chore: include ollama-cli in workspace

djinn-rs/Cargo.toml → Cargo.toml

@@ -1,6 +1,6 @@
 [workspace]
-members = [ "djinn-cli","djinn-core", "djinn-server"]
-default-members = ["djinn-cli"]
+members = [ "djinn-cli","djinn-core", "djinn-server", "ollama-cli" ]
+default-members = ["djinn-cli", "ollama-cli"]
 resolver = "2"
 
 [workspace.package]

README.md

@@ -1,22 +1,37 @@
-# djinn
+# `djinn-rs`
 
-Right now the Rust implementation is most complete.
+a Rust playground
+for running language models
+and some other machine learning models.
 
-Run with command line arguments:
+## `djinn-cli`
+the main entrypoint
+for running either
+a server (with `djinn-server`)
+or a one shot inference.
 
-```sh
-$ cd djinn-rs
-$ cargo run -- --help
-# example prompt
-$ cargo run --release -- llama --disable-flash-attention --model-version v1 --dtype f32
- "In a world" --sample-len 256 --repeat-penalty 50
-```
+## `djinn-core`
+
+core ML functionality and model implementations.
+
+## `djinn-server`
 
-## supported models
+an HTTP server
+with an API
+and HTMX front-end
+for running models
 
-- Llama 1
-- YOLOv8
+# examples
+
+run the server on a Macbook M-series:
+
+```sh
+cargo run --release --features djinn-core/mac -- server-config --name test
+```
 
-### TODO
-- Llama 2
-- [LoRA](https://github.com/EricLBuehler/candle-lora)
+breakdown:
+    - `cargo run --release` to build in release mode for best performance
+    - `--features djinn-core/mac` to enable CoreML acceleration
+    - `--` everything before this are `cargo` args and everything after are `djinn` args
+    - `server-config` command to run the server from a config file
+    - `--name test` to run the config named `test`, in `./configs/server/test.toml`

ollama/src/tui/messages/mod.rs → ollama-cli/src/tui/messages/mod.rs

@@ -21,8 +21,6 @@ pub mod widget_item;
 pub mod widget_rendering;
 pub mod widget_state;
 
-const TEST_OUTPUT: &str = include_str!("../../../example_output.txt");
-
 #[derive(Clone, Debug)]
 pub struct MessagesViewModel {
     /// Streaming response from the model before it's finished.
@@ -37,7 +35,7 @@ impl Default for MessagesViewModel {
     fn default() -> Self {
         Self {
             model_stream: Default::default(),
-            messages: VecDeque::from([Message::Assistant(TEST_OUTPUT.into())]),
+            messages: VecDeque::from([Message::Assistant("".into())]),
             state: Default::default(),
         }
     }