Use this Verity Mod Ollama setup guide after the Forge mod and its dependencies already reach the Minecraft title screen. It separates Ollama, the local model, Verity JE configuration and voice features so you can identify exactly where a failed response begins.
Version and source check: July 22, 2026
Install Ollama on the same Windows computer that runs Minecraft, pull one model, confirm the local service answers at http://localhost:11434, then select the Ollama path in Verity JE's configuration and enter the exact installed model tag.
Test typed chat before microphone or speech output. If Ollama answers outside Minecraft but Verity does not, the problem is normally the mod configuration, model tag or blocked local connection rather than Forge.
Finish the Minecraft layer first; Ollama cannot repair a broken Forge profile.
Start with a clean Minecraft 1.20.1 Forge instance containing Verity JE 5.7.3, GeckoLib and YACL. Launch once and confirm all three projects appear in Forge's Mods screen. If Minecraft crashes before the title screen, fix the loader, Java runtime or dependency versions before installing a language model.
Ollama runs the model on your own computer, so available memory, storage and CPU or GPU speed affect response time. Choose a smaller model for the first test. A slow response can be normal on modest hardware; a connection refusal means the local service is not reachable.
| Layer | Required check | If it fails |
|---|---|---|
| Minecraft | Forge 1.20.1 reaches title screen | Repair loader or dependencies |
| Verity JE | 5.7.3 appears in Mods list | Verify the JAR and profile |
| Ollama | Local service is running | Restart Ollama |
| Model | Exact tag appears in ollama list | Pull or correct the tag |
| Voice | Text chat works first | Check audio only after AI works |
Install Ollama from its official Windows download and open it once. The default local API uses port 11434. Keep Minecraft and Ollama on the same computer for the simplest setup; localhost refers to the current computer, not another device on your network.
Use a terminal to confirm the application is available. The status request is a service test, not a Verity test. If it fails here, do not change Minecraft files.
ollama --version
A version response confirms the command is available.
ollama list
Copy the model name and tag exactly as shown.
Invoke-RestMethod -Uri http://localhost:11434/api/tags
A JSON response means the local service can answer.
Forge loads Verity JE, GeckoLib and YACL.
The local service loads the selected language model.
The mod sends typed prompts to the local endpoint.
Pull one model before opening the Verity configuration. The command below is an example; use the model recommended by the current maintainer tutorial when it specifies one, and match the installed tag exactly. Larger models usually need more memory and take longer to answer.
Run a short prompt in the terminal. This isolates Ollama from Verity. If the model cannot answer here, reinstalling the mod will not solve the problem.
ollama pull llama3.2
Use a smaller compatible model for the first diagnostic test.
ollama run llama3.2
Enter a short prompt, wait for a reply, then exit the session.
Open the Verity configuration through the mod's current YACL-based settings flow and choose the local Ollama option instead of the online Groq path. Use http://localhost:11434 unless the maintainer's current instructions explicitly require another address. Enter the same model tag shown by ollama list; capitalization and suffixes such as :latest or :8b can matter.
Save the configuration, fully close Minecraft and launch the same Forge profile again while Ollama remains running. Avoid publishing screenshots that include unrelated credentials or private paths. Ollama itself normally does not require a cloud API key for a local-only connection.
Use the current in-game configuration entry provided through YACL.
Choose the local model path rather than the Groq online path.
Use the reachable localhost address and the exact installed model tag.
Keep Ollama running, close Minecraft completely and relaunch the profile.
Create a disposable test world and verify one layer at a time. Start with typed interaction because it removes microphone permissions, speech recognition and output-device selection from the test. A successful text response proves that Minecraft, Verity, the endpoint and the model can complete one round trip.
Then test whether responses remain usable while moving through the world. Only after that should you enable microphone input and spoken output. This sequence prevents an audio problem from being mistaken for an Ollama connection failure.
| Test | Expected result | What it proves |
|---|---|---|
| ollama run | Model replies in terminal | Model is installed and usable |
| API tags | Installed models return as JSON | Local service is reachable |
| Typed Verity chat | Companion returns text | Mod-to-Ollama path works |
| Second prompt | Conversation continues | Model tag and context are stable |
| Microphone and speech | Audio works after text | Voice layer is configured |
Treat the error message as a boundary marker. Connection refused points to Ollama or the endpoint. Model not found points to a tag mismatch. Minecraft crashes point back to Forge, Java or dependencies. Very slow answers usually indicate that the selected model is too large for available hardware.
Do not download random replacement JARs to fix an AI error. The verified Verity JE version remains 5.7.3 on July 22, 2026, and its Modrinth file still returns a Java archive with the published size. Keep the known-good mod profile and change one AI setting at a time.
| Symptom | Likely cause | Next action |
|---|---|---|
| Connection refused | Ollama stopped or wrong address | Open Ollama and retest /api/tags |
| Model not found | Tag does not match ollama list | Copy the exact installed tag |
| Long pause or timeout | Model exceeds available resources | Try a smaller model and close heavy apps |
| Verity loads but never replies | Wrong provider or unsaved config | Select Ollama, save and restart |
| Text works but voice fails | Audio permission or device issue | Keep AI settings and troubleshoot audio |
| Minecraft crashes | Forge, Java or dependency issue | Read latest.log and repair the mod profile |
No. Verity JE presents Groq as an online path and Ollama as a local path. Use one configured provider.
For Ollama on the same Windows computer, start with http://localhost:11434 unless current maintainer instructions say otherwise.
Begin with a smaller model that your computer can run reliably, and use the exact model recommended by the current Verity tutorial when one is specified.
The configured model name usually does not exactly match the tag returned by ollama list.
Yes. Use ollama run and the local /api/tags endpoint before troubleshooting Verity.
Yes. The Modrinth API check on July 22, 2026 still lists 5.7.3 for Forge 1.20.1.
These sources verify the current Verity release path and Ollama's local service behavior.