diff --git a/.gitignore b/.gitignore
index 54dfe61..2d7a42e 100644
--- a/.gitignore
+++ b/.gitignore
@@ -56,3 +56,6 @@ Cargo.lock
 
 backend/target/
 
+.env
+
+.env
diff --git a/backend/.gitignore b/backend/.gitignore
new file mode 100644
index 0000000..4c49bd7
--- /dev/null
+++ b/backend/.gitignore
@@ -0,0 +1 @@
+.env
diff --git a/backend/Cargo.toml b/backend/Cargo.toml
index 8589859..7e278a8 100644
--- a/backend/Cargo.toml
+++ b/backend/Cargo.toml
@@ -6,10 +6,14 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-tokio = { version= "1", features = ["full"] }
+openai_api_rust = "0.1.8"
+diesel = { version = "2.1.0", features = ["postgres"] }
+dotenvy = "0.15"
+uuid = { version = "1.7.0", features = ["v5", "fast-rng", "macro-diagnostics"] }
+tokio = { version = "1", features = ["full"] }
 tokio-stream = "0.1.6"
+async-trait = "0.1.77"
 warp = "0.3"
 serde = { version = "1.0", features = ["derive"]}
 serde_json = "1.0"
 futures = { version = "0.3", default-features=false}
-uuid = { version = "1.1.2", features = ["v4", "fast-rng", "macro-diagnostics"]}
diff --git a/backend/README.md b/backend/README.md
new file mode 100644
index 0000000..da43389
--- /dev/null
+++ b/backend/README.md
@@ -0,0 +1,6 @@
+# Setup
+1. Install Rust
+2. cargo build
+3. echo DATABASE_URL=postgres://username:password@localhost/diesel_demo > .env
+4. diesel setup
+5. cargo run
diff --git a/backend/src/cache.rs b/backend/src/cache.rs
new file mode 100644
index 0000000..ca27e21
--- /dev/null
+++ b/backend/src/cache.rs
@@ -0,0 +1,22 @@
+use crate::kube::{KubeId, Kube};
+use std::vec::Vec;
+
+pub struct Recipe {
+    items: Vec<KubeId>,
+}
+
+impl Recipe {
+    pub fn new(items: Vec<KubeId>) -> Self {
+        let mut items = items;
+        items.sort();
+        Recipe { items }
+    }
+    pub fn hash(&self) -> u64 {
+        let big_key = self.items.iter().fold(0, |acc, x| acc ^ x.as_u128());
+        (big_key >> 64 & big_key) as u64
+    }
+}
+
+pub struct PsqlCache {
+
+}
diff --git a/backend/src/cache_test.rs b/backend/src/cache_test.rs
new file mode 100644
index 0000000..19ae28e
--- /dev/null
+++ b/backend/src/cache_test.rs
@@ -0,0 +1,20 @@
+#[cfg(test)]
+mod tests {
+    // Note this useful idiom: importing names from outer (for mod tests) scope.
+    use super::*;
+
+    #[test]
+
+    fn test_kube_cache_hash_is_not_order_sensitive() {
+        let mut items = vec![];
+        items.push(KubeId::new("a"));
+        items.push(KubeId::new("b"));
+        items.push(KubeId::new("b"));
+        items.push(KubeId::new("c"));
+        let recipe1 = Recipe::new(items.clone());
+
+        items.reverse();
+        let recipe2 = Recipe::new(items.clone());
+        assert_eq!(recipe1.hash(), recipe2.hash());
+    }
+}
diff --git a/backend/src/kube.rs b/backend/src/kube.rs
index 33e690b..5c82bcd 100644
--- a/backend/src/kube.rs
+++ b/backend/src/kube.rs
@@ -1,7 +1,36 @@
-#[derive(Debug, PartialEq)]
+use uuid::Uuid;
+
+#[derive(PartialEq, Eq, PartialOrd, Ord, Debug)]
+pub struct KubeId {
+    uuid: Uuid,
+}
+
+impl KubeId {
+    pub fn new(name: &str) -> Self {
+        let mut name = name.to_string();
+        name.push_str("kube");
+        KubeId {
+            uuid: Uuid::new_v5(&Uuid::NAMESPACE_DNS, name.as_bytes()),
+        }
+    }
+
+    pub fn as_u128(&self) -> u128 {
+        self.uuid.as_u128()
+    }
+}
+
+#[derive(PartialEq, Debug)]
 pub struct Kube {
-    name: String,
-    uuid: String, //uuid type?
