| command | action |
|---|---|
<message> |
send an in-game message |
/command |
execute a minecraft command |
:help |
show neuroflayer commands |
:status |
show neuroflayer status |
:llm <on/off> |
toggle replies |
:memory <on/off> |
toggle memory. "off" clears history |
:forget |
clear conversation history and cancel pending replies |
:ask <question> |
ask the LLM; answers in the console |
:parser [index|name] |
show available message parsers | select a specific one |
:respawn |
respawn after death |
:quit |
disconnect and end session |
| variable | default | description |
|---|---|---|
MINECRAFT_HOST |
required | minecraft server address |
MINECRAFT_PORT |
25565 |
minecraft server port |
MINECRAFT_USERNAME |
required | username for offline-mode login |
MINECRAFT_VERSION |
auto | minecraft version used by mineflayer |
MINECRAFT_LOG_MESSAGES |
true |
log minecraft messages to the console |
LLM_URL |
required | API endpoint |
LLM_MODEL |
required | model used to generate replies |
LLM_PROMPT_PATH |
./prompt |
path to the instruction prompt |
LLM_TIMEOUT_MS |
45000 |
maximum time spent waiting for an LLM response |
LLM_MAX_TOKENS |
512 |
max tokens generated per answer |
LLM_TEMPERATURE |
0.9 |
sampling temperature from 0 to 2; lower values make replies more predictable, higher values make them more varied |
MAX_MESSAGE_LENGTH |
256 |
character limit per outgoing message |
MAX_PENDING_MESSAGES |
8 |
maximum questions waiting in the queue |
QUEUE_TTL_SECONDS |
120 |
seconds before a queued request expires |
MEMORY_MAX_MESSAGES |
30 |
maximum messages used as context |
MEMORY_TTL_SECONDS |
600 |
seconds of inactivity before memory clears |
requires Node.js 22+ and a running LLM server supporting the grammar parameter, such as llama.cpp. from the neuroflayer directory, run the following:
npm install
cp -n .env.example .env
touch promptfill in the required .env values and modify prompt for your needs, then:
npm start