chore(dynamo-run): Fix echo_core for EOS tokens (#759)

"echo_core" is an engine that echoes the post-processed request back to you so you can see the template. Good for testing. It needed an extra flag set to work correctly.

docs/guides/dynamo_run.md

@@ -398,6 +398,13 @@ The `echo_core` engine accepts pre-processed requests and echoes the tokens back
 dynamo-run in=http out=echo_core --model-path <hf-repo-checkout>
 ```
 
+Note that to use it with `in=http` you need to tell the post processor to ignore stop tokens from the template by adding `nvext.ignore_eos` like this:
+```
+curl -N -d '{"nvext": {"ignore_eos": true}, "stream": true, "model": "Qwen2.5-3B-Instruct", "max_completion_tokens": 4096, "messages":[{"role":"user", "content": "Tell me a story" }]}' ...
+```
+
+The default `in=text` sets that for you.
+
 #### echo_full
 
 The `echo_full` engine accepts un-processed requests and echoes the prompt back as the response.

launch/dynamo-run/src/input/text.rs

@@ -13,6 +13,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use dynamo_llm::protocols::openai::nvext::NvExt;
 use dynamo_llm::types::openai::chat_completions::{
     NvCreateChatCompletionRequest, OpenAIChatCompletionsStreamingEngine,
 };
@@ -110,6 +111,10 @@ async fn main_loop(
             .temperature(0.7)
             .n(1) // only generate one response
             .build()?;
+        let nvext = NvExt {
+            ignore_eos: Some(true),
+            ..Default::default()
+        };
 
         // TODO We cannot set min_tokens with async-openai
         // if inspect_template {
@@ -117,7 +122,10 @@ async fn main_loop(
         //     req_builder.min_tokens(8192);
         // }
 
-        let req = NvCreateChatCompletionRequest { inner, nvext: None };
+        let req = NvCreateChatCompletionRequest {
+            inner,
+            nvext: Some(nvext),
+        };
 
         // Call the model
         let mut stream = engine.generate(Context::new(req)).await?;

lib/llm/src/http/service/openai.rs

@@ -148,7 +148,10 @@ async fn completions(
         ..request.inner
     };
 
-    let request = CompletionRequest { inner, nvext: None };
+    let request = CompletionRequest {
+        inner,
+        nvext: request.nvext,
+    };
 
     // todo - make the protocols be optional for model name
     // todo - when optional, if none, apply a default
@@ -233,9 +236,10 @@ async fn chat_completions(
         stream: Some(true),
         ..request.inner
     };
+
     let request = NvCreateChatCompletionRequest {
         inner: inner_request,
-        nvext: None,
+        nvext: request.nvext,
     };
 
     // todo - make the protocols be optional for model name