Repository navigation
feat(eval): --background-enrichment for the agentic QA path - #121
Merged
Merged
Conversation
Runs write-path enrichment in Background mode (instant ingest + concurrent off-critical-path enrichment) and waits for the queue to drain before querying, so a bounded write-side experiment can enrich at the LLM backend's full serving concurrency instead of serially. Composes with --exclude-superseded / --distill. Sets enrichment_mode=Background + enable_knowledge_graph; the agentic loop calls wait_for_enrichment after ingest. Flag test added. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
![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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Wires the async write-path enrichment (#117) into the agentic eval so a bounded write-side experiment can enrich at the LLM backend's full serving concurrency instead of serially.
What
--background-enrichmentflag → the agentic-QA memory config usesEnrichmentMode::Background(+enable_knowledge_graph). Ingest only enqueues (instant); a background worker enriches concurrently off the critical path.memory.wait_for_enrichment()after each conversation's ingest, before querying, so retrieval reflects the completed enrichment (distilled facts, marked supersessions).--exclude-superseded/--distill. Flag test added; both feature configs build.Why
This is the eval side of the async-enrichment work: with a concurrent-capable LLM backend, per-turn extraction/resolution that was serial (~hours) runs 8-way concurrent (~minutes). Verified end-to-end plumbing (background worker enriching 8-way against a concurrent backend, no hang).
Note
The write-side capability A/B this enables (does dropping frontier-detected supersessions help a weak answerer) was not run to completion: the OpenAI/ChatGPT OAuth account hit its usage limit (
usage_limit_reached, planprolite, resets in ~5 days) mid-experiment, which blocks both the frontier reasoner and the codex judge. The wiring + plumbing are validated; the A/B is a single run once quota resets or on a backend with headroom.🤖 Generated with Claude Code