Revoke any token that has already been exposed publicly.

Groq route

Fix Groq 401 and status 429 in the right order

For 401, replace the credential once

Groq documents 401 as a request without valid authentication credentials. Confirm that Groq is the provider selected in the Verity configuration. Create or rotate the key in Groq Console, paste only the key value into the local mod configuration, save, and restart the game. Avoid quotation marks, leading spaces, trailing spaces, or a key copied from the wrong provider.

Do not cycle through many keys without reading the code. If a fresh, private key still gives 401, confirm the config was saved and that an older config file is not overriding the in-game setting. If the code changes to 429, authentication now works and the next issue is quota. That is progress, not a reason to download the jar again.

For 429, inspect the exhausted limit

Groq defines 429 as too many requests in a time window. Its current documentation tracks requests per minute and day, tokens per minute and day, plus audio limits. Limits apply at the organization level, not to each key separately. That is why making another key in the same organization can produce the same failure.

Stop repeated prompts, open the Groq limits page, identify which counter reached its ceiling, and wait for the displayed reset. Then send one short text prompt. Test speech only after text works. If the limit returns quickly, reduce prompt frequency or context use, choose a supported lower-demand model, use a suitable paid tier, or follow the maintainer-linked local Ollama route if your computer can run it.

Provider split

Do not paste a Groq fix into OpenRouter or Ollama

OpenRouter has its own key, credits, and model IDs

Some current Verity-related configurations and add-ons expose an OpenRouter path, but an OpenRouter key is not a Groq key. Select OpenRouter in the local configuration before entering an OpenRouter token. If the account is out of credits or the requested model is unavailable, OpenRouter returns its own structured error. Read the provider message and model ID instead of assuming every failure is Groq 429.

Model names can change, and a model may be routed through different upstream providers. Use the current OpenRouter model list and the option supported by your installed build. Avoid copying a model ID from an old video without checking it. If text begins working but speech remains silent, move to the voice troubleshooting route; the API text path has already succeeded.

Ollama removes the cloud key but adds a local service

The Verity JE project links a dedicated Ollama setup tutorial for fully local operation. Local does not mean automatic. Ollama must be installed and running, the configured model must be present, and the mod must connect to the local service address it expects. A connection refused message usually means the service is not listening or a firewall, address, or port differs from the configuration.

Confirm the model can answer through Ollama before opening Minecraft. Then launch a clean profile and test a short prompt. Watch RAM, CPU or GPU load, and response time. A model that is too large can appear silent while it loads or runs slowly. Choose a smaller supported model before treating latency as file corruption. Never install an unknown executable that claims it can "unlock" local Verity AI.

Avoid false fixes

Actions that usually waste time

Download

Replacing the jar for 429

A rate limit belongs to the provider account and time window. Reinstalling the same release cannot reset it.

Keys

Publishing a credential

Support only needs a sanitized error. Revoke a key immediately if it appears in a screenshot or public post.

Voice

Changing AI when text works

A text response proves the model path works. Check TTS, output device, language, and voice dependencies next.

Retries

Spamming prompts during a limit

Repeated retries do not repair quota. Wait for the provider reset and retest with a single short prompt.

Support-ready result

Know when the connection is actually fixed

Pass text before voice

A clean verification has a simple sequence. Start Minecraft with the intended profile. Confirm the Verity entry and world behavior. Send one short text question. Wait for a complete text answer. Send a second question only after the first finishes. If both responses arrive without an error, the AI connection is working at that moment. You can then test speech recognition and text-to-speech separately.

Run the same two-prompt test after changing a key, provider, or local model. Changing one variable makes the result understandable. If the error returns, record the new code and provider dashboard state. A repeatable failure with sanitized evidence is suitable for the maintainer support channel; a vague statement that Verity is broken is not.

When reinstalling is reasonable

Reinstall only when the file is missing, corrupted, from the wrong edition, or incompatible with the profile. Use the checked maintainer record, confirm the filename and Minecraft version, and test without unrelated mods. Reinstalling is also reasonable after a failed partial download, but not as a response to a provider status code.

If the API succeeds outside Minecraft but the same valid configuration consistently fails inside a clean profile, collect the launcher log, exact Verity JE version, Forge version, provider, model, and sanitized error. That is the point where a mod configuration or compatibility report is more useful than another account or another file mirror.

FAQ

Verity Mod API connection questions

Why does Verity Mod say API connection failed?

The selected provider did not return a usable response. Read the status code and provider message. A 401 points to authentication, 403 to permission, 404 to a model or endpoint, 429 to rate limits, and 5xx to a service-side problem. No-code failures need a network, config, Ollama service, or log check.

How do I fix status 429?

Stop sending prompts, inspect the provider limits, identify the exhausted request, token, or audio window, wait for its reset, and test one short text prompt. For Groq, limits apply at the organization level, so a second key in the same organization may share the same ceiling.

Does API connection failed mean the Verity file is broken?

Usually not. The error normally appears after the mod has loaded far enough to call its AI provider. Check the provider path first. Revisit the file only if Minecraft cannot load the mod or the file does not match the game version and loader.

Can I share my API key with someone fixing the mod?

No. Share the provider, model, status code, timestamp, and sanitized error text. Revoke any key that was pasted into a public chat, support post, screenshot, stream, or form.