fix(dynamo-run): Text input doesn't need a name (#80)

For the `echo` and `pystr` engines we previously required the user to pass `--model-name <x>` so we would have a name for the model. If the input is HTTP we do need this to match on the users' JSON request.

If the input is Text we don't need a name. So if the input is Text and we don't already have a name for the model, give it one.

launch/dynamo-run/README.md

@@ -170,12 +170,12 @@ Build: `cargo build --release --features python`
 If the Python engine wants to receive and returns strings - it will do the prompt templating and tokenization itself - run it like this:
 
 ```
-dynamo-run out=pystr:/home/user/my_python_engine.py --name <model-name>
+dynamo-run out=pystr:/home/user/my_python_engine.py
 ```
 
 - The `request` parameter is a map, an OpenAI compatible create chat completion request: https://platform.openai.com/docs/api-reference/chat/create
 - The function must `yield` a series of maps conforming to create chat completion stream response (example below).
-- The `--name` flag is the name we serve the model under, if using an HTTP front end.
+- If using an HTTP front-end add the `--model-name` flag. This is the name we serve the model under.
 
 The file is loaded once at startup and kept in memory.
 
@@ -218,7 +218,7 @@ dynamo-run out=pytok:/home/user/my_python_engine.py --model-path <hf-repo-checko
 {"token_ids":[791],"tokens":None,"text":None,"cum_log_probs":None,"log_probs":None,"finish_reason":None}
 ```
 
-- Command like flag `--model-path` which must point to a Hugging Face repo checkout containing the `tokenizer.json`. The `--name` flag is optional. If not provided we use the HF repo name (directory name) as the model name.
+- Command like flag `--model-path` which must point to a Hugging Face repo checkout containing the `tokenizer.json`. The `--model-name` flag is optional. If not provided we use the HF repo name (directory name) as the model name.
 
 Example engine:
 ```

launch/dynamo-run/src/lib.rs

@@ -36,6 +36,10 @@ pub use opt::{Input, Output};
 /// concatenations.
 const ENDPOINT_SCHEME: &str = "dyn://";
 
+/// When `in=text` the user doesn't need to know the model name, and doesn't need to provide it on
+/// the command line. Hence it's optional, and defaults to this.
+const INVISIBLE_MODEL_NAME: &str = "dynamo-run";
+
 /// How we identify a python string endpoint
 #[cfg(feature = "python")]
 const PYTHON_STR_SCHEME: &str = "pystr:";
@@ -81,12 +85,21 @@ pub async fn run(
         .or(flags.model_path_flag)
         .and_then(|p| p.canonicalize().ok());
     // Serve the model under the name provided, or the name of the GGUF file or HF repo.
-    let model_name = flags.model_name.or_else(|| {
-        model_path
-            .as_ref()
-            .and_then(|p| p.iter().last())
-            .map(|n| n.to_string_lossy().into_owned())
-    });
+    let model_name = flags
+        .model_name
+        .or_else(|| {
+            model_path
+                .as_ref()
+                .and_then(|p| p.iter().last())
+                .map(|n| n.to_string_lossy().into_owned())
+        })
+        .or_else(|| {
+            if in_opt == Input::Text {
+                Some(INVISIBLE_MODEL_NAME.to_string())
+            } else {
+                None
+            }
+        });
     // Load the model deployment card, if any
     // Only used by some engines, so without those feature flags it's unused.
     #[allow(unused_variables)]

launch/dynamo-run/src/opt.rs

@@ -17,6 +17,7 @@ use std::fmt;
 
 use crate::ENDPOINT_SCHEME;
 
+#[derive(PartialEq)]
 pub enum Input {
     /// Run an OpenAI compatible HTTP server
     Http,