feat(vector/hnsw): add per‑query ef and distance_threshold to similar_to, fix early termination #9514

joelamming · 2025-10-19T20:30:12Z

Hugely appreciative of the Dgraph team’s work. Native vector search integrated directly into a graph database is kind of a no brainer today. Deployed Dgraph (both vanilla and customised) in systems with 1M+ vectors guiding deep traversal queries across 10M+ nodes -- tight coupling of vector search with graph traversal at massive scale gets us closer to something that could represent the fuzzy nuances of everything in an enterprise. Certainly not the biggest deployment your team will have seen, but this PR fixes an under‑recall edge case in HNSW and introduces opt‑in, per‑query controls that let users dial recall vs latency safely and predictably. I’ve had this running in production for a while and thought it worth proposing to main.

Summary
- Fix incorrect early termination in the HNSW bottom layer that could stop before collecting k neighbours.
- Extend similar_to with optional per‑query ef and distance_threshold (string or JSON‑like fourth argument).
- Backwards compatible: default 3‑arg behaviour of similar_to is unchanged.
Motivation
- In narrow probes, the bottom‑layer search could exit at a local minimum before collecting k, hurting recall.
- No per‑query ef meant recall vs latency trade‑offs required global tuning or inflating k (and downstream work).
- This PR corrects the termination logic and adds opt‑in knobs so users can increase exploration only when needed.
Changes (key files)
- tok/hnsw/persistent_hnsw.go: fix early termination, add SearchWithOptions/SearchWithUidAndOptions, apply ef override at upper layers and max(k, ef) at bottom layer, apply distance_threshold in the metric domain (Euclidean squared internally, cosine as 1 − sim).
- tok/index/index.go: add VectorIndexOptions and OptionalSearchOptions (non‑breaking).
- worker/task.go: parse optional fourth argument to similar_to (ef, distance_threshold), thread options, route to optional methods when provided, guard zero/negative k.
- tok/index/search_path.go: add SearchPathResult helper.
- Tests: tok/hnsw/ef_recall_test.go adds
  - TestHNSWSearchEfOverrideImprovesRecall
  - TestHNSWDistanceThreshold_Euclidean
  - TestHNSWDistanceThreshold_Cosine
- CHANGELOG.md: Unreleased entry for HNSW fix and per‑query options.
Backwards compatibility
- No default behaviour changes. The three‑argument similar_to(attr, k, vector_or_uid) is unchanged.
- ef and distance_threshold are optional, unsupported metrics safely ignore the threshold.
Performance
- No overhead without options.
- With ef, bottom‑layer candidate size becomes max(k, ef) (as in HNSW), cost scales accordingly.
- Threshold filtering is a cheap pass over candidates, squaring Euclidean thresholds avoids extra roots.
Rationale and alignment
- Matches HNSW semantics: ef_search controls exploration/recall, k controls output size.
- Aligns with Typesense’s per‑query ef and distance_threshold semantics for familiarity.

Checklist

Code compiles correctly and linting passes locally
For all code changes, an entry added to the CHANGELOG.md describing this PR
Tests added for new functionality / regression tests for the bug fix
For public APIs/new features, docs PR will be prepared and linked here after initial review

darkcoderrises · 2025-10-19T20:38:52Z

Hey, Could you send a copy of the PR to https://github.com/predictable-labs/dgraph/pulls also? I am the author of the bugs that you have fixed. I have made a fork of the repo to be able to make changes myself. I have fixed some bugs in the vector index that makes it much faster.

joelamming · 2025-10-19T21:10:39Z

Hey, Could you send a copy of the PR to https://github.com/predictable-labs/dgraph/pulls also? I am the author of the bugs that you have fixed. I have made a fork of the repo to be able to make changes myself. I have fixed some bugs in the vector index that makes it much faster.

Cheers for the quick look. I’ve opened a mirror PR against your fork here: predictable-labs#17

Your branch is ahead of upstream so it shows conflicts in tok/hnsw/persistent_hnsw.go and worker/task.go. You're welcome to cherrypick from my branch into your tree. I could otherwise rebase onto predictable-labs/dgraph:main and resolve the conflicts on my side (will be tomorrow).

