v0.7.0 — Sync, Tri-Modal Query, OPFS, Integrations

AgentDB v0.7.0

Major New Features

• CRDT-Based Sync Engine — Replicate AgentDB instances with Hybrid Logical Clocks, operation logs, and configurable conflict resolution (Last-Writer-Wins / First-Writer-Wins)
• Tri-Modal Hybrid Query — Blend vector ANN + graph traversal + FTS BM25 in a single query with tunable alpha/beta/gamma weights
• WASM OPFS Persistence — Browser databases now survive page reloads via the Origin Private File System API (load-save pattern with SQLite serialize/deserialize)
• LangChain Integration — langchain-agentdb package implementing VectorStore + BaseChatMessageHistory
• LlamaIndex Integration — llamaindex-agentdb package implementing BasePydanticVectorStore + BaseChatStore
• Ruby SDK — Full FFI bindings (40+ functions) with idiomatic Ruby API, RSpec test suite

CLI Improvements

• agentdb version — print version + target triple
• agentdb export <path> [--table <name>] — dump database as SQL statements
• agentdb completions --shell <bash|zsh|fish|powershell> — generate shell completions

Test Coverage

• 73 new async API tests (Rust)
• 91 new Node.js tests covering full API surface
• Python test suite (10 test classes)
• 16 sync module tests
• 8 tri-modal query tests

Dependency Migrations

• rand 0.8 → 0.10 (thread_rng → rng, gen → random)
• criterion 0.5 → 0.8 (removed deleted html_reports feature)

CI/CD Fixes

• Fixed Snap/WinGet/Chocolatey publish workflows (root cause: actions/checkout@v7 doesn't exist)
• crates.io publish: pinned MSRV 1.95.0, added --locked
• Coverage threshold raised to 80%, now includes async+ffi code paths
• Node.js smoke test class name bug fixed

Documentation

• ROADMAP.md aligned with actual v0.6.0 release content
• README roadmap table corrected
• Manual publish steps documented for Snap/WinGet/Chocolatey/crates.io

---

Full Changelog: v0.6.0...v0.7.0

Full Changelog: v0.6.0...v0.7.0

v0.6.0 — AI-Native Architecture

AI-Native Architecture for Autonomous Agents

AgentDB v0.6.0 introduces 5 new storage layers, an MCP server interface, and full multi-language binding coverage — making it a complete embedded database purpose-built for AI agents.

New Storage Layers

Layer	Purpose
Tool Registry	Register tools with JSON Schema parameters, log invocations with latency
Audit Log	Immutable append-only provenance trail with actor/action/reason
Context Window	Token-budgeted context management with priority and relevance scoring
Prompt Templates	Versioned templates with {{placeholder}} rendering and model hints
Data Labels	Privacy-by-design classification (PII, sensitive, internal, etc.)

MCP Server Interface

Full Model Context Protocol implementation:

• JSON-RPC 2.0 transport
• tools/list — 21 tools with JSON Schema input definitions
• tools/call — invoke any AgentDB operation
• resources/list / resources/read — database statistics

Multi-Language Bindings

All 5 new layers are exposed across:

• Rust (core) — direct API
• C FFI — 14 new functions in agentdb.h
• Node.js (napi-rs) — TypeScript declarations included
• Go (cgo) — idiomatic Go wrappers
• Java (JNI) — AgentDB.java with full Javadoc
• C# (P/Invoke) — AgentDB.cs with XML docs
• WASM (wasm-bindgen) — browser-ready, JSON interchange
• Async (Tokio) — AsyncToolStore, AsyncAuditStore, AsyncContextStore, AsyncPromptStore, AsyncLabelStore

Stats

• 13 storage layers total
• 231+ tests passing
• Zero clippy warnings
• Schema v5 with 6 new tables

Install

# Rust
cargo add datacules-agentdb

# Node.js
npm install @datacules/agentdb

# Python
pip install datacules-agentdb

Full Changelog: v0.5.3...v0.6.0

v0.5.3 — API Ergonomics & Serde

What's New

Added

• query_json_params(sql, params) — parameterized SQL queries on core, async, FFI, and Node.js APIs. Prevents SQL injection when interpolating user-supplied values.
• metadata parameter on create_workflow — FFI, Node.js, and Go bindings now accept optional metadata JSON.
• relation filter on graph traversal — FFI, Node.js, Go, Java, and C# bindings now accept an optional relation string to filter traversed edges.
• Serialize/Deserialize derives on all public data types (DbStats, VectorEntry, SearchResult, SearchOptions, BatchEntry, Node, Edge, TraversalOptions, TraversalResult, FtsResult, HybridResult, MessageSearchResult, Conversation, Message, Workflow, WorkflowStep, Trace).
• Clone derive on AgentDB — enables sharing the database handle across threads/tasks.

Fixed

• close() / Drop double-flush — calling close() then dropping no longer reindexes HNSW indexes twice.
• Vector search N+1 metadata queries — replaced per-result SQL lookups with a single batch WHERE id IN (...) query.
• agentdb_open not marked unsafe — now consistent with all other FFI exports.
• Clippy clean — zero warnings with --all-targets --features ffi.

