feat: adding http clients and recorded response stream

[ryanolson]

Summary by CodeRabbit

• New Features

  • Introduced advanced HTTP clients for streaming large language model (LLM) responses, supporting multiple client types and enhanced request context management.
  • Added performance recording and analysis tools for streaming LLM responses, enabling detailed metrics and timing analysis.

• Bug Fixes

  • Improved error handling and context propagation in streaming chat completion requests.

• Tests

  • Added comprehensive integration tests for HTTP clients, covering streaming, error handling, and context management.

• Chores

  • Updated dependencies to enable additional features for improved security and flexibility.
  • Relaxed concurrency requirements for certain traits to enhance compatibility.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Walkthrough

This change introduces a comprehensive HTTP client module for streaming LLM responses, supporting OpenAI, NVIDIA custom, and BYOT (Bring Your Own Types) protocols. It adds performance recording utilities, new integration tests for client-server streaming, and modifies trait bounds for engine context providers. Dependency features are updated to enable "rustls" and "byot" support.

Changes

File(s)	Change Summary
Cargo.toml, lib/llm/Cargo.toml	Enabled "rustls" and "byot" features for async-openai and async-stream dependencies.
lib/llm/src/http.rs, lib/llm/src/lib.rs	Added new public module declarations: client (http.rs) and perf (lib.rs).
lib/llm/src/http/client.rs	Added a new HTTP client module supporting streaming LLM responses for OpenAI, NVIDIA, and BYOT, with context and error handling.
lib/llm/src/perf.rs	Added a new performance recording module for streaming responses, with timestamping, metrics, and recording stream wrappers.
lib/llm/tests/http-service.rs	Added integration tests for all new HTTP client types, covering streaming, error, and context propagation scenarios.
lib/runtime/src/engine.rs	Relaxed trait bounds for AsyncEngineContextProvider by removing the Sync requirement.

Sequence Diagram(s)

sequenceDiagram
    participant Test as Test Runner
    participant Client as HTTP Client (OpenAI/NV/BYOT)
    participant Service as HTTP Service
    participant Engine as LLM Engine

    Test->>Client: Create streaming chat completion request
    Client->>Service: POST /chat/completions (stream: true)
    Service->>Engine: Forward request to model engine
    Engine-->>Service: Stream responses
    Service-->>Client: Stream response chunks
    Client-->>Test: Yield streamed responses (with context)

Loading

sequenceDiagram
    participant Client as HTTP Client
    participant Perf as RecordingStream
    participant Consumer as Stream Consumer

    Client->>Perf: Stream LLM responses
    Perf->>Perf: Timestamp and record each response
    Perf-->>Consumer: Yield responses as stream
    Perf->>Consumer: Send performance metrics on completion

Loading

Possibly related PRs

• ai-dynamo/dynamo#1216: Removes bring-your-own-engine feature and related code. This is related as the current PR introduces BYOT client support, while the referenced PR removes BYOT engine support.

Poem

🐇
New clients hop in, streaming with flair,
OpenAI, NVIDIA, BYOT—handled with care!
Performance is tracked with a timestamped stream,
Context flows smoothly, like a bunny’s daydream.
Tests now abound, so bugs can’t hide—
In this LLM meadow, the rabbits take pride!

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

• Review comments: Directly reply to a review comment made by CodeRabbit. Example:
  • I pushed a fix in commit <commit_id>, please review it.
  • Explain this complex logic.
  • Open a follow-up GitHub issue for this discussion.
• Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
  • @coderabbitai explain this code block.
  • @coderabbitai modularize this function.
• PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
  • @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
  • @coderabbitai read src/utils.ts and explain its main purpose.
  • @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
  • @coderabbitai help me debug CodeRabbit configuration file.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

• @coderabbitai pause to pause the reviews on a PR.
• @coderabbitai resume to resume the paused reviews.
• @coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
• @coderabbitai full review to do a full review from scratch and review all the files again.
• @coderabbitai summary to regenerate the summary of the PR.
• @coderabbitai generate sequence diagram to generate a sequence diagram of the changes in this PR.
• @coderabbitai resolve resolve all the CodeRabbit review comments.
• @coderabbitai configuration to show the current CodeRabbit configuration for the repository.
• @coderabbitai help to get help.

Other keywords and placeholders

• Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
• Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
• Add @coderabbitai anywhere in the PR title to generate the title automatically.

Documentation and Community

• Visit our Documentation for detailed information on how to use CodeRabbit.
• Join our Discord Community to get help, request features, and share feedback.
• Follow us on X/Twitter for updates and announcements.

[coderabbitai]

Actionable comments posted: 7

🧹 Nitpick comments (1)

lib/llm/src/perf.rs (1)

118-121: Improve average calculation to handle large response counts.

The division by u32 could overflow for very large streams.

         let total_time = self.total_duration;
         let intervals = self.responses.len() - 1;
-        Some(total_time / intervals as u32)
+        Some(Duration::from_nanos(
+            total_time.as_nanos() as u64 / intervals as u64
+        ))

📜 Review details

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between f00d700 and 81cbd9c.

