SYMPTOM-FIRST TROUBLESHOOTING
Verity Mod Not Working? Start With the Symptom
A crash, unanswered chat, and a silent microphone are different problems. Identify your edition, project, and last successful step, then test one change at a time. Reinstalling a file cannot fix every provider outage or unsupported feature.
FIRST CHECK
Find the failing layer before changing settings
01Record your Minecraft version, filename, and creator. Java's Verity JE, PnTMC's Verity — Bedrock Edition, and Utlover65's Verity BE (Official) have separate procedures. Do not substitute another project's commands.
Back up important worlds. Use a disposable world or separate launcher instance and keep working files. Change one variable, repeat the same test, and record the result; changing several settings together hides the cause.
| Symptom | Start here |
|---|---|
| Java closes before the title screen | Check loader, game version, dependencies, and the first log error. |
| The world loads but AI chat fails | Check the configured provider or local service and the exact error. |
| Typed replies work but voice does not | Test microphone input and speech output separately. |
| PnTMC Bedrock loses its connection | Check the platform helper and reconnect inside the world. |
| Utlover65 Bedrock shows only three dots | Check its documented backend connection and service status. |
JAVA DOES NOT LAUNCH
Separate a loader failure from an AI failure
02JE currently lists Forge builds for Minecraft 1.20.1 and an older NeoForge build for 1.21.1. Fabric is not listed. Match the individual release record; renaming a JAR does not change its loader.
Check that GeckoLib matches your Minecraft version and loader. If the loader reports another missing dependency, such as YetAnotherConfigLib, follow that release's requirements. Do not install arbitrary dependency versions from an unrelated profile.
- Open the selected launcher's game directory. Keep only one Verity JE JAR; do not combine stable, beta, and legacy builds.
- Test the matching loader, Verity, and documented dependencies in a clean instance. If it works, restore other mods in small groups.
- Read latest.log and the crash report. Find the first missing dependency, incompatible version, or class-loading message, not just the final exit code.
- Save the error. If the clean instance still fails, report that reproduction; changing unrelated AI settings will not explain a loader failure.
JE publisher and release records ↗ · JE dependency records ↗ · Installation checks →
THE GAME LOADS, CHAT DOES NOT
Use the actual AI error, not a reinstall loop
03A title screen does not verify AI. Follow your installed JE branch's tutorial; 6.0+ instructions are not universal. Compare provider, model, endpoint, and credential privately.
The maintainer describes these codes as diagnostic categories, not guaranteed explanations. Record the provider and redacted error. Without a code, note whether the request times out, is rejected, or never starts.
| Code | What to check | Next action |
|---|---|---|
| 401 | Invalid, expired, or missing key | Check the key in the correct provider setting; never post it publicly. |
| 403 | Required permissions or regional availability | Check provider access rules. Do not treat this as a Minecraft loader error. |
| 429 | Request rate limit reached | Pause repeated requests and check limits; consider a documented alternative setup. |
| 503 / 529 | Service unavailable | Check provider status and retry later, or configure another supported provider. |
Maintainer's JE 6.0+ setup and error-code guide ↗ · AI setup by Java branch →
LOCAL AI
Distinguish a cold Ollama request from a broken connection
04Test Ollama before Minecraft. Model loading affects response time, so a slow first reply alone does not prove Verity is misconfigured. Compare a cold request with the next request.
Ollama normally binds to 127.0.0.1 on port 11434. Keep same-computer tests local. Do not open router ports or expose the service to the internet as a first-line fix.
- Confirm Ollama is running. Compare the installed model's exact name and tag with your setup.
- Send one short prompt directly through Ollama. If it fails, investigate that error first.
- After a successful reply, repeat the request in Minecraft. Do not stack requests during loading.
- Use ollama ps to inspect the loaded model and CPU/GPU allocation. Record whether only the first reply or every reply is slow.
ollama ls
ollama ps Ollama loading, memory, and local connection FAQ ↗ · Ollama command reference ↗
TEXT WORKS, AUDIO FAILS
Test typed chat, microphone input, and speech separately
05Try one typed message using your project's interaction method. If a reply arrives, preserve the working AI settings and investigate audio separately.
Check microphone permission and the selected input device; test it outside Minecraft. For speech output, check the output device and your release's speech settings. Report whether your words are received without spoken output, or not received at all.
JE supports multilingual AI text, but its publisher tested speech-to-text and text-to-speech only in English. Include a short English test when reporting speech problems.
BEDROCK: PNTMC ONLY
Check the 4.0 helper and reconnect after entering the world
06PnTMC's 4.0 setup uses a platform-specific helper on the same device. Importing the .mcaddon alone is insufficient. Obtain helper files through the maintainer's guide.
The guide requires WebSockets, Beta APIs, and Cheats. Review its platform connection settings deliberately; do not disable security tools or broaden network access casually.
- Check that your platform helper is running. Read its error output if it stopped.
- Confirm the world settings, enter the world, and run the localhost command below. Reconnect after re-entering the world.
- On iOS with aShell, type chat normally. Microphone input and !talk are unsupported; another permission will not add them.
/connect 127.0.0.1:3000 BEDROCK: UTLOVER65 ONLY
Three dots and missing dialogue have different checks
07Utlover65's FAQ links three-dot replies to a missing debugger connection or a backend problem. Get the current command from this project's tutorial or support channel. PnTMC's localhost command is not interchangeable.
A missing dialogue box points to disabled Beta APIs or an outdated add-on. Check world settings and your release. This project requires Cheats; typed messages begin with Verity followed by your message.
If a previously working setup now returns dots, check backend status before importing repeatedly. Report when it last worked and whether a fresh test world behaves identically.
Utlover65's setup, FAQ, and support links ↗ · Compare the two Bedrock projects →
WHEN THE CHECKS DO NOT RESOLVE IT
Send a small, reproducible, redacted support report
08Use the relevant maintainer's support channel. This independent site cannot restore accounts or fix hosted backends. Explain the last successful step and first failed step.
Redact API keys, authorization headers, tokens, private endpoints, usernames, and personal folder details from logs and screenshots. Keep error codes and non-secret context. Revoke or rotate any exposed key through its provider; deleting a post is insufficient.
- Edition, creator, Verity filename/version, Minecraft version, and Java loader if applicable.
- Operating system and exact symptom: crash, chat, slow replies, microphone, or speech output.
- Shortest reproduction; results from a clean instance or direct Ollama test.
- Redacted error wording, latest.log or crash report, and PnTMC helper errors.
- Changes since it worked and the result of each attempted check.
PLAYER QUESTIONS
Frequently asked questions
01 Should I reinstall Verity when AI chat stops working? +
Not automatically. First distinguish a local loading problem from a provider error. Reinstalling the same file does not change a rejected credential or unavailable backend.
02 Can I use any Bedrock connection command? +
No. Confirm the project creator before following connection instructions. PnTMC's local helper and Utlover65's backend use separate setup procedures.
03 Should I share my whole configuration for support? +
Review it first and remove keys, tokens, private endpoints, and personal details. Share the exact error and relevant non-secret settings, not credentials.
04 Does a failed microphone test mean the AI is broken? +
No. Test typed conversation first, then microphone input and speech output separately. An unsupported platform feature cannot be fixed by changing a working AI key.