NornicDB version history

- Added GPU-accelerated HNSW construction for CUDA and Vulkan - Fixed multi-MATCH relationship variable binding issue in Cypher queries - Improved Cypher map key parsing for keys containing colons - Restored Bolt 4.x datetime/time encoding compatibility

- Upgraded llama.cpp and refreshed dependencies across Go modules and UI npm packages. - Fixed Cypher Graphiti ingest correctness and improved performance. - Expanded test coverage for Graphiti and Graphify scenarios. - Regenerated GraphQL code to accommodate upstream library changes.

## [v1.1.9] - [Go](https://www.youtube.com/watch?v=AKZ1oa-LbjQ) - 6.24.2026 (for Ez, who never stops racing) Full Graphiti support - special thanks to @joefrancia for all the testing and co-profiling with me. includes #227 and Added followup fixes an edge condition race on the cache and audit logging race. ### Changed - **Cypher hot paths were tightened across execution, planning, and clause handling**: - structural expression evaluation now uses dedicated fast paths for common operators and property/literal forms; - hot candidate paths compile `WHERE` predicates earlier; - `CALL`-tail retrieval and composite subquery binding preservation were both corrected and optimized; - keyword scanning now caches results and narrows stale-index fallback behavior. - **Vector-search execution was made more selective and more passive during ingest**: - query-time vector index warmups are avoided on read paths; - vector fast paths now use owned search-readiness gates instead of forcing full warmup semantics; - Graphiti ingest keeps vector reads and writes passive where appropriate; - exact file-store candidates are now included for property-vector node querie

## [v1.1.7] - [Black Math](https://www.youtube.com/watch?v=HzjNAnEfvxc) - 6.16.2026 ## You can now install and run using Homebrew! `brew tap --trust orneryd/nornicdb && brew install nornicdb && brew services start nornicdb` ### Changed - **llama.cpp bundle/build refresh** - Updated bundled llama.cpp artifacts and synchronized build surfaces (Makefile, Windows batch scripts, shell/PowerShell build scripts, and Docker llama images) to the current pinned integration version. - **Embedding model defaults hardened for local GGUF stability**: - Embedding context features now default `flash_attn` to disabled (`0`) instead of auto for safer startup across Metal/CUDA backends. - `NORNICDB_EMBEDDING_FLASH_ATTN=0` is treated as an explicit override (not silently ignored). - **GPU-assisted HNSW construction path retuned for lower latency and lower allocation pressure**: - Build-time candidate generation now uses a Metal batched matrix/top-k primitive plus graph-beam expansion instead of per-query GPU search calls. - Hot-path build code now reuses graph snapshot scratch state and batch flattening buffers, avoiding repeated per-batch/per-iteration map/slice churn.

## [v1.1.6] - [Cherry Bomb](https://www.youtube.com/watch?v=_EBvXpjudf8) - 2026-06-12 Release focused on Neo4j/Graphiti compatibility hardening, vector-search performance correctness, and a new offline admin import/export workflow. ### Added - **`nornicdb-admin` offline tooling** for full-database import/export workflows: - `database import full` with multi-file node/relationship sources, deterministic reports, structured exit codes, and fail-fast source validation. - Neo4j CSV package round-trip support, including schema export/import surfaces and full-schema package generation for operational migrations. - **Operations documentation for admin workflows**, including import/export usage, package layout, and recovery behavior. ### Changed - **Cypher DDL and clause parsing migrated further from regex matching to keyword/token scanning** across schema/index and helper paths, improving parser determinism and reducing fragile query-shape coupling. - **Configuration/docs alignment for async writes**: YAML and environment-variable behavior is now documented and validated consistently for production operators. ### Fixed - **Graphiti-blocking write-path correctn

## WE PUT THE DEMON IN PRI*DE MON*TH <p align="center"> <img src="https://raw.githubusercontent.com/orneryd/NornicDB/refs/heads/main/docs/assets/logos/nornicdb-logo-pride.svg" alt="NornicDB Logo" width="200"/> </p> ## [v1.1.5] - [Bicycle](https://www.youtube.com/watch?v=uzZxbF2Zdu0) - 2026-06-03 Post-`v1.1.4` stabilization focused on Cypher/Bolt correctness, storage resilience, and deterministic behavior under real Neo4j-driver query shapes. This range also includes broad coverage expansion; all test/coverage work is bundled under a single item below. ### Changed - **Storage embedding persistence now shards oversized chunk payloads in Badger.** Large embedding chunks are written/read using size-aware shard keys with backward-compatible decode logic for legacy single-value chunks. This prevents value-size failures on valid large embeddings while keeping existing data readable. - **Search query chunk scoring now uses best-match chunk score instead of average score.** This improves relevance for multi-chunk documents where one strong matching chunk should dominate ranking. - **Version metadata updated** to reflect the current post-`v1.1.4` development state. ###

## [v1.1.4] - [Barracuda](https://www.youtube.com/watch?v=PeMvMNpvB5M) - 2026-05-29 Hotfix v1.1.3-v1.1.4: for merge prop preservation with nullable values fixes #186 and fixes the publishing for docker images. bumping version to make sure it's all in sync. Maintenance release: **llama.cpp upgraded to b9410** with configurable per-model context features, plus several storage and Bolt correctness fixes discovered through expanded test coverage. No on-disk format changes; existing databases upgrade transparently. ### Added - **Per-model llama.cpp context feature passthrough.** Each model domain (embedding, rerank, Heimdall) now accepts env-driven llama.cpp context parameters so operators can tune models that require non-default settings (e.g. MTP-trained models, CLS pooling, rank pooling): - Embedding: `NORNICDB_EMBEDDING_CTX_TYPE`, `NORNICDB_EMBEDDING_POOLING_TYPE`, `NORNICDB_EMBEDDING_ATTENTION_TYPE`, `NORNICDB_EMBEDDING_FLASH_ATTN` - Rerank: `NORNICDB_RERANK_CTX_TYPE`, `NORNICDB_RERANK_POOLING_TYPE`, `NORNICDB_RERANK_ATTENTION_TYPE`, `NORNICDB_RERANK_FLASH_ATTN` - Heimdall: `NORNICDB_HEIMDALL_CTX_TYPE`, `NORNICDB_HEIMDALL_POOLING_TYPE`, `NORNICDB_HEIMDALL_ATTE

