docs: explain the indexing cost model (cold vs incremental vs model swap)#610
Draft
j-sperling wants to merge 1 commit into
Draft
docs: explain the indexing cost model (cold vs incremental vs model swap)#610j-sperling wants to merge 1 commit into
j-sperling wants to merge 1 commit into
Conversation
Deduplication makes indexing cost event-driven, and users reading "slow indexing" reports are usually seeing one of two full-cost events: a cold index or an embedding model change. Document the three cost tiers, the model-name-in-chunk-ID consequence (a model swap re-embeds the entire corpus and can require a collection rebuild), and the three throughput knobs for local providers, with measured reference numbers. Add a FAQ entry for the slow-first-index question linking to it.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
docs/architecture.md, directly after Deduplication: indexing cost is event-driven, and the three tiers (steady-state incremental, cold index, embedding model change) have very different costs. The non-obvious one for users: the model name is part of every composite chunk ID, so changingembedding.model/providerre-embeds the entire corpus and, on a dimension change, requires a collection rebuild — worth treating as a scheduled migration.embedding.batch_size) with hardware-qualified reference measurements (default onnx model on an Apple M-series CPU: ~10 texts/s; batch 64 ~11% faster than 32).Test plan
mkdocs build --strictclean (anchors/links resolve, including the FAQ dimension-mismatch cross-reference)