Tests for both parts are in tok/hnsw/ef_recall_test.go, and CHANGELOG.md has an Unreleased entry. Let me know which route you prefer and I'll see about implementing tomorrow.

darkcoderrises · 2025-10-20T09:10:19Z

@joelamming Thanks a lot for sending your changes here also. I can take a look and convert your changes. Basically we have introduced a new Partioned HNSW which is supposed to be much faster. We also figured out a race condition in the hnsw tree which led to less recall. We were able to improve accuracy significantly even without your changes in the new hnsw.

matthewmcneely · 2025-10-23T21:33:23Z

@joelamming Thanks for this contribution, truly amazing. Could I ask that you create an integration test in /query/vector/vector_test.go that illustrates the new arguments/functionality? This will help with documentation and general understanding of this new functionality.

Also, a few nits:

Can you not use the term "neighbour", instead "neighbor"? I know we do it wrong on this side of the pond, but other places in the codebase use the North American version
Can you install/run trunk fmt against your changes? That's the only (required) failing check

Thanks!

joelamming · 2025-10-24T23:16:09Z

Thanks for the nudge -- I’ve pushed a few follow-up commits:

Added TestSimilarToOptionsIntegration in query/vector/vector_test.go. It tries both the "ef=…" string syntax and the JSON-style {distance_threshold: …} options and passes against a fresh compose cluster (go test ./query/vector -run TestSimilarToOptionsIntegration -tags=integration -count=1)
Normalised the spelling to “neighbor” in the new vector/HNSW paths (tok/hnsw/persistent_hnsw.go, tok/index/search_path.go)
Ran trunk fmt over the tree, checks are clean now

While tracking down the integration failure I also found that Euclidean distance_threshold was being compared to the squared distance. Never surfaced in production as we've rarely used it but it can be an important part of full featured vector search. I fixed the filtering to use the raw metric distance and refreshed the regression tests (TestHNSWDistanceThreshold_Euclidean). Everything including the integration test now lines up with the expected semantics

joelamming · 2025-10-26T10:03:34Z

@matthewmcneely I see the CI failures from the GitHub Actions runs. I've identified and fixed the parser regression introduced by the similar_to JSON argument support.

The lexer changes to support similar_to(pred, k, vec, {ef: 64, distance_threshold: 0.45}) altered error handling for a specific class of malformed queries, causing three parser tests to fail. Not one that surfaced for us in production but makes sense to keep in dgraph main.

Yesterday:

Fixed the lexer/parser interaction to properly validate braces
Updated affected tests with documentation explaining the (minor) error message trade-off
Added comprehensive test coverage for edge cases
All local tests passing: go test ./dql

Today I'm running the full CI-equivalent suite on an ubuntu-noble-24.04-amd64 EC2 runner (same config as the Actions hosts). For each harness run I'm doing the standard ./t -r setup/teardown and recopying dgraph into $(go env GOPATH)/bin. As we speak:

go clean -testcache
./t -r
cp dgraph/dgraph "$(go env GOPATH)/bin/dgraph"
./t --suite=core
./t --suite=vector
./t -r
go clean -testcache
go test -v -timeout=90m -failfast -tags=integration2 ./...

Please hold off on reviewing until I've confirmed everything is clean on the EC2 instance and I've finished pushing the fixes. Should have results later today.

Will provide detailed docs on the architectural tradeoff in detail in the next push -- happy to discuss the approach once you have a chance to review.

Thanks for your patience!

joelamming · 2025-10-26T18:24:52Z

@matthewmcneely CI tests all pass on my end after bumping to the 32GB instance

Guarded similar_to's JSON literal parsing in the parser so only the 4th argument can hold
{...} while the lexer stays strict elsewhere
Restored legacy “Unrecognized character…” errors for non-similar_to stray braces
Updated parser tests to reflect the restored messaging and added coverage for the
supported/unsupported call shapes

Ran tests:

go test ./dql
./t -r && cp dgraph/dgraph "$(go env GOPATH)/bin/dgraph"
./t --suite=core
./t --suite=vector
./t --suite=load
./t --suite=systest
./t --suite=ldbc
go test -v -timeout=90m -failfast -tags=integration2 ./...
go test -v -timeout=120m -failfast -tags=upgrade
github.com/hypermodeinc/dgraph/v25/{acl,worker,query}
go clean -testcache between suites and recopying dgraph each time per EC2 workflow

