Godot LLM

Isn't it cool to utilize large language model (LLM) to generate contents for your game? LLM has great potential in NPC models, game mechanics and design assisting. Thanks for technology like llama.cpp, "small" LLM, such as llama-3-8B, run reasonably well locally on lower-end machine without a good GPU. I want to experiment LLM in Godot but I couldn't find any good library, so I decided to create one here.

⚠ While LLM is less controversial than image generation models, there can still be legal issues when LLM contents are integrated in games, I have created another page to document some relevant information

Table of Contents

1. Quick start
   • Install
   • Text Generation: GDLlama node
   • Text Embedding: GDEmbedding node
   • Multimodal Text Generation: GDLlava node
   • Vector Database: LlmDB node
   • Template/Demo
2. Retrieval Augmented Generation (RAG)
3. Roadmap
4. Documentation
   • Inspector Properties: GDLlama, GDEmbedding, and GDLlava
   • GDLlama Functions and Signals
   • GDEmbedding Functions and Signals
   • GDLlava Functions and Signals
   • Text generation with Json schema
   • LlmDB Functions and Signals
   • LlmDBMetaData
5. FAQ
6. Compile from Source

Quick Start

Install

1. Get Godot LLM directly from the asset library, or download the vulkan or cpu zip file from the release page, and unzip it to place it in the addons folder in your godot project
2. Now you should be able to see GdLlama, GdEmbedding, GDLlava, and LlmDB nodes in your godot editor. You can add them to a scene in Godot editor, or initialize themm directly by .new().

Text Generation: GDLlama node

1. Download a supported LLM model in GGUF format (recommendation: Meta-Llama-3-8B-Instruct-Q5_K_M.gguf), move the file to somewhere in your godot project
2. Set up your model with GDScript, point model_path to your GGUF file. The default n_predict = -1 generates an infinite sequence, we want it to be shorter here

func _ready():
    var gdllama = GDLlama.new()
    gdllama.model_path = "./models/Meta-Llama-3-8B-Instruct.Q5_K_M.gguf" ##Your model path
    gdllama.n_predict = 20

3. Generate text starting from "Hello"

    var generated_text = gdllama.generate_text_simple("Hello")
    print(generated_text)

4. Text generation is slow, you may want to call gdllama.run_generate_text("Hello", "", "") to run the generation in background, then handle the generate_text_updated or generate_text_finished signals

    gdllama.generate_text_updated.connect(_on_gdllama_updated)
    gdllama.run_generate_text("Hello", "", "")

func _on_gdllama_updated(new_text: String):
    print(new_text)

Text Embedding: GDEmbedding node

1. Download a supported embedding model in GGUF format (recommendation: mxbai-embed-large-v1.Q5_K_M.gguf), move the file to somewhere in your godot project
2. Set up your model with GDScript, point model_path to your GGUF file

func _ready():
    var gdembedding= GDEmbedding.new()
    gdembedding.model_path = "./models/mxbai-embed-large-v1.Q5_K_M.gguf"

3. Compute the embedded vector of "Hello world" in PackedFloat32Array

    var array: PackedFloat32Array = gdembedding.compute_embedding("Hello world")
    print(array)

4. Compute the similarity between "Hello" and "World"

    var similarity: float = gdembedding.similarity_cos_string("Hello", "World")
    print(similarity)

5. Embedding computation can be slow, you may want to call gdembedding.run_compute_embedding("Hello world") or gdembedding.run_similarity_cos_string("Hello", "Worlld") to run the computation in background, then handle the compute_embedding_finished and similarity_cos_string_finished signals

    gdembedding.compute_embedding_finished.connect(_on_embedding_finished)
    gdembedding.run_compute_embedding("Hello world")

func _on_embedding_finished(embedding: PackedFloat32Array):
    print(embedding)

    gdembedding.similarity_cos_string_finished.connect(_on_embedding_finished)
    gdembedding.run_similarity_cos_string("Hello", "Worlld")

func _on_similarity_finished(similarity: float):
    print(similarity)

Note that the current implementation only allows one thread running per node, avoid calling 2 run_* methods consecutively:

    ## Don't do this, this will hang your UI
    gdembedding.run_compute_embedding("Hello world")
    gdembedding.run_similarity_cos_string("Hello", "Worlld")

Instead, always wait for the finished signal or check gdembedding.is_running() before calling a run_* function.