Binding Parity

All new APIs are available in: Rust (core + async), C FFI, Node.js (napi-rs), Go (cgo), Java (JNI), C# (P/Invoke).

Full Changelog: v0.5.2...v0.5.3

Full Changelog: v0.5.0...v0.5.3

v0.5.0 — Full SDK parity

What's new in v0.5.0

API completeness pass — all SDKs now match the Rust core

9 new FFI operations exposed across Go, Node.js, Java, and C#:

• vector_delete — remove a vector by id
• drop_collection — drop a collection and all its data
• reindex — force HNSW rebuild
• graph_get_node — retrieve a single graph node
• graph_delete_node — remove a node and its edges
• graph_delete_edge — remove a specific directed edge
• fts_delete — remove a document from FTS
• fts_optimize — merge FTS index segments
• workflow_fail — mark a workflow as failed with an error message

hybridQuery filter support: all SDKs now accept an optional MongoDB-style filter_json parameter to narrow vector candidates during hybrid graph+vector queries.

9-field DbStats: collections, vectors, nodes, edges, conversations, messages, workflows, workflow_steps, traces — all SDKs updated.

Node.js additions: upsert_with_text (combined FTS+vector write), Collection.delete.

All 133 tests pass. MSRV: 1.85.0.

What's Changed

• chore(ci): bump actions/checkout from 4 to 7 by @dependabot[bot] in #19
• chore(ci): bump softprops/action-gh-release from 2 to 3 by @dependabot[bot] in #20
• chore(ci): bump peter-evans/repository-dispatch from 3 to 4 by @dependabot[bot] in #21
• chore(ci): bump actions/setup-node from 4 to 6 by @dependabot[bot] in #23
• chore(ci): bump codecov/codecov-action from 4 to 7 by @dependabot[bot] in #24
• chore(deps): bump @napi-rs/cli from 2.18.4 to 3.7.2 in /nodejs by @dependabot[bot] in #27
• chore(deps): bump tempfile from 3.14.0 to 3.27.0 by @dependabot[bot] in #28

New Contributors

• @dependabot[bot] made their first contribution in #19

Full Changelog: v0.4.5...v0.5.0

v0.4.5

Full Changelog: v0.4.4...v0.4.5

v0.4.4

Full Changelog: v0.4.3...v0.4.4

v0.4.3

Full Changelog: v0.4.2...v0.4.3

v0.4.2

Full Changelog: v0.4.1...v0.4.2

v0.4.1

Full Changelog: v0.4.0...v0.4.1

v0.4.0 — AI-Native Features + Multi-Language SDKs

AgentDB v0.4.0

AgentDB is now the complete embedded database for AI agents.

One file. Eight layers. Zero servers. Nine language SDKs.

---

New Features

AI-Native Layers

• Conversation threading — _adb_conversations + _adb_messages tables with full CRUD API (db.conversations())
• Workflow persistence — _adb_workflows + _adb_workflow_steps for durable multi-step agent tasks (db.workflows())
• Reasoning traces — _adb_traces with tree-structured chain-of-thought and recursive CTE traversal (db.traces())

Developer Experience

• Transaction API — db.transaction(|tx| { ... }) for ACID closures + db.execute_batch() for atomic multi-statement execution
• Interactive CLI shell — agentdb shell <path> or agentdb -i <path> — readline-style REPL with multi-line SQL, dot-commands (.stats, .collections, .inspect, .help, .quit)

Multi-Language SDKs

• Go (go/) — cgo wrapper, full FFI surface coverage
• Java (java/) — JNI wrapper with Maven POM, try-with-resources
• C# / .NET (dotnet/) — P/Invoke wrapper with NuGet project, IDisposable

Infrastructure

• Dockerfile — multi-stage build (rust:1.75 builder + debian:bookworm-slim runtime), OCI labels, VOLUME /data, <30MB final image
• Schema v2 — 7 new tables + 5 indexes (additive, backward-compatible)

---

10 Attributes — All Verified ✅

#	Attribute	Status
1	Serverless, embedded architecture	✅
2	Zero-configuration setup	✅
3	Single-file portable database format	✅
4	Cross-platform (Linux, Windows, macOS, Docker, WSL)	✅
5	Simple installation and packaging	✅
6	Standalone CLI with interactive shell	✅
7	ACID transactions, WAL, crash recovery	✅
8	Small footprint, minimal dependencies	✅
9	SDK integration (Rust, Python, JS, Go, Java, C#, C/C++, WASM)	✅
10	AI-native features (memory, conversations, vectors, search, workflows, traces)	✅

---

Install / Upgrade

# Rust
cargo install datacules-agentdb

# Python
pip install --upgrade datacules-agentdb

# Node.js
npm install @datacules/agentdb@latest

# Docker
docker pull ghcr.io/hvrcharon1/agentdb:v0.4.0
docker run -v $(pwd):/data agentdb shell mydb.agentdb

# Homebrew / Scoop / Chocolatey / Snap / WinGet
brew upgrade agentdb
scoop update agentdb
choco upgrade agentdb
snap refresh agentdb
winget upgrade Datacules.AgentDB

Full Changelog: v0.3.4...v0.4.0

Full Changelog: v0.3.4...v0.4.0