+    pub id: KubeId,
+    pub name: String,
+}
+impl Kube {
+    pub fn new(name: String) -> Kube {
+        Kube {
+            id: KubeId::new(name.as_str()),
+            name,
+        }
+    }
 }
 
 // we should have a placeholder ''loading'' cube we can send over if api is slow
diff --git a/backend/src/lib.rs b/backend/src/lib.rs
index 824e2f3..e76e302 100644
--- a/backend/src/lib.rs
+++ b/backend/src/lib.rs
@@ -2,5 +2,7 @@ pub mod grid;
 pub mod space;
 pub mod kube;
 pub mod player;
+pub mod llm;
+pub mod cache;
 
 type Coordinate = [u64; 2];
diff --git a/backend/src/llm.rs b/backend/src/llm.rs
new file mode 100644
index 0000000..c94aecb
--- /dev/null
+++ b/backend/src/llm.rs
@@ -0,0 +1,52 @@
+use crate::kube::Kube;
+use async_trait::async_trait;
+
+/// Trait for interacting with an LLM.
+#[async_trait]
+pub trait LLM {
+    /// Send a query to the LLM and get a [`std::string::String`] response.
+    async fn query(input: &str) -> String;
+    /// Ask the LLM to combine the given Kubes and return a new Kube.
+    async fn combine(&self, kubes: &[Kube]) -> Kube;
+}
+
+/// A fake LLM that functions very basically, not processing the input in any meaningful way. This is most useful for testing functionality of other features which use LLMs.
+pub struct FakeLLM {
+}
+impl FakeLLM {
+    fn new() -> FakeLLM {
+        FakeLLM {  }
+    }
+}
+
+#[async_trait]
+impl LLM for FakeLLM {
+    async fn query(input: &str) -> String {
+        format!("This is a response to: {input}")
+    }
+    async fn combine(&self, kubes: &[Kube]) -> Kube {
+        let mut new_string = String::new();
+        for kube in kubes {
+            new_string.push_str(kube.name.as_str());
+        }
+        Kube::new(new_string)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    #[tokio::test]
+    async fn fake_combine_test() {
+        let kubes = vec![
+            Kube::new(String::from("water")),
+            Kube::new(String::from("glass")),
+        ];
+        let fake_llm = FakeLLM::new();
+        let response_kube = fake_llm.combine(&kubes).await;
+        assert_eq!(
+            String::from("waterglass"),
+            response_kube.name,
+        );
+    }
+}
diff --git a/backend/src/schema.rs b/backend/src/schema.rs
new file mode 100644
index 0000000..f1dc5de
--- /dev/null
+++ b/backend/src/schema.rs
@@ -0,0 +1,33 @@
+// @generated automatically by Diesel CLI.
+
+diesel::table! {
+    kube_recipe_lines (recipe_id) {
+        recipe_id -> Uuid,
+        input_id -> Uuid,
+    }
+}
+
+diesel::table! {
+    kube_recipes (id) {
+        id -> Uuid,
+        output_id -> Uuid,
+    }
+}
+
+diesel::table! {
+    kubes (id) {
+        id -> Uuid,
+        #[max_length = 255]
+        name -> Varchar,
+    }
+}
+
+diesel::joinable!(kube_recipe_lines -> kube_recipes (recipe_id));
+diesel::joinable!(kube_recipe_lines -> kubes (input_id));
+diesel::joinable!(kube_recipes -> kubes (output_id));
+
+diesel::allow_tables_to_appear_in_same_query!(
+    kube_recipe_lines,
+    kube_recipes,
+    kubes,
+);
diff --git a/diesel.toml b/diesel.toml
new file mode 100644
index 0000000..be61e3e
--- /dev/null
+++ b/diesel.toml
@@ -0,0 +1,9 @@
+# For documentation on how to configure this file,
+# see https://diesel.rs/guides/configuring-diesel-cli
+
+[print_schema]
+file = "backend/src/schema.rs"
+custom_type_derives = ["diesel::query_builder::QueryId"]
+
+[migrations_directory]
+dir = "migrations"
diff --git a/migrations/.keep b/migrations/.keep
new file mode 100644
index 0000000..e69de29
diff --git a/migrations/00000000000000_diesel_initial_setup/down.sql b/migrations/00000000000000_diesel_initial_setup/down.sql
new file mode 100644
index 0000000..a9f5260
--- /dev/null
+++ b/migrations/00000000000000_diesel_initial_setup/down.sql
@@ -0,0 +1,6 @@
+-- This file was automatically created by Diesel to setup helper functions
+-- and other internal bookkeeping. This file is safe to edit, any future
+-- changes will be added to existing projects as new migrations.