Multimodal Text Generation: GDLlava node

1. Download a supported multimodal model in GGUF format (recommendation: llava-phi-3-mini-int4.gguf), be aware that there are two files needed - a gguf language model and a mmproj model (typical name *mmproj*.gguf), move the files to somewhere in your godot project
2. Set up your model with GDScript, point model_path and mmproj_path to your corresponding GGUF files

func _ready():
    var gdllava = GDLlava.new()
    gdllava.model_path = "./models/llava-phi-3-mini-int4.gguf"
    gdllava.mmproj_path = "./models/llava-phi-3-mini-mmproj-f16.gguf"

3. Load an image (svg, png, or jpg, other format may also works as long as it is supported by Godot), or use your game screen (viewport) as a image

    var image = Image.new()
    image.load("icon.svg")

    ## Or load the game screen instead
    #var image = get_viewport().get_texture().get_image()

4. Generate text to provide "Provide a full description" for the image

    var generated_text = gdllava.generate_text_image("Provide a full description", image)
    print(generated_text)

5. Text generation is slow, you may want to call gdllama.run_generate_text("Hello", "", "") to run the generation in background, then handle the generate_text_updated or generate_text_finished signals

    gdllava.generate_text_updated.connect(_on_gdllava_updated)
    gdllava.run_generate_text_image("Provide a full description", image)

func _on_gdllava_updated(new_text: String):
    print(new_text)

Vector database: LlmDB node

1. LlmDB node extends GDEmbedding node, follow the previous section to download a model and set up the model_path

func _ready():
	var db = LlmDB.new()
	db.model_path = "./models/mxbai-embed-large-v1.Q5_K_M.gguf"

2. Open a database, which creates a llm.db file and connect to it by default

	db.open_db()

3. Set up the structure of the metadata of your textual data, the first metadata should always be an id field with String as the data type, here we use the LlmDBMetaData.create_text, LlmDBMetaData.create_int, and LlmDBMetaData.create_real functions to define the structure of metadata with the corresponding data type.

	db.meta = [
		LlmDBMetaData.create_text("id"),
		LlmDBMetaData.create_int("year"),
        LlmDBMetaData.create_real("attack")
	]

4. Different models create embedding vectors of different sizes, calibrate the embedding_size property before creating tables

    db.calibrate_embedding_size()

5. Create tables based on the metadata, By default, these table are created:

• llm_table_meta: which store the metadata for a particular id
• llm_table: store texts with metadata and embedding
• Some tables with names containing llm_table_virtual: tables for embedding similarity computation

Note that your .meta property should always match the metadata columns in the database before any storing or retrieving operation, consider setting your .meta property within the _ready() function or within the inspector.

    db.create_llm_tables()

6. Store a piece of text with metadata dictionary specifying the year, note that you can leave out some of the metadata if it is not relevant to the text. If the input text is longer than chunk_size, the function will automatically break it down into smaller pieces to fit in the chunk_size.

    var text = "Godot is financially supported by the Godot Foundation, a non-profit organization formed on August 23rd, 2022 via the KVK (number 87351919) in the Netherlands. The Godot Foundation is responsible for managing donations made to Godot and ensuring that such donations are used to enhance Godot. The Godot Foundation is a legally independent organization and does not own Godot. In the past, the Godot existed as a member project of the Software Freedom Conservancy."

    db.store_text_by_meta({"year": 2024}, text)

7. Retrieve 3 of the most similar text chunks to godot where the year is 2024:

    print(db.retrieve_similar_texts("godot", "year=2024", 3))

8. Depending on the embedding model, storing and retrieving can be slow, consider using the run_store_text_by_meta function, run_retrieve_similar_texts function, and the retrieve_similar_text_finished signal to store and retrieve texts in background. Also, call close_db() when the database is no longer in use.

Template/Demo

The godot-llm-template provides a rather complete demonstration on different functionalities of this plugin

Retrieval-Augmented Generation (RAG)

This plugin now has all the essentaial components for simple Retrieval-Augmented Generation (RAG). You can store information about your game world or your character into the vector database, retrieve relevant texts to enrich your prompt, then generate text for your game, the generated text can be stored back to the vector database to enrich future prompt. RAG complement the shortcoming of LLM - the limited context size force the model to forget earlier information, and with RAG, information can be stored in a database to become long-term memory, and only relevant information are retrieve to enrich the prompt to keep the prompt within the context size.