Given the runner matches Actions (Ubuntu 24.04, 16 vCPU/32 GiB) I don’t anticipate surprises on CI

Many thanks!

joelamming · 2025-11-01T19:34:56Z

Hi @matthewmcneely

Just to follow up on this -- the new commits are pushed and it looks like the CI workflows are just waiting for approval to run

Thanks for your time!

matthewmcneely · 2025-11-02T18:11:56Z

CI workflows are just waiting for approval

We have some issues with our benchmarks repo following our switch of accounts. Waiting for that to be sorted out.

matthewmcneely · 2025-11-03T03:49:54Z

@joelamming Could you merge our main branch? If has fixes for paths in two CI actions that need updates. Thanks.

…ns, add tests

matthewmcneely · 2025-12-22T21:39:17Z

Hey @joelamming, Finally able to get back on this. I've been wrestling with the syntax for those optional parameters to the similar_to function. I'm sure you gave it quite a lot of thought too. Unfortunately, DQL has evolved with differing concepts for positional and optional arguments.

Would you be opposed to keeping the syntax more in-line with how our directives take optional arguments? Take @recurse(depth: 3, loop: true) for example, both arguments are optional. For similar_to, it might look like this:

similar_to(vpred, 3, $vec) # the current function syntax
similar_to(vpred, 3, $vec, ef: 2)
similar_to(vpred, 4, $vec, distance_threshold: 1.5)
similar_to(vpred, 4, $vec, ef: 64, distance_threshold: 0.5)
similar_to(vpred, 3, $vec, ef: $effort) #effort is a query variable

Pros:

This minimizes new concepts in the DQL syntax
This would allow the support of DQL query/block variables for those params (note the last example)

Cons:

You need to rewrite the parser

Let me know your thoughts. If you're too busy let me know and I'll have a stab at it.

joelamming · 2025-12-23T23:16:52Z

@matthewmcneely -- working on this now. Did indeed wrestle with the syntax and chose to keep in line with the original similar_to. But didn't seem quite in line with the rest of the codebase. Currently testing your suggestion. Will get back with some results/commits shortly

matthewmcneely · 2025-12-24T01:19:43Z

@joelamming Great, thanks for taking it on. BTW, are you OK with my adding your synthetic test to my BEIR eval benchmark in the dgraph-benchmarks repo?

joelamming · 2025-12-24T02:25:16Z

BTW, are you OK with my adding your synthetic test to my BEIR eval benchmark in the dgraph-benchmarks repo?

Of course -- please feel free to include.

I've pushed a commit that implements your suggestions. It passes my local tests and the synthetic recall test. Should be good for CI. Thanks!

