Repository navigation
perf(samples): pre-position agents in parallel - #84
Merged
Merged
Conversation
Pre-positioning ran each agent's Authorization-T + Notification-T sequence serially; every step is LLM-bound (prompt generation plus agent-side validation, with retries), so startup spent ~169s with four agents-worth of sequential model calls. Run each agent's sequence on its own worker (bounded at 4 concurrent). Agents are independent: separate transports/channels per operation, separate remote tasks, and the workflow executor already exercises concurrent sends over one transport. Failure semantics are unchanged: per-agent failures are counted and skipped, never aborting other agents, and the subscription handle is still transferred to the lifecycle owner before the acknowledgement wait. Also raise the samples LlmHelper default A2AT_LLM_MAX_TOKENS to 4000: field logs show qwen3 reasoning content exhausting 2000 tokens, which truncated validation output and forced retries. Signed-off-by: Zhoujie628 <zhoujie628@users.noreply.github.com>
…tems The gates calibrated on the core modules also cover the samples module: the demo entry method and the vendor SSE adapter carry cohesive methods slightly over the length budget, intrinsic wire-compatibility complexity, and conventional lowercase static-final holders (loggers, mappers, flags). Suppress per file so new code elsewhere stays gated. Signed-off-by: Zhoujie628 <zhoujie628@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pre-positioning ran each agent Authorization-T + Notification-T sequence serially; every step is LLM-bound (prompt generation plus agent-side validation, with retries), so startup spent ~169s on sequential model calls.
Run each agent sequence on its own worker (bounded at 4 concurrent). Agents are independent: separate transports/channels per operation and separate remote tasks - the workflow executor already exercises concurrent sends over one transport. Failure semantics unchanged: per-agent failures are counted and skipped, and the subscription handle is still transferred to the lifecycle owner before the acknowledgement wait.
Also raise the samples LlmHelper default A2AT_LLM_MAX_TOKENS to 4000 (field logs show qwen3 reasoning content exhausting 2000 tokens, truncating validation output and forcing retries), and suppress accepted legacy demo/vendor-adapter checkstyle items in the shared suppressions file.