To get started, you may try the following format for your prompt input:

Document:
{retrieved text}

Question:
{your prompt}

Roadmap

Features

• Platform (backend): windows (cpu, vulkan), macOS(cpu, metal), Linux (cpu, vulkan), Android (cpu)
  • macOS support is on best effort basis since I don't have a mac myself
• llama.cpp-based features
  • Text generation
  • Embedding
  • Multimodal
• Vector database integration based on sqlite and sqlite-vec
  • Split text to chunks
  • Store text embedding
  • Associate metadata with text
  • Retrieve text by embedding similarity and sql constraints on metadata

TODO

• iOS: build should be trivial, but an apple developer ID is needed to run thhe build
• Add in-editor documentation, waiting for proper support in Godot 4.3
• Add utility functions to generate useful prompts, such as llama guard 2
• Download models directly from huggingface
• Automatically generate json schema from data classes in GDSCript
• More llama.cpp features
• mlc-llm integration
• Any suggestion?

Documentation

Inspector Properties: GDLlama, GDEmbedding, and GDLlava

There are 3 base nodes added by this plugin: GdLlama, GdEmbedding, and GdLlava. Each type of node owns a set of properties which affect the computational performance and the generated output. Some of the properties belong to more than one node, and they generally have similar meaning for all types of node.

• Model Path: location of your GGUF model
• Mmproj Path location of your mmproj GGUF file, for GdLlava only
• Instruct: question and answer interactive mode
• Interactive: custom interactive mode, you should set your reverse_prompt, input_prefix, and input_suffix to set up a smooth interaction
• Reverse Prompt: AI stops to wait for user input after seeing this prompt being generated, a good example is "User:"
• Input Prefix: append before every user input
• Input Suffix: append after every user input
• Should Output prompt: whether the input prompt should be included in the output
• Should Output Special: whether the special (e.g., beginning of sequence and ending of sequence) token should be included in the output
• Context Size: number of tokens the model can process at a time
• N Predict: number of new tokens to generate, generate infinite sequence if -1
• N Keep: when the model run out of context size, it starts to forget about earlier context, set this variable to force the model to keep a number of the earliest tokens to keep the conversation relevant
• Temperature: the higher the temperature, the more random the generated text
• Penalty Repeat: penalize repeated sequence, diabled if -1
• Penalty Last N: the number of latest token to consider when penalizing repeated sequence, disabled if 0, Context Size if -1
• Penalilze Nl: penallize newline token
• Top K: only sample from this amount of tokens with the highest probabilities, disabled if 0
• Top P: only sample from tokens within this cumulative probability, disabledd if 1.0
• Min P: only sample from tokens with at least this probability, disabledd if 0.0
• N Thread: number of cpu threads to use
• N GPU Layer: number of layer offloaded to GPU
• Main GPU: the main GPU for computation
• Split Mode: how the computation will be distributed if there are multiple GPU in your systemm (0: None, 1: Layer, 2: Row)
• Escape: process escape character in input prompt
• N Batch: maximum number of tokens per iteration during continuous batching
• N Ubatch: maximum batch size for computation

GdLlama functions and signals

Functions

• generate_text_simple(prompt: String) -> String: generate text from prompt
• generate_text_json(prompt: String, json: String) -> String: generate text in a format enforced by a json schema, see the following section
• generate_text_grammar(prompt: String, grammar: String) -> String: generate text in a format enforced by GBNF grammar
• generate_text(prompt: String, grammar: String, json: String) -> String: a wrapper function, run generate_text_gramma if grammar is non-empty, runs generate_text_json if json is non-empty, run generate_text_simple otherwise
• run_generate_text(prompt: String, grammar: String, json: String) -> Error: run generate_text in background, rely on signals to recieve generated text, note that only one background thread is allowd for a GDLlama node, calling this function when the background thread is still running will freeze the logic until the background thread is done
• input_text(input: String): input text to interactively generate text (with either Instruct or Interactive enabled) with the model, only works if the model is waiting for intput, inputing an empty string means the model should continue to generate what it has been generating
• stop_generate_text(): stop text generation, clean up the model and the background thread
• is_running() -> bool: whether the background thread is running
• is_waiting_input() -> bool: whether the model is waiting for input text (with either Instruct or Interactive enabled)

Signals

• generate_text_finished(text: String) : emitted with the full generated text when a text generation is completed. When either Instruct or Interactive enabled, this signal is emitted after the whole interaction is finished
• generate_text_updated(new_text: String): instead of waiting the full generated text, this signal is emited whenever a new token (part of the text sequence) is generated, which forms a stream of strings
• input_wait_started(): the model is now starting to wait for user input, happens when either Instruct or Interactive are enabled and the model stop generating text in the middle of the conversation to wait for further input from the user.