…uct metrics (#9559) ### Summary This PR fixes a critical bug in HNSW vector search where **cosine similarity and dot product metrics returned incorrect results**. The search algorithm was treating all metrics as distance metrics (lower is better), causing similarity metrics (higher is better) to return the *worst* matches instead of the best. ### Problem The HNSW implementation had two issues with similarity-based metrics: 1. **Search phase**: The candidate heap in persistent_hnsw.go::searchPersistentLayer always used a min-heap, which pops the lowest value first. For similarity metrics where higher values are better, this caused the algorithm to explore the worst candidates first and terminate prematurely. 2. **Edge pruning phase**: The helper.go::addNeighbors function used a fixed comparison (`>`) when pruning edges, which is correct for distance metrics but inverted for similarity metrics. This resulted in keeping the worst edges instead of the best. ### Root Cause The original code assumed all metrics behave like distance metrics: ```go // Always used min-heap (pops lowest first) candidateHeap := *buildPersistentHeapByInit(elements) // Edge pruning always used > comparison compare: func(i, j uint64) bool { return ph.distance_betw(..., i, ...) > ph.distance_betw(..., j, ...) } ``` For **Euclidean distance**, lower values = better matches → min-heap is correct. For **Cosine/DotProduct similarity**, higher values = better matches → need max-heap. ### Solution #### 1. Added candidateHeap interface with metric-aware heap selection ```go type candidateHeap[T c.Float] interface { Len() int Pop() minPersistentHeapElement[T] Push(minPersistentHeapElement[T]) PopLast() minPersistentHeapElement[T] } func buildCandidateHeap[T c.Float](array []minPersistentHeapElement[T], isSimilarityMetric bool) candidateHeap[T] { if isSimilarityMetric { return &maxHeapWrapper[T]{...} // Pops highest first } return &minHeapWrapper[T]{...} // Pops lowest first } ``` #### 2. Added isSimilarityMetric flag to SimilarityType ```go type SimilarityType[T c.Float] struct { // ... existing fields isSimilarityMetric bool // true for cosine, dotproduct; false for euclidean } ``` #### 3. Fixed edge pruning comparison in addNeighbors ```go compare: func(i, j uint64) bool { distI := ph.distance_betw(ctx, tc, uuid, i, &inVec, &outVec) distJ := ph.distance_betw(ctx, tc, uuid, j, &inVec, &outVec) return !ph.simType.isBetterScore(distI, distJ) } ``` ### Files Changed | File | Changes | |------|---------| | tok/hnsw/heap.go | Added candidateHeap interface, minHeapWrapper, maxHeapWrapper, and buildCandidateHeap factory | | tok/hnsw/helper.go | Added isSimilarityMetric field to SimilarityType; fixed edge pruning comparison | | tok/hnsw/persistent_hnsw.go | Updated searchPersistentLayer to use metric-aware candidate heap | | tok/hnsw/persistent_hnsw_test.go | Added comprehensive unit tests for heap behavior and search correctness | ### Testing Added new tests covering: - TestCandidateHeapMinHeap: Verifies min-heap pops in ascending order - TestCandidateHeapMaxHeap: Verifies max-heap pops in descending order - TestCandidateHeapPushPop: Tests Push/Pop operations for both heap types - TestCandidateHeapPopLast: Tests PopLast for both types - TestSimilarityTypeIsSimilarityMetric: Verifies flag is set correctly for each metric - TestSearchReturnsCorrectOrderForAllMetrics: End-to-end test for Euclidean, Cosine, and DotProduct - TestEdgePruningKeepsBestEdges: Verifies edge pruning keeps best edges for each metric ### Performance Note This fix builds on PR #9514 which corrected the early termination condition. Together, these changes ensure HNSW search explores the correct number of candidates and returns properly ordered results. Users experiencing slower insert/search times compared to v25.1.0 can tune performance by lowering efConstruction and efSearch parameters when creating your vector indexes. Lower values trade recall for speed. The default values (efConstruction=128, efSearch=64) prioritize recall. ### GenAI Notice Parts of this implementation and all of the testing was generated using Claude Opus 4.5 (thinking). ### Checklist - [x] The PR title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/#summary) syntax, leading with `fix:`, `feat:`, `chore:`, `ci:`, etc. - [x] Code compiles correctly and linting (via trunk) passes locally - [x] Tests added for new functionality, or regression tests for bug fixes added as applicable Fixes #9558 ### Benchmarks Our BEIR SciFact Information Retrieval Benchmarks now show recall rates close to or exceeding acceptable and excellent performance for all metrics. ``` ============================================================================================================================================ NDCG@k Comparison ============================================================================================================================================ NDCG@1: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████░░░░░░░░░░░░ 0.6200 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████░░░░░░░░░░░░░░░ 0.5200 Dgraph v25.1.0 (euclidean) ███████████████░░░░░░░░░░░░░░░ 0.5000 Dgraph v25.1.0 (cosine) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2767 Dgraph v25.1.0 (dotproduct) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2867 Dgraph staged-fix (euclidean) ███████████████░░░░░░░░░░░░░░░ 0.5233 Dgraph staged-fix (cosine) ███████████████░░░░░░░░░░░░░░░ 0.5300 Dgraph staged-fix (dotproduct) ███████████████░░░░░░░░░░░░░░░ 0.5167 NDCG@3: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ████████████████████░░░░░░░░░░ 0.6700 BEIR Acceptable (Acceptable baseline (384-dim)) ██████████████████░░░░░░░░░░░░ 0.6000 Dgraph v25.1.0 (euclidean) ████████████████░░░░░░░░░░░░░░ 0.5588 Dgraph v25.1.0 (cosine) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3043 Dgraph v25.1.0 (dotproduct) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3164 Dgraph staged-fix (euclidean) █████████████████░░░░░░░░░░░░░ 0.5918 Dgraph staged-fix (cosine) █████████████████░░░░░░░░░░░░░ 0.5957 Dgraph staged-fix (dotproduct) █████████████████░░░░░░░░░░░░░ 0.5830 NDCG@5: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ████████████████████░░░░░░░░░░ 0.6900 BEIR Acceptable (Acceptable baseline (384-dim)) ██████████████████░░░░░░░░░░░░ 0.6300 Dgraph v25.1.0 (euclidean) █████████████████░░░░░░░░░░░░░ 0.5858 Dgraph v25.1.0 (cosine) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3197 Dgraph v25.1.0 (dotproduct) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3290 Dgraph staged-fix (euclidean) ██████████████████░░░░░░░░░░░░ 0.6168 Dgraph staged-fix (cosine) ██████████████████░░░░░░░░░░░░ 0.6240 Dgraph staged-fix (dotproduct) ██████████████████░░░░░░░░░░░░ 0.6081 NDCG@10: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) █████████████████████░░░░░░░░░ 0.7000 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████████░░░░░░░░░░░ 0.6500 Dgraph v25.1.0 (euclidean) ██████████████████░░░░░░░░░░░░ 0.6118 Dgraph v25.1.0 (cosine) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3305 Dgraph v25.1.0 (dotproduct) ██████████░░░░░░░░░░░░░░░░░░░░ 0.3423 Dgraph staged-fix (euclidean) ███████████████████░░░░░░░░░░░ 0.6461 Dgraph staged-fix (cosine) ███████████████████░░░░░░░░░░░ 0.6505 Dgraph staged-fix (dotproduct) ███████████████████░░░░░░░░░░░ 0.6369 NDCG@100: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) █████████████████████░░░░░░░░░ 0.7200 BEIR Acceptable (Acceptable baseline (384-dim)) ████████████████████░░░░░░░░░░ 0.6800 Dgraph v25.1.0 (euclidean) ███████████████████░░░░░░░░░░░ 0.6418 Dgraph v25.1.0 (cosine) ██████████░░░░░░░░░░░░░░░░░░░░ 0.3445 Dgraph v25.1.0 (dotproduct) ██████████░░░░░░░░░░░░░░░░░░░░ 0.3555 Dgraph staged-fix (euclidean) ████████████████████░░░░░░░░░░ 0.6794 Dgraph staged-fix (cosine) ████████████████████░░░░░░░░░░ 0.6849 Dgraph staged-fix (dotproduct) ████████████████████░░░░░░░░░░ 0.6707 ============================================================================================================================================ MAP@k Comparison ============================================================================================================================================ MAP@1: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████░░░░░░░░░░░░ 0.6000 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████░░░░░░░░░░░░░░░ 0.5000 Dgraph v25.1.0 (euclidean) ██████████████░░░░░░░░░░░░░░░░ 0.4812 Dgraph v25.1.0 (cosine) ███████░░░░░░░░░░░░░░░░░░░░░░░ 0.2586 Dgraph v25.1.0 (dotproduct) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2747 Dgraph staged-fix (euclidean) ███████████████░░░░░░░░░░░░░░░ 0.5046 Dgraph staged-fix (cosine) ███████████████░░░░░░░░░░░░░░░ 0.5112 Dgraph staged-fix (dotproduct) ██████████████░░░░░░░░░░░░░░░░ 0.4979 MAP@3: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ███████████████████░░░░░░░░░░░ 0.6400 BEIR Acceptable (Acceptable baseline (384-dim)) █████████████████░░░░░░░░░░░░░ 0.5700 Dgraph v25.1.0 (euclidean) ████████████████░░░░░░░░░░░░░░ 0.5357 Dgraph v25.1.0 (cosine) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2883 Dgraph v25.1.0 (dotproduct) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3022 Dgraph staged-fix (euclidean) ████████████████░░░░░░░░░░░░░░ 0.5663 Dgraph staged-fix (cosine) █████████████████░░░░░░░░░░░░░ 0.5707 Dgraph staged-fix (dotproduct) ████████████████░░░░░░░░░░░░░░ 0.5579 MAP@5: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ███████████████████░░░░░░░░░░░ 0.6600 BEIR Acceptable (Acceptable baseline (384-dim)) █████████████████░░░░░░░░░░░░░ 0.5900 Dgraph v25.1.0 (euclidean) ████████████████░░░░░░░░░░░░░░ 0.5544 Dgraph v25.1.0 (cosine) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2993 Dgraph v25.1.0 (dotproduct) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3113 Dgraph staged-fix (euclidean) █████████████████░░░░░░░░░░░░░ 0.5838 Dgraph staged-fix (cosine) █████████████████░░░░░░░░░░░░░ 0.5902 Dgraph staged-fix (dotproduct) █████████████████░░░░░░░░░░░░░ 0.5755 MAP@10: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ████████████████████░░░░░░░░░░ 0.6700 BEIR Acceptable (Acceptable baseline (384-dim)) ██████████████████░░░░░░░░░░░░ 0.6000 Dgraph v25.1.0 (euclidean) █████████████████░░░░░░░░░░░░░ 0.5676 Dgraph v25.1.0 (cosine) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3045 Dgraph v25.1.0 (dotproduct) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3175 Dgraph staged-fix (euclidean) █████████████████░░░░░░░░░░░░░ 0.5987 Dgraph staged-fix (cosine) ██████████████████░░░░░░░░░░░░ 0.6035 Dgraph staged-fix (dotproduct) █████████████████░░░░░░░░░░░░░ 0.5900 MAP@100: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ████████████████████░░░░░░░░░░ 0.6800 BEIR Acceptable (Acceptable baseline (384-dim)) ██████████████████░░░░░░░░░░░░ 0.6100 Dgraph v25.1.0 (euclidean) █████████████████░░░░░░░░░░░░░ 0.5746 Dgraph v25.1.0 (cosine) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3074 Dgraph v25.1.0 (dotproduct) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3203 Dgraph staged-fix (euclidean) ██████████████████░░░░░░░░░░░░ 0.6060 Dgraph staged-fix (cosine) ██████████████████░░░░░░░░░░░░ 0.6113 Dgraph staged-fix (dotproduct) █████████████████░░░░░░░░░░░░░ 0.5977 ============================================================================================================================================ RECALL@k Comparison ============================================================================================================================================ Recall@1: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████░░░░░░░░░░░░ 0.6000 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████░░░░░░░░░░░░░░░ 0.5000 Dgraph v25.1.0 (euclidean) ██████████████░░░░░░░░░░░░░░░░ 0.4812 Dgraph v25.1.0 (cosine) ███████░░░░░░░░░░░░░░░░░░░░░░░ 0.2586 Dgraph v25.1.0 (dotproduct) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2747 Dgraph staged-fix (euclidean) ███████████████░░░░░░░░░░░░░░░ 0.5046 Dgraph staged-fix (cosine) ███████████████░░░░░░░░░░░░░░░ 0.5112 Dgraph staged-fix (dotproduct) ██████████████░░░░░░░░░░░░░░░░ 0.4979 Recall@3: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) █████████████████████░░░░░░░░░ 0.7300 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████████░░░░░░░░░░░ 0.6500 Dgraph v25.1.0 (euclidean) █████████████████░░░░░░░░░░░░░ 0.5984 Dgraph v25.1.0 (cosine) █████████░░░░░░░░░░░░░░░░░░░░░ 0.3248 Dgraph v25.1.0 (dotproduct) ██████████░░░░░░░░░░░░░░░░░░░░ 0.3377 Dgraph staged-fix (euclidean) ███████████████████░░░░░░░░░░░ 0.6384 Dgraph staged-fix (cosine) ███████████████████░░░░░░░░░░░ 0.6401 Dgraph staged-fix (dotproduct) ██████████████████░░░░░░░░░░░░ 0.6284 Recall@5: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ███████████████████████░░░░░░░ 0.7900 BEIR Acceptable (Acceptable baseline (384-dim)) █████████████████████░░░░░░░░░ 0.7200 Dgraph v25.1.0 (euclidean) ███████████████████░░░░░░░░░░░ 0.6638 Dgraph v25.1.0 (cosine) ██████████░░░░░░░░░░░░░░░░░░░░ 0.3632 Dgraph v25.1.0 (dotproduct) ███████████░░░░░░░░░░░░░░░░░░░ 0.3697 Dgraph staged-fix (euclidean) ████████████████████░░░░░░░░░░ 0.6988 Dgraph staged-fix (cosine) █████████████████████░░░░░░░░░ 0.7088 Dgraph staged-fix (dotproduct) ████████████████████░░░░░░░░░░ 0.6888 Recall@10: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) █████████████████████████░░░░░ 0.8400 