Repository navigation
docs(eval): enrichment concurrency is near-linear (7.8× at concurrency 8) with a parallel backend - #118
Merged
Conversation
…y 8) with a parallel backend Isolates harness concurrency from the LLM backend: re-ran the deferred-enrichment path against a concurrent mock endpoint (fixed 1s/call). Synchronous 844.7s vs deferred 108.6s = 7.8x (~97.5% efficiency at concurrency 8). Brackets the earlier real-7B result (1.1x, ollama-serialized): the async concurrency is real and near-linear; realized speedup = the backend's effective serving parallelism. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
There was a problem hiding this comment.
Your trial has ended. Reactivate Greptile to resume code reviews.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Proves the async write-path enrichment concurrency (merged in #117) actually delivers — the earlier real-7B run showed only 1.1× because ollama serialized the requests, not because of the code.
Re-ran the same
--deferred-enrichmentpath (same conv-0, same 8-wayenrich_pending) against a mock OpenAI-compatible endpoint that serves requests concurrently (fixed 1s/call) — a genuinely parallel-capable backend:7.8× at concurrency 8 (~97.5% efficiency). The two measurements bracket the truth:
So the async code's concurrency is real and near-linear; the realized wall-clock speedup equals the LLM backend's effective serving parallelism. Docs-only; the real-LLM number remains the honest end-to-end figure for a single local model.
🤖 Generated with Claude Code