GDEmbedding functions and signals

Functions

• compute_embedding(prompt: String) -> PackedFloat32Array: compute the embedding vector of a prompt
• similarity_cos_array(array1: PackedFloat32Array, array2: PackedFloat32Array) -> float: compute the cosine similarity between two embedding vectors, this is a fast function, no model is loaded
• similarity_cos_string(s1: String, s2: String) -> float: compute the cosine similarity between two strings
• run_compute_embedding(prompt: String) -> Error: run compute_embedding(prompt: String) in background, rely on the compute_embedding_finished signal to recieve the embedding vector, note that only one background thread is allowd for a GDEmbedding node, calling this function when the background thread is still running will freeze the logic until the background thread is done
• run_similarity_cos_string(s1: String, s2: String) -> Error: run similarity_cos_string in background, rely on the compute_similairty_finished signal to recieve the cosine similairty, note that only one background thread is allowd for a GDEmbedding node, calling this function when the background thread is still running will freeze the logic until the background thread is done
• is_running() -> bool: whether the background thread is running

Signals

• compute_embedding_finished(embedding: PackedFloat32Array): emitted when run_compute_embedding is completed
• similarity_cos_string_finished(similarity: float): emitted when run_similarity_cos_string is completed

GDLlava functions and signals

functions

• generate_text_base64(prompt: String, image_base64: String) -> String: generate text based on a prompt and a base64 string which encodes a jpg or png image
• generate_text_image(prompt: String, image: Image) -> String: generate text based on a prompt and an Image object in Godot
• run_generate_text_base64(prompt: String, image_base64: String) -> Error: run generate_text_base64 in background, rely on signals to recieve generated text, note that only one background thread is allowd for a GDLlava node, calling this function when the background thread is still running will freeze the logic until the background thread is done
• run_generate_text_base64(prompt: String, image: Image) -> Error: run generate_text_base64 in background, rely on signals to recieve generated text, note that only one background thread is allowd for a GDLlava node, calling this function when the background thread is still running will freeze the logic until the background thread is done
• stop_generate_text(): stop text generation, clean up the model and the background thread
• is_running() -> bool: whether the background thread is running

Signals

• generate_text_finished(text: String) : emitted with the full generated text when a text generation is completed
• generate_text_updated(new_text: String): instead of waiting the full generated text, this signal is emited whenever a new token (part of the text sequence) is generated, which forms a stream of strings

Text generation with Json schema

Suppose you want to generate a character with:

• name: a string from 3 character to 20 character
• birthday: a string with a specific date format
• weapon: either "sword", "bow", or "wand
• description: a text with minimum 10 character

You should first create a GDLlama node, and turn Should Output prompt and Should Output Special off either by inspector or by script:

should_output_prompt = false
should_output_special = false

Construct the following _person_schema dictionary in GDScript:

var _person_schema = {
	"type": "object",
	"properties": {
		"name": {
			"type": "string",
			"minLength": 3,
			"maxLength": 20,
		},
		"birthday": {
			"type": "string",
			"format": "date"
		},
		"weapon": {
			 "enum": ["sword", "bow", "wand"],
		},
		"description": {
			"type": "string",
			"minLength": 10,
		},
	},
	"required": ["name", "birthday", "weapon", "description"]
}

Then convert it to a json string

var person_schema: String = JSON.stringify(_person_schema)

Supposed you are interested in a "Main character in a magic world", you can generate the character using the generate_text_json(prompt, json_scheme) of the GDLlama node:

var json_string: String = generate_text_json(prompt, json_scheme)

Note that text generation is slow, you may want to use run_generate_text(prompt, "", json_scheme) to run the generation in background, then handle generate_text_finished to receive the generated text.

json_string should look like this:

{"birthday": "2000-05-12", "description": "A young wizard with a pure heart and a mischievous grin. He has a wild imagination and a love for adventure. He is always up for a challenge and is not afraid to take risks.", "name": "Eryndor Thorne", "weapon": "wand"}