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████████████░░░░░░░ 0.7800 Dgraph v25.1.0 (euclidean) ██████████████████████░░░░░░░░ 0.7368 Dgraph v25.1.0 (cosine) ███████████░░░░░░░░░░░░░░░░░░░ 0.3950 Dgraph v25.1.0 (dotproduct) ████████████░░░░░░░░░░░░░░░░░░ 0.4074 Dgraph staged-fix (euclidean) ███████████████████████░░░░░░░ 0.7808 Dgraph staged-fix (cosine) ███████████████████████░░░░░░░ 0.7834 Dgraph staged-fix (dotproduct) ███████████████████████░░░░░░░ 0.7701 Recall@100: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ████████████████████████████░░ 0.9500 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████████████████░░░ 0.9000 Dgraph v25.1.0 (euclidean) ██████████████████████████░░░░ 0.8717 Dgraph v25.1.0 (cosine) █████████████░░░░░░░░░░░░░░░░░ 0.4589 Dgraph v25.1.0 (dotproduct) █████████████░░░░░░░░░░░░░░░░░ 0.4658 Dgraph staged-fix (euclidean) ████████████████████████████░░ 0.9350 Dgraph staged-fix (cosine) ████████████████████████████░░ 0.9417 Dgraph staged-fix (dotproduct) ███████████████████████████░░░ 0.9250 ============================================================================================================================================ PRECISION@k Comparison ============================================================================================================================================ Precision@1: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████░░░░░░░░░░░░ 0.6200 BEIR Acceptable (Acceptable baseline (384-dim)) ███████████████░░░░░░░░░░░░░░░ 0.5200 Dgraph v25.1.0 (euclidean) ███████████████░░░░░░░░░░░░░░░ 0.5000 Dgraph v25.1.0 (cosine) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2767 Dgraph v25.1.0 (dotproduct) ████████░░░░░░░░░░░░░░░░░░░░░░ 0.2867 Dgraph staged-fix (euclidean) ███████████████░░░░░░░░░░░░░░░ 0.5233 Dgraph staged-fix (cosine) ███████████████░░░░░░░░░░░░░░░ 0.5300 Dgraph staged-fix (dotproduct) ███████████████░░░░░░░░░░░░░░░ 0.5167 Precision@3: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████████████████ 0.2700 BEIR Acceptable (Acceptable baseline (384-dim)) █████████████████████████░░░░░ 0.2300 Dgraph v25.1.0 (euclidean) ████████████████████████░░░░░░ 0.2178 Dgraph v25.1.0 (cosine) █████████████░░░░░░░░░░░░░░░░░ 0.1211 Dgraph v25.1.0 (dotproduct) █████████████░░░░░░░░░░░░░░░░░ 0.1211 Dgraph staged-fix (euclidean) █████████████████████████░░░░░ 0.2311 Dgraph staged-fix (cosine) █████████████████████████░░░░░ 0.2322 Dgraph staged-fix (dotproduct) █████████████████████████░░░░░ 0.2278 Precision@5: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████████████████ 0.1800 BEIR Acceptable (Acceptable baseline (384-dim)) ██████████████████████████░░░░ 0.1600 Dgraph v25.1.0 (euclidean) ████████████████████████░░░░░░ 0.1480 Dgraph v25.1.0 (cosine) █████████████░░░░░░░░░░░░░░░░░ 0.0827 Dgraph v25.1.0 (dotproduct) █████████████░░░░░░░░░░░░░░░░░ 0.0807 Dgraph staged-fix (euclidean) █████████████████████████░░░░░ 0.1553 Dgraph staged-fix (cosine) ██████████████████████████░░░░ 0.1573 Dgraph staged-fix (dotproduct) █████████████████████████░░░░░ 0.1533 Precision@10: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ██████████████████████████████ 0.1000 BEIR Acceptable (Acceptable baseline (384-dim)) ██████████████████████████░░░░ 0.0900 Dgraph v25.1.0 (euclidean) █████████████████████████░░░░░ 0.0837 Dgraph v25.1.0 (cosine) █████████████░░░░░░░░░░░░░░░░░ 0.0453 Dgraph v25.1.0 (dotproduct) █████████████░░░░░░░░░░░░░░░░░ 0.0447 Dgraph staged-fix (euclidean) ██████████████████████████░░░░ 0.0887 Dgraph staged-fix (cosine) ██████████████████████████░░░░ 0.0887 Dgraph staged-fix (dotproduct) ██████████████████████████░░░░ 0.0873 Precision@100: -------------------------------------------------------------------------------------------------------------------------------------------- BEIR Excellent (State-of-the-art baseline (768-dim)) ████████████████████████████░░ 0.0100 BEIR Acceptable (Acceptable baseline (384-dim)) ████████████████████████████░░ 0.0100 Dgraph v25.1.0 (euclidean) ███████████████████████████░░░ 0.0100 Dgraph v25.1.0 (cosine) ██████████████░░░░░░░░░░░░░░░░ 0.0053 Dgraph v25.1.0 (dotproduct) ██████████████░░░░░░░░░░░░░░░░ 0.0051 Dgraph staged-fix (euclidean) █████████████████████████████░ 0.0106 Dgraph staged-fix (cosine) ██████████████████████████████ 0.0107 Dgraph staged-fix (dotproduct) █████████████████████████████░ 0.0105 ``` --------- Co-authored-by: Joe Lamming <191030909+joelamming@users.noreply.github.com>

HNSW fixes and tests

f31851a

joelamming requested a review from a team October 19, 2025 20:30

Update comment in persistent_hnsw.go

16e8ead

Merge branch 'hypermodeinc:main' into main

408ec42

Fix HNSW distance threshold filtering

69c1eb0

Working on CI test failures

2c9c391

Merge branch 'dgraph-io:main' into main

b508ec2

joelamming and others added 9 commits November 3, 2025 08:51

Merge branch 'dgraph-io:main' into main

eb1ecc2

Merge branch 'dgraph-io:main' into main

1e01940

Merge branch 'main' into main

9db5bc7

fix: polish PR - licence header, line lengths, error on unknown optio…

4af9bc8

…ns, add tests

Merge branch 'main' into main

a2b3ea7

Merge branch 'main' into main

a934b0e

Merge branch 'main' into main

8d355a8

Merge branch 'main' into main

1e05072

Merge branch 'main' into main

c1dd555

joelamming requested a review from a team as a code owner December 17, 2025 00:40

similar_to options improvement

ecee550

matthewmcneely approved these changes Dec 29, 2025

View reviewed changes

matthewmcneely merged commit 2c84a01 into dgraph-io:main Dec 29, 2025
24 of 25 checks passed

matthewmcneely mentioned this pull request Jan 8, 2026

fix(vector): Fix similarity-based HNSW search for cosine and dot product metrics #9559

Merged

3 tasks

feat(vector/hnsw): add per‑query ef and distance_threshold to similar_to, fix early termination #9514

feat(vector/hnsw): add per‑query ef and distance_threshold to similar_to, fix early termination #9514

Uh oh!

Conversation

joelamming commented Oct 19, 2025

Uh oh!

darkcoderrises commented Oct 19, 2025

Uh oh!

joelamming commented Oct 19, 2025

Uh oh!

darkcoderrises commented Oct 20, 2025

Uh oh!

matthewmcneely commented Oct 23, 2025

Uh oh!

joelamming commented Oct 24, 2025

Uh oh!

joelamming commented Oct 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

joelamming commented Oct 26, 2025

Ran tests:

Uh oh!

joelamming commented Nov 1, 2025

Uh oh!

matthewmcneely commented Nov 2, 2025

Uh oh!

matthewmcneely commented Nov 3, 2025

Uh oh!

matthewmcneely commented Dec 22, 2025

Uh oh!

joelamming commented Dec 23, 2025

Uh oh!

matthewmcneely commented Dec 24, 2025

Uh oh!

joelamming commented Dec 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

3 participants

joelamming commented Oct 26, 2025 •

edited

Loading

joelamming commented Dec 24, 2025 •

edited

Loading