+
+DROP FUNCTION IF EXISTS diesel_manage_updated_at(_tbl regclass);
+DROP FUNCTION IF EXISTS diesel_set_updated_at();
diff --git a/migrations/00000000000000_diesel_initial_setup/up.sql b/migrations/00000000000000_diesel_initial_setup/up.sql
new file mode 100644
index 0000000..d68895b
--- /dev/null
+++ b/migrations/00000000000000_diesel_initial_setup/up.sql
@@ -0,0 +1,36 @@
+-- This file was automatically created by Diesel to setup helper functions
+-- and other internal bookkeeping. This file is safe to edit, any future
+-- changes will be added to existing projects as new migrations.
+
+
+
+
+-- Sets up a trigger for the given table to automatically set a column called
+-- `updated_at` whenever the row is modified (unless `updated_at` was included
+-- in the modified columns)
+--
+-- # Example
+--
+-- ```sql
+-- CREATE TABLE users (id SERIAL PRIMARY KEY, updated_at TIMESTAMP NOT NULL DEFAULT NOW());
+--
+-- SELECT diesel_manage_updated_at('users');
+-- ```
+CREATE OR REPLACE FUNCTION diesel_manage_updated_at(_tbl regclass) RETURNS VOID AS $$
+BEGIN
+    EXECUTE format('CREATE TRIGGER set_updated_at BEFORE UPDATE ON %s
+                    FOR EACH ROW EXECUTE PROCEDURE diesel_set_updated_at()', _tbl);
+END;
+$$ LANGUAGE plpgsql;
+
+CREATE OR REPLACE FUNCTION diesel_set_updated_at() RETURNS trigger AS $$
+BEGIN
+    IF (
+        NEW IS DISTINCT FROM OLD AND
+        NEW.updated_at IS NOT DISTINCT FROM OLD.updated_at
+    ) THEN
+        NEW.updated_at := current_timestamp;
+    END IF;
+    RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
diff --git a/migrations/2024-03-02-174331_create_init_db/down.sql b/migrations/2024-03-02-174331_create_init_db/down.sql
new file mode 100644
index 0000000..2488ed7
--- /dev/null
+++ b/migrations/2024-03-02-174331_create_init_db/down.sql
@@ -0,0 +1,5 @@
+-- This file should undo anything in `up.sql`
+
+DROP TABLE kubes_recipe_lines;
+DROP TABLE kubes_recipes;
+DROP TABLE kubes;
diff --git a/migrations/2024-03-02-174331_create_init_db/up.sql b/migrations/2024-03-02-174331_create_init_db/up.sql
new file mode 100644
index 0000000..fa9479a
--- /dev/null
+++ b/migrations/2024-03-02-174331_create_init_db/up.sql
@@ -0,0 +1,20 @@
+-- Your SQL goes here
+
+CREATE TABLE kubes (
+  id uuid primary key,
+  name varchar(255) not null
+);
+
+CREATE INDEX k_i on kubes(name);
+
+CREATE TABLE kube_recipes (
+  id uuid primary key,
+  output_id uuid not null references kubes(id)
+);
+
+CREATE TABLE kube_recipe_lines (
+  recipe_id uuid primary key references kube_recipes(id),
+  input_id uuid not null references kubes(id)
+);
+
+CREATE INDEX krl_i ON kube_recipe_lines(recipe_id, input_id);