Now, the generated data is ready, you can parse back to a dictionary or other object to use the data.

var dict: Dictionary = {}
var json = JSON.new()
var error = json.parse(json_string)
if (error == OK):
		dict = json.data

print(dict["name"]) ##Eryndor Thorne

Inspector properties: LlmDB

LlmDB extends GDEmbedding and shares all its properties, check the section above for the relevant information. Additionally, LlmDB has

• Meta: an array of LlmDBMetaData Resource which defines the structure of the metadata. LlmDBMetaData contains Data Name which define the name of a metadata, and Data Type (0=integer, 1=real, 2=text, 3=blob) to define the data type of the metadata. Meta should be non-empty, and the first element of Meta should always be an id with text as the Data Type.
• dB Dir: the directory of the database file, default is the root directory of the project
• dB File: the file name of the database file, default is llm.db
• Table Name: defines the name of the tables created by the create_llm_tables function
• Embedding Size: the vector size of the embedding computed by the model, used in the create_llm_tables function
• Absolute Separators: an array of String. When storing a piece of text, the text will be first separated by the String defines here, the separation process will stop if the separated text is shorter than Chunk Size or all the separators here have been processed. The default are \n and \n\n, which are displayed as empty space in the inspector.
• Chunk Separators: an array of String. After the Absolute Separators are processed, one of the separators (first one that works) here will be chosen to further separated the piece of texts, then the pieces are grouped up to chunks to fulfill the requirements of Chunk Size and Chunk Overlap
• Chunk Size: any text chunk should not exceed this size, unless the separation function fails to fulfill the requirement after iteratoring through the iterators
• Chunk Overlap: the maximum overlap between neighbouring text chunks, the algorithm will try to create the biggest overlap possible fulfilling this constraint

LlmDB Functions and Signals

Besides the functions and signals from GDEmbedding, LlmDB has a few more functions and signals

Functions

• calibrate_embedding_size(): calibrate Embedding Size to the correct number based on the model in model_path
• open_db(): create a dB_File at dB_Dir if the file doesn't exist, then connect to the database
• close_db(): terminate the connection to the database
• execute(statement: String) execute an sql statement, turn on Verbose stdout in Project Settings to see the log generated by this statement
• create_llm_tables(): create a table with name Table Name if the table doesn't exist, a Table Name + _meta table to store pre-defined metadata by id, and some _virtual tables to
• drop_table(p_table_name: String): drop a table with a specific name
• drop_llm_tables(p_table_name: String): drop all tables (except the sqlite_sequence table which is created automatically for autoincrement) created by create_llm_tables(), i.e., p_table_name, p_table_name + _meta and every table with a name containing p_table_name + _virtual
• has_table(p_table_name: String) -> bool: whether a table with this name exists
• is_table_valid(p_table_name: String) -> bool: whether the table contains valid metadata, i.e., all elements in .meta properties exist in the table and the data types are correct
• store_meta(meta_dict: Dictionary): store a set of meta data to table Table Name + _meta with id as the primary key, such that you can call store_text_by_id by id instead of inputting the full metadata dictionary through store_text_by_meta
• has_id(id: String, p_table_name: String) -> bool: whether the table has a specific id stored
• split_text(text: String) -> PackedStringArray: split a piece of text first by all Absolute Separators, then by one of the appropiate Chunk Separators, such that any text chunk is shorter than Chunk Size (measured in character), and the overlap is close to but not greater than Chunk Overlap. If the algorithm failed to satisfy the contraints, there will be an error message printed out and the returned chunk will be greater than the Chunk Size
• store_text_by_id(id: String, text: String): split the text and store the chunks in the database, be aware that store_meta should have been called previously such that the id with the corresponding meta is already in the database
• run_store_text_by_id(id: String, text: String) -> Error: run store_text_by_id in background, emits store_text_finished signal when finished
• store_text_by_meta(meta_dict: Dictionary, text: String): split the text and store the chunks in the database with the metadata defined in meta_dict, be aware that the metadata should be valid, every key should be a name stored in the .meta property and the corresponding type should be correct
• run_store_text_by_meta(meta_dict: Dictionary, text: String) -> Error run store_text_by_meta in background, emits store_text_finished signal when finished
• retrieve_similar_texts(text: String, where: String, n_results: int) -> PackedStringArray: retrieve n_results most similar text chunks to text, where should be empty or an sql WHERE clause to filter the chunks by metadata
• run_retrieve_similar_texts(text: String, where: String, n_results: int) -> Error: run retrieve_similar_texts in background, and emits a retrieve_similar_texts_finished signal once it is done

Signals

• store_text_finished: emitted when run_store_text_by_id or run_store_text_by_meta is finished
• retrieve_similar_texts_finished(array: PackedStringArray): contains an array of String, emitted when run_retrieve_similar_texts is finished

LlmDBMetaData

This is a simple resource class that forms the meta array property in LlmDB. It has two properties:

• data_name: a String that defines the name of this metadata
• data_type: an int that defines the data type of this metadata (0=integer, 1=real, 2=text, 3=blob), note that inputing an integer here is not recommended since it can be confusing, use the inspector properties, the LlmDBMetaData enum or the function below instead
• LlmDBMetaDataType enum:
  • LlmDBMetaData.INTEGER = 0
  • LlmDBMetaData.REAL = 1
  • LlmDBMetaData.TEXT = 2
  • LlmDBMetaData.BLOB = 3

There are 4 static functions to create LlmDBMetaData

• create_int(data_name: String) -> LlmDBMetaData: create a LlmDBMetaData with type int (0)
• create_real(data_name: String) -> LlmDBMetaData: create a LlmDBMetaData with type real (1)
• create_text(data_name: String) -> LlmDBMetaData: create a LlmDBMetaData with type text (2)
• create_blob(data_name: String) -> LlmDBMetaData: create a LlmDBMetaData with type blob (3), note that blob data type support is still a work-in-progress

Alternatively, you can use this static function to create LlmDBMetaData

• create(data_name: String, data_type: int) -> LlmDBMetaData: create a corresponding LlmDBMetaData by data_name and data_type, it is recommended to use the enum instead of int for data_type

FAQ

1. How to get more debug message?

Turn on Verbose stdout in Project Settings, consider running Godot from a terminal to get additional logging messages.

2. Does it support languages other than English?

Yes, the plugin uses utf8 encoding so it has multilingual support naturally. However, a language model may be trained with English data only and it won't be able to generate text other than English, choose the language model based on your need.

3. Strange tokens in generated text, such as <eot_id> when Should Output Special is off.

You are always welcome to open an issue. However, be aware that the standard of GGUF format can be changed to support new features and models, such that the bug can come from the model side instead of within this plugin. For example, some older llama 3 GGUF model may not be compatible with the latest format, you may try to search for a newer model with fixes such as this.

4. You are running Arch linux (or its derivatives such as Manjaro) and your Godot Editor crash.

The Arch build of Godot is bugged when working with GDExtension, download Godot from the official website instead.

5. You have a discrete GPU and you see unable to load model error, you have make sure that the model parameters are correctly set.

There is currently a bug on vulkan backend if you have multiple drivers installed for the same GPU, try to turn Split Mode to NONE (0) and set your Main GPU manually (starting from 0) to see if it works.

Compile from source

Install build tools and Vulkan SDK for your operating system, then clone this repository

git clone https://github.com/Adriankhl/godot-llm.git
cd godot-llm
git submodule update --init --recursive
mkdir build
cd build

Run cmake.

On Windows:

cmake .. -GNinja -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DLLAMA_NATIVE=OFF -DLLAMA_VULKAN=ON -DCMAKE_EXPORT_COMPILE_COMMANDS=1 -DCMAKE_BUILD_TYPE=Release

On Linux:

cmake .. -GNinja -DLLAMA_NATIVE=OFF -DCMAKE_EXPORT_COMPILE_COMMANDS=1 -DLLAMA_VULKAN=ON -DCMAKE_BUILD_TYPE=Release

Vulkan build works for Windows and Linux, if you want a cpu build, set -DLLAMA_VULKAN=OFF instead.

For Android, set $NDK_PATH to your android ndk directory, then:

cmake .. -GNinja -DCMAKE_TOOLCHAIN_FILE=$NDK_PATH\cmake\android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_C_FLAGS="-mcpu=generic" -DCMAKE_CXX_FLAGS="-mcpu=generic" -DCMAKE_BUILD_TYPE=Release

You may want to adjust the compile flags for Android to suit different types of CPU.

Then compile and install by ninja:

ninja -j4
ninja install

The folder ../install/gpu/addons/godot_llm (cpu instead of gpu for cpu build) can be copy directly to the addons folder of your godot project.