feat(compiler): opt-in top-K retrieval for concepts-plan (bound plan prompt as the KB grows)#92
Open
cnndabbler wants to merge 2 commits into
Open
Conversation
The concepts-plan step injects every existing concept/entity brief, so the prompt grows O(N) with the KB (~2k->18k tokens at a few hundred concepts), hurting speed/cost and the model's ability to reconcile against the right existing pages (-> near-duplicate concepts). Add an opt-in top-K relevance filter (openkb/retrieval.py, dependency-free TF-IDF cosine over brief lines, query = doc summary). Off by default (concepts_plan_retrieval / concepts_plan_retrieval_k in config), so behaviour is unchanged unless enabled. The select_relevant_briefs() interface is swappable for an embedding-based ranker later. Prototype for the O(N)->O(K) plan-context scaling discussion.
Benchmark on a real 335-concept KB (ground truth = summary concept links): TF-IDF recall@40=0.90 vs embeddings 0.79. Dependency-free TF-IDF wins on this LLM-generated corpus (high lexical overlap) and is free per-doc, so it stays the default; embedding ranker kept as an option for higher-drift corpora.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
The
concepts-planstep injects every existing concept/entity brief into the prompt (_read_concept_briefs/_read_entity_briefsread the wholeconcepts/andentities/dirs). So the plan prompt grows O(N) with the KB. On a 165-doc KB this was observed climbing from ~2k tokens early to ~15–18k tokens at a few hundred concepts — which hurts cost/latency on every subsequent doc and degrades the model's ability to reconcile a new doc against the right existing pages (it starts creating near-duplicate concepts).What this does (opt-in, default off)
Adds
openkb/retrieval.pyand a small block in_compile_conceptsthat, only when enabled, keeps the top-K briefs most relevant to the current doc's summary instead of all of them. Two config keys (default off → behaviour byte-identical):select_relevant_briefs_embed, provider injected — no SDK dependency in the module) is included for higher-drift corpora / future hybrid use.Benchmark (recall@K vs. ground-truth concept links)
On a real 335-concept / 489-entity KB, using each summary's
[[concepts/X]]links as ground truth and the summary as the query (scripts/bench_retrieval.py):TF-IDF wins here (LLM-generated briefs share heavy lexical overlap with summaries) and is free per-doc, so it's the default. K=40 recovers ~90% of the relevant concepts at 12% of the full-inject prompt size.
Tradeoff
At K=40, ~10% of relevant existing concepts may fall outside the window for a given doc (small risk of a duplicate concept) in exchange for a bounded plan prompt as the KB scales. Off by default; tune
..._khigher to trade prompt size for recall.Tests:
tests/test_retrieval.py(ranker behaviour) + full suite green.