📒 Files selected for processing (8)

• Cargo.toml (1 hunks)
• lib/llm/Cargo.toml (1 hunks)
• lib/llm/src/http.rs (1 hunks)
• lib/llm/src/http/client.rs (1 hunks)
• lib/llm/src/lib.rs (1 hunks)
• lib/llm/src/perf.rs (1 hunks)
• lib/llm/tests/http-service.rs (3 hunks)
• lib/runtime/src/engine.rs (1 hunks)

🧰 Additional context used

🧠 Learnings (5)

Cargo.toml (1)

Learnt from: kthui
PR: ai-dynamo/dynamo#1424
File: lib/runtime/src/pipeline/network/egress/push_router.rs:204-209
Timestamp: 2025-06-13T22:07:24.843Z
Learning: The codebase uses async-nats version 0.40, not the older nats crate. Error handling should use async_nats::error::Error variants, not nats::Error variants.

lib/runtime/src/engine.rs (1)

Learnt from: PeaBrane
PR: ai-dynamo/dynamo#1236
File: lib/llm/src/mocker/engine.rs:140-161
Timestamp: 2025-06-17T00:50:44.845Z
Learning: In Rust async code, when an Arc<Mutex<_>> is used solely to transfer ownership of a resource (like a channel receiver) into a spawned task rather than for sharing between multiple tasks, holding the mutex lock across an await is not problematic since there's no actual contention.

lib/llm/Cargo.toml (1)

Learnt from: kthui
PR: ai-dynamo/dynamo#1424
File: lib/runtime/src/pipeline/network/egress/push_router.rs:204-209
Timestamp: 2025-06-13T22:07:24.843Z
Learning: The codebase uses async-nats version 0.40, not the older nats crate. Error handling should use async_nats::error::Error variants, not nats::Error variants.

lib/llm/tests/http-service.rs (2)

Learnt from: kthui
PR: ai-dynamo/dynamo#1424
File: lib/runtime/src/pipeline/network/egress/push_router.rs:204-209
Timestamp: 2025-06-13T22:07:24.843Z
Learning: The codebase uses async-nats version 0.40, not the older nats crate. Error handling should use async_nats::error::Error variants, not nats::Error variants.

Learnt from: oandreeva-nv
PR: ai-dynamo/dynamo#1195
File: lib/llm/tests/block_manager.rs:150-152
Timestamp: 2025-06-02T19:37:27.666Z
Learning: In Rust/Tokio applications, when background tasks use channels for communication, dropping the sender automatically signals task termination when the receiver gets `None`. The `start_batching_publisher` function in `lib/llm/tests/block_manager.rs` demonstrates this pattern: when the `KVBMDynamoRuntimeComponent` is dropped, its `batch_tx` sender is dropped, causing `rx.recv()` to return `None`, which triggers cleanup and task termination.

lib/llm/src/perf.rs (1)

Learnt from: oandreeva-nv
PR: ai-dynamo/dynamo#1195
File: lib/llm/tests/block_manager.rs:150-152
Timestamp: 2025-06-02T19:37:27.666Z
Learning: In Rust/Tokio applications, when background tasks use channels for communication, dropping the sender automatically signals task termination when the receiver gets `None`. The `start_batching_publisher` function in `lib/llm/tests/block_manager.rs` demonstrates this pattern: when the `KVBMDynamoRuntimeComponent` is dropped, its `batch_tx` sender is dropped, causing `rx.recv()` to return `None`, which triggers cleanup and task termination.

🧬 Code Graph Analysis (1)

lib/llm/src/perf.rs (3)

lib/runtime/src/engine.rs (18)

• new (225-227)
• from (104-106)
• poll_next (234-239)
• context (169-169)
• context (245-247)
• context (262-264)
• context (268-270)
• context (443-446)
• context (458-460)
• fmt (251-258)
• id (126-126)
• stop (156-156)
• stop_generating (153-153)
• kill (161-161)
• is_stopped (129-129)
• is_killed (137-137)
• stopped (141-141)
• killed (145-145)

lib/llm/src/http/client.rs (22)

• new (71-78)
• new (193-200)
• new (262-267)
• new (365-369)
• new (424-428)
• new (482-486)
• into_async_engine_stream (287-297)
• poll_next (273-275)
• poll_next (309-311)
• context (279-281)
• context (315-317)
• fmt (129-137)
• fmt (327-331)
• fmt (339-343)
• id (142-144)
• stop (146-150)
• stop_generating (152-155)
• kill (157-161)
• is_stopped (163-165)
• is_killed (167-169)
• stopped (171-173)
• killed (175-178)

lib/llm/src/http/service/openai.rs (1)

• responses (400-505)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)

• GitHub Check: Build and Test - vllm

🔇 Additional comments (7)

lib/llm/src/lib.rs (1)

28-28: LGTM: Clean module addition.

The perf module declaration is properly placed and follows the existing module organization pattern.

lib/llm/src/http.rs (1)

16-16: LGTM: Appropriate module addition.

The client module declaration complements the existing service module and is properly organized.

Cargo.toml (1)

49-49: Confirmed valid features for async-openai 0.29.0

The features "rustls" and "byot" are both valid for version 0.29.0 of the async-openai crate.

lib/llm/tests/http-service.rs (3)

17-47: Imports are well-organized and necessary.

The new imports properly support the HTTP client integration tests.

---

485-533: Well-structured test fixtures with proper port isolation.

The use of rstest fixtures with configurable ports is a good practice that enables parallel test execution.

---

574-580: Good practice: Partial stream consumption for testing.

Consuming only the first few responses is an efficient way to verify streaming behavior without waiting for the entire stream.

lib/llm/src/http/client.rs (1)

90-114: Excellent implementation of hierarchical context cancellation.

The use of Tokio's child_token() ensures proper propagation of cancellation from parent to child contexts.