feat: OpenAI compatible http service (#123)

Signed-off-by: Ryan Olson <ryanolson@users.noreply.github.com>
Co-authored-by: Ryan McCormick <rmccormick@nvidia.com>
Co-authored-by: Neelay Shah <neelays@nvidia.com>

llm/rust/triton-llm/tests/snapshots/openai_completions__valid_samples-8.snap

+---
+source: triton-llm/tests/openai_completions.rs
+description: "should have prompt, model, and echo fields"
+expression: sample.request
+---
+{
+  "model": "gpt-3.5-turbo",
+  "prompt": "What is the meaning of life?",
+  "echo": true
+}

llm/rust/triton-llm/tests/snapshots/openai_completions__valid_samples-9.snap

+---
+source: triton-llm/tests/openai_completions.rs
+description: "should have prompt, model, and stream fields"
+expression: sample.request
+---
+{
+  "model": "gpt-3.5-turbo",
+  "prompt": "What is the meaning of life?",
+  "stream": true
+}

llm/rust/triton-llm/tests/snapshots/openai_completions__valid_samples.snap

+---
+source: triton-llm/tests/openai_completions.rs
+description: should have only prompt and model fields
+expression: sample.request
+---
+{
+  "model": "gpt-3.5-turbo",
+  "prompt": "What is the meaning of life?"
+}

runtime/rust/Cargo.lock

@@ -2326,7 +2326,7 @@ dependencies = [
 
 [[package]]
 name = "triton-distributed"
-version = "0.1.2"
+version = "0.1.3"
 dependencies = [
  "anyhow",
  "assert_matches",

runtime/rust/Cargo.toml

@@ -17,7 +17,7 @@
 name = "triton-distributed"
 description = "Distributed GenAI Inference Framework"
 readme = "README.md"
-version = "0.1.2" # TODO: Centralize Version Automation
+version = "0.1.3" # TODO: Centralize Version Automation
 edition = "2021"
 authors = ["NVIDIA"]
 license = "Apache-2.0"

runtime/rust/examples/hello_world/Cargo.toml

@@ -15,7 +15,7 @@
 
 [package]
 name = "hello_world"
-version = "0.1.0"
+version = "0.1.3"
 edition = "2021"
 
 [dependencies]

runtime/rust/examples/hello_world/rust-toolchain.toml

-[toolchain]
-channel = "1.84.1"

runtime/rust/python-wheel/Cargo.lock

@@ -1433,9 +1433,9 @@ dependencies = [
 
 [[package]]
 name = "once_cell"
-version = "1.20.2"
+version = "1.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1261fe7e33c73b354eab43b1273a57c8f967d0391e80353e51f764ac02cf6775"
+checksum = "945462a4b81e43c4e3ba96bd7b49d834c6f61198356aa858733bc4acf3cbe62e"
 
 [[package]]
 name = "openssl-probe"
@@ -2594,7 +2594,7 @@ dependencies = [
 
 [[package]]
 name = "triton-distributed"
-version = "0.1.2"
+version = "0.1.3"
 dependencies = [
  "anyhow",
  "async-nats",
@@ -2632,7 +2632,7 @@ dependencies = [
 
 [[package]]
 name = "triton_distributed_py3"
-version = "0.1.1"
+version = "0.1.3"
 dependencies = [
  "futures",
  "once_cell",

runtime/rust/python-wheel/Cargo.toml

@@ -15,7 +15,7 @@
 
 [package]
 name = "triton_distributed_py3"
-version = "0.1.1"
+version = "0.1.3"
 edition = "2021"
 authors = ["NVIDIA"]
 license = "Apache-2.0"
@@ -30,10 +30,10 @@ crate-type = ["cdylib"]
 
 
 [dependencies]
-triton-distributed = { version = "0.1.1", path = "../" }
+triton-distributed = { version = "0.1.3", path = "../" }
 
 futures = "0.3"
-once_cell = "1"
+once_cell = "1.20.3"
 serde = "1"
 serde_json = "1.0.138"
 tokio = { version = "1", features = ["full"] }

runtime/rust/python-wheel/pyproject.toml

@@ -16,7 +16,7 @@
 
 [project]
 name = "triton-distributed-rs"
-version = "0.1.1"
+version = "0.1.3"
 description = "Distributed LLM Framework"
 # readme = "README.md"
 authors = [

runtime/rust/python-wheel/rust-toolchain.toml

-[toolchain]
-channel = "1.84.1"

runtime/rust/rust-toolchain.toml

-[toolchain]
-channel = "1.84.1"

runtime/rust/src/pipeline/context.rs

@@ -17,7 +17,7 @@
 //! There are two context object defined in this module:
 //!
 //! - [`Context`] is an input context which is propagated through the processing pipeline,
-//!    up to the point where the input is pass to an [`nim_llm_async_engine::AsyncEngine`] for processing.
+//!    up to the point where the input is pass to an [`triton_distributed::engine::AsyncEngine`] for processing.
 //! - [`StreamContext`] is the input context transformed into to a type erased context that maintains the inputs
 //!   registry and visitors. `StreamAdaptors` will amend themselves to the [`StreamContext`] to allow for the
 
@@ -299,7 +299,7 @@ impl<T: Send + Sync + 'static> From<Context<T>> for StreamContext {
     }
 }
 
-// TODO - refactor here - this came from the nim-llm-async-engine crate
+// TODO - refactor here - this came from the triton-llm-async-engine crate
 
 use tokio::sync::watch::{channel, Receiver, Sender};