docs(schema): document HNSW int8 quantization option#508
Closed
kriszyp wants to merge 0 commit into
Closed
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
🚀 Preview DeploymentYour preview deployment is ready! 🔗 Preview URL: https://preview.harper-documentation.harperfabric.com/pr-508 This preview will update automatically when you push new commits. |
| | `optimizeRouting` | `0.5` | Heuristic aggressiveness for omitting redundant connections (0 = off, 1 = most aggressive) | | ||
| | `mL` | computed from `M` | Normalization factor for level generation | | ||
| | `efSearchConstruction` | `50` | Max nodes explored during search | | ||
| | `quantization` | _(full precision)_ | `"int8"` stores each indexed vector as 8-bit scalar-quantized values plus a per-vector scale instead of float32 — roughly a 5× smaller index and substantially faster search, at a small recall cost (~1%). Omit for full-precision float32. Only the index is quantized; the full-precision vector on the record is unchanged. | |
Member
There was a problem hiding this comment.
I think default should be "float32" for this option to make more sense. _(full precision)_ is not a value. I believe right now this option only supports two values, but will there be more in the future? If not, and this is truly a binary configuration, should it not reflect that better? Like maybe making it a toggleable int8Quantization: boolean ?
4 tasks
kriszyp
force-pushed
the
kris/hnsw-int8-quantization-docs
branch
from
7da20a0 to
dcd5dc2
Compare
kriszyp
added a commit
that referenced
this pull request
Jun 13, 2026
* docs(v5.1): release notes, deployment tracking ops, deploy_component updates - Add 5.1.md release notes covering: models/AI, @embed directive, MCP server, deployment tracking, HNSW int8 quantization, and replication improvements - Update deploy_component docs: urlPath, install_allow_scripts params, deployment_id response - Document new deployment operations: list_deployments, get_deployment, get_deployment_payload, delete_deployment_payload - Document hdb_deployment record schema (fields, phases, peer_results) Note: models/AI detail, MCP reference, and HNSW quantization have separate PRs (#523, #507/#516, #508) — this PR adds the release notes overview and the deployment tracking operations which had no coverage. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * style: run prettier on changed files Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: remove cross-plugin MCP link that breaks Docusaurus build The release-notes and reference doc plugins are separate; relative .md links between them resolve incorrectly. Removing until PR #507 (MCP reference section) merges and can be linked with an absolute path. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * docs(5.1): expand release notes — middleware/routing, caching, LOCAL_ONLY, HARPER_CONFIG, RocksDB, migrateOnStart, upgrade improvements --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
🚀 Preview DeploymentYour preview deployment is ready! 🔗 Preview URL: https://preview.harper-documentation.harperfabric.com/pr-508 This preview will update automatically when you push new commits. |
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
2 participants
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
Documents the new
quantization: "int8"option for HNSW vector indexes — adds a row to the HNSW parameter table and a short example.Context
Pairs with HarperFast/harper#894 (optional int8 vector quantization for the HNSW index): ~5× smaller index, substantially faster search, ~1% recall cost; opt-in, the record's full-precision vector is unchanged. Making it the default is tracked separately (HarperFast/harper#932).
Generated with the assistance of an LLM (Claude Opus 4.8).