## [v1.1.2] [Purple Haze](https://www.youtube.com/watch?v=WGoDaYjdfSg) - 2026-05-26 Headline release: **Bolt over WebSocket** lands end-to-end so browser-based Neo4j drivers connect to NornicDB without a proxy, and **per-database BM25 + vector index master switches** ship as a first-class memory and warmup-cost lever for multi-tenant deployments. Three independently reported Cypher correctness regressions (mcp-neo4j-memory) are fixed with deeply-asserted parity against Neo4j 5.x DDL and Lucene wildcard semantics. A profile-led overhaul of the shortestPath traversal stack drops latency ~400× on the demo workload. No on-disk format changes; existing `v1.1.x` databases upgrade transparently. # shortestPath benchmark — 500K nodes / ~3.97M edges Single-process traversal latency at scale. Generated by: ``` go test ./pkg/cypher/ -run TestLargeScaleShortestPath_HopBuckets \ -timeout 30m -largescale -v ``` Hardware: Apple M3 Max. Storage chain: BadgerDB → AsyncEngine → NamespacedEngine → StorageExecutor (the same chain the HTTP server uses). ## Setup - 500,000 `Star` nodes across 1,000 sectors (500 per sector) - 3,969,418 `HYPERLANE` edges (forward+reverse), avg ~

## [v1.1.1] - [Santeria](https://www.youtube.com/watch?v=AEYN5w4T_aM) - 2026-05-19 Patch release focused on multi-tenant MVCC isolation, MERGE concurrency correctness, and a discoverable consumer-facing documentation surface (skills, migration scripts, unified Bolt error shape). No on-disk format changes; existing `v1.1.0` databases upgrade transparently. ## Check out the demo on the hidden `http://localhost:7474/demo` route in your browser. click any two nodes to see the distance path light up and check the latency for hops <img width="1717" height="915" alt="Screenshot 2026-05-19 at 8 13 11 PM" src="https://github.com/user-attachments/assets/3e838d4b-884c-44c6-8cfa-9a853ce26b30" /> ### Added - **`/demo` galaxy route** — interactive 3D force-directed graph (lazy-loaded `3d-force-graph` + `three.js`) that procedurally seeds a `d3_demo` database with a Fibonacci-sphere sector layout, links sectors via gateway hyperlanes, and exposes click-two-stars-to-traverse so the deep multi-hop `shortestPath` path lights up in purple. A purple-themed HUD tracks live `shortestPath` latency only — seed and catalog calls are excluded so the bars reflect query speed, not page load. Al

## [v1.1.0] - [Trendsetter](https://www.youtube.com/watch?v=PyrX8c1HGZw) - 2026-05-15 # NornicDB v1.1.0 Release Notes > ⚠️ ⚠️ ⚠️ **READ BEFORE UPGRADING — STORAGE FORMAT CHANGES** ⚠️ ⚠️ ⚠️ > > This release ships **breaking storage changes** that require a one-time > in-place migration on first start. Once your data files are upgraded, > **older versions of NornicDB will not be able to read them**. > > 🦀 **Back up your data directory before upgrading.** 🦀 > > Run the database with the `--upgrade-storage` flag (or, in the macOS > installer, tick the upgrade checkbox on the first-run wizard, or in the > tray settings panel restart with the flag enabled). This is a rolling > upgrade — when it completes, the engine compacts the on-disk layout to > the new format and resumes normal operation. > > After this release, no further major storage overhauls are planned. 🤟 --- **Release tag:** `v1.1.0` **Range covered:** `v1.0.45..v1.1.0` **Combined repository delta:** 644 files changed, +170,459 / −9,816 lines across 91 non-merge commits. This document is a single rolled-up changelog for everything shipped in v1.1.0 across both the `v1.1.0-preview` (2026-05

## [v1.0.45] - [Go](https://youtu.be/TCYsY5B8hcQ?t=81) - 2026-05-08 ### Added - **Broader hot-path coverage for `UNWIND MATCH SET` updates**: - added a batched `UNWIND` + `MATCH` + `SET` hot path so more update-heavy Cypher write workloads can stay on the optimized execution path. ### Changed - **Browser query-results grid behavior**: - updated the web UI to `@ornery/ui-grid-*` `1.0.5`, bringing the latest resize and scroll tracking behavior from the new grid runtime. - query-results selection now relies on the grid's per-column opt-out behavior for the synthetic Select column, so hidden sort and group controls follow the legacy ui-grid behavior instead of rendering as disabled header buttons. - **Cypher write-path routing**: - tightened `MATCH ... SET` hot-path routing so optimized update execution only applies when uniqueness guarantees are present, reducing incorrect fast-path routing for ambiguous match shapes. - **Maintenance and dependency refreshes**: - refreshed UI grid dependencies from the `1.0.4` line to `1.0.5` and regenerated the UI lockfile to match the published package graph. - updated package metadata and peer-dependency wiring i