Skip to content

v2.0.0: The Agent-Native Retrieval Engine

Latest
Latest

Choose a tag to compare

View all tags
@atul-fusionpact atul-fusionpact released this 06 Mar 11:57
· 4 commits to main since this release
v2.0.0
423df19

Changelog

All notable changes to FusionPact will be documented in this file.

[2.0.0] - 2026-03-06

🚀 Major Release — The Agent-Native Retrieval Engine

This release transforms FusionPact from a vector database into a full hybrid retrieval engine combining vector search, reasoning-based tree retrieval, and agent memory.

Added

  • TreeIndex — Hierarchical document indexing with LLM-powered reasoning-based retrieval. Inspired by the insight that similarity ≠ relevance. Supports Markdown, HTML, and plain text.
  • HybridRetriever — Multi-strategy retrieval combining vector (HNSW), tree reasoning, and keyword (BM25) search with Reciprocal Rank Fusion.
  • Adaptive Retrieval — System learns which strategy works best for different query patterns.
  • Multi-Agent Orchestration — AgentOrchestrator for coordinating multiple AI agents with message routing, capability-based task delegation, and collaborative retrieval.
  • Shared Memory — Cross-agent knowledge sharing pool with access control.
  • Conversation Memory — Thread-aware chat history management per agent.
  • LLM Provider — Unified interface for Ollama, OpenAI, and Anthropic for tree reasoning.
  • HTTP API Server — RESTful API alongside MCP server.
  • Comprehensive Documentation — Full API reference, architecture docs, integration guides.
  • ATTRIBUTION.md — Clear attribution guidelines for users.

Changed

  • Repositioned from "Agent-Native Vector Database" to "Agent-Native Retrieval Engine"
  • Enhanced MCP server with hybrid retrieval and orchestration tools
  • Improved RAG pipeline with hybrid retrieval support
  • Better CLI with demo, serve, mcp, and bench commands

Architecture

FusionPact 2.0 Architecture:
├── Core: FusionEngine + HNSWIndex
├── Index: TreeIndex (reasoning-based)
├── Retrieval: HybridRetriever (vector + tree + keyword fusion)
├── Memory: AgentMemory (episodic + semantic + procedural + shared)
├── Orchestration: AgentOrchestrator (multi-agent)
├── RAG: RAGPipeline (one-click ingestion)
├── MCP: MCPServer (stdio + HTTP)
└── Embedders: Ollama, OpenAI, Mock

[1.0.0] - 2025-01-15

Initial Release

  • HNSW vector indexing
  • Agent Memory (episodic, semantic, procedural)
  • MCP server
  • RAG pipeline
  • Multi-tenancy
  • Ollama + OpenAI embedders

Built by [FusionPact Technologies Inc.](https://fusionpact.com)# Changelog

All notable changes to FusionPact will be documented in this file.

[2.0.0] - 2026-03-06

🚀 Major Release — The Agent-Native Retrieval Engine

This release transforms FusionPact from a vector database into a full hybrid retrieval engine combining vector search, reasoning-based tree retrieval, and agent memory.

Added

  • TreeIndex — Hierarchical document indexing with LLM-powered reasoning-based retrieval. Inspired by the insight that similarity ≠ relevance. Supports Markdown, HTML, and plain text.
  • HybridRetriever — Multi-strategy retrieval combining vector (HNSW), tree reasoning, and keyword (BM25) search with Reciprocal Rank Fusion.
  • Adaptive Retrieval — System learns which strategy works best for different query patterns.
  • Multi-Agent Orchestration — AgentOrchestrator for coordinating multiple AI agents with message routing, capability-based task delegation, and collaborative retrieval.
  • Shared Memory — Cross-agