diff --git a/src/llm/README.md b/src/llm/README.md
new file mode 100644
index 00000000..8d75d02e
--- /dev/null
+++ b/src/llm/README.md
@@ -0,0 +1,7 @@
+# Local LLM Integration
+1. Clone and build [llama.cpp](https://github.com/ggerganov/llama.cpp) on the same machine where you will run your uqbar node
+  - follow their README for details on how to do this. In most cases simply running `make` works
+  - make sure to get your model as a .gguf file
+2. Within the llama.cpp directory, run this command in llama.cpp on the same machine you will run your uqbar node: `./server --port <PORT>`
+  - Note: you can pass in whatever other command line arguments to the llama cpp server you want depending on your preferences/hardware/model/etc.
+3. Run your Uqbar node with `--features llm` and `--llm http://localhost:<PORT>`. For example `cargo +nightly run --features llm --release home --rpc wss://eth-sepolia.g.alchemy.com/v2/<YOUR_API_KEY> --llm http://localhost:<PORT>`
diff --git a/src/llm.rs b/src/llm/mod.rs
similarity index 99%
rename from src/llm.rs
rename to src/llm/mod.rs
index 7417164e..ce53d802 100644
--- a/src/llm.rs
+++ b/src/llm/mod.rs
@@ -1,7 +1,10 @@
 use crate::types::*;
+use crate::llm::types::*;
 use anyhow::Result;
 use reqwest::Response as ReqwestResponse;
 
+mod types;
+
 pub async fn llm(
     our_name: String,
     send_to_loop: MessageSender,
diff --git a/src/llm/types.rs b/src/llm/types.rs
new file mode 100644
index 00000000..3fd2be01
--- /dev/null
+++ b/src/llm/types.rs
@@ -0,0 +1,86 @@
+use serde::{Deserialize, Serialize};
+use thiserror::Error;
+
+#[derive(Debug, Serialize, Deserialize)]
+pub struct LlmPrompt {
+    pub prompt: String,
+    pub n_predict: u64,
+}
+
+#[derive(Clone, Serialize, Deserialize, Debug)]
+pub struct LlmResponse {
+    pub content: String,
+    pub generation_settings: GenerationSettings,
+    pub model: String,
+    pub prompt: String,
+    pub slot_id: u64,
+    pub stop: bool,
+    pub stopped_eos: bool,
+    pub stopped_limit: bool,
+    pub stopped_word: bool,
+    pub stopping_word: String,
+    pub timings: Timings,
+    pub tokens_cached: u64,
+    pub tokens_evaluated: u64,
+    pub tokens_predicted: u64,
+    pub truncated: bool,
+}
+
+#[derive(Clone, Serialize, Deserialize, Debug)]
+pub struct GenerationSettings {
+    pub frequency_penalty: f64,
+    pub grammar: String,
+    pub ignore_eos: bool,
+    pub logit_bias: Vec<serde_json::Value>, // This should be changed to the appropriate type
+    pub mirostat: u64,
+    pub mirostat_eta: f64,
+    pub mirostat_tau: f64,
+    pub model: String,
+    pub n_ctx: u64,
+    pub n_keep: u64,
+    pub n_predict: u64,
+    pub n_probs: u64,
+    pub penalize_nl: bool,
+    pub presence_penalty: f64,
+    pub repeat_last_n: u64,
+    pub repeat_penalty: f64,
+    pub seed: u64,
+    pub stop: Vec<serde_json::Value>, // This should be changed to the appropriate type
+    pub stream: bool,
+    pub temp: f64,
+    pub tfs_z: f64,
+    pub top_k: u64,
+    pub top_p: f64,
+    pub typical_p: f64,
+}
+
+#[derive(Clone, Serialize, Deserialize, Debug)]
+pub struct Timings {
+    pub predicted_ms: f64,
+    pub predicted_n: u64,
+    pub predicted_per_second: f64,
+    pub predicted_per_token_ms: f64,
+    pub prompt_ms: f64,
+    pub prompt_n: u64,
+    pub prompt_per_second: f64,
+    pub prompt_per_token_ms: f64,
+}
+
+#[derive(Error, Debug, Serialize, Deserialize)]
+pub enum LlmError {
+    #[error("llm: rsvp is None but message is expecting response")]
+    BadRsvp,
+    #[error("llm: no json in request")]
+    NoJson,
+    #[error(
+        "llm: JSON payload could not be parsed to LlmPrompt: {error}. Got {:?}.",
+        json
+    )]
+    BadJson { json: String, error: String },
+    #[error("llm: http method not supported: {:?}", method)]
+    BadMethod { method: String },
+    #[error("llm: failed to execute request {:?}", error)]
+    RequestFailed { error: String },
+    #[error("llm: failed to deserialize response {:?}", error)]
+    DeserializationToLlmResponseFailed { error: String },
+}
diff --git a/src/types.rs b/src/types.rs
index 8a6d2b86..7fe2018e 100644
--- a/src/types.rs
+++ b/src/types.rs
@@ -1060,90 +1060,3 @@ pub enum EncryptorMessage {
     Encrypt(EncryptAction),
 }
 // encryptor End
-
-// llm.rs Start
-#[derive(Debug, Serialize, Deserialize)]
-pub struct LlmPrompt {
-    pub prompt: String,
-    pub n_predict: u64,
-}
-
-#[derive(Clone, Serialize, Deserialize, Debug)]
-pub struct LlmResponse {
-    pub content: String,
-    pub generation_settings: GenerationSettings,
-    pub model: String,
-    pub prompt: String,
-    pub slot_id: u64,
-    pub stop: bool,
-    pub stopped_eos: bool,
-    pub stopped_limit: bool,
-    pub stopped_word: bool,
-    pub stopping_word: String,
-    pub timings: Timings,
-    pub tokens_cached: u64,
-    pub tokens_evaluated: u64,
-    pub tokens_predicted: u64,
-    pub truncated: bool,
-}
-
-#[derive(Clone, Serialize, Deserialize, Debug)]
-pub struct GenerationSettings {
-    pub frequency_penalty: f64,
-    pub grammar: String,
-    pub ignore_eos: bool,
-    pub logit_bias: Vec<serde_json::Value>, // This should be changed to the appropriate type
-    pub mirostat: u64,
-    pub mirostat_eta: f64,
-    pub mirostat_tau: f64,
-    pub model: String,
-    pub n_ctx: u64,
-    pub n_keep: u64,
-    pub n_predict: u64,
-    pub n_probs: u64,
-    pub penalize_nl: bool,
-    pub presence_penalty: f64,
-    pub repeat_last_n: u64,
-    pub repeat_penalty: f64,
-    pub seed: u64,
-    pub stop: Vec<serde_json::Value>, // This should be changed to the appropriate type
-    pub stream: bool,
-    pub temp: f64,
-    pub tfs_z: f64,
-    pub top_k: u64,
-    pub top_p: f64,
-    pub typical_p: f64,
-}
-
-#[derive(Clone, Serialize, Deserialize, Debug)]
-pub struct Timings {
-    pub predicted_ms: f64,
-    pub predicted_n: u64,
-    pub predicted_per_second: f64,
-    pub predicted_per_token_ms: f64,
-    pub prompt_ms: f64,
-    pub prompt_n: u64,
-    pub prompt_per_second: f64,
-    pub prompt_per_token_ms: f64,
-}
-
-#[derive(Error, Debug, Serialize, Deserialize)]
-pub enum LlmError {
-    #[error("llm: rsvp is None but message is expecting response")]
-    BadRsvp,
-    #[error("llm: no json in request")]
-    NoJson,
-    #[error(
-        "llm: JSON payload could not be parsed to LlmPrompt: {error}. Got {:?}.",
-        json
-    )]
-    BadJson { json: String, error: String },
-    #[error("llm: http method not supported: {:?}", method)]
-    BadMethod { method: String },
-    #[error("llm: failed to execute request {:?}", error)]
-    RequestFailed { error: String },
-    #[error("llm: failed to deserialize response {:?}", error)]
-    DeserializationToLlmResponseFailed { error: String },
-}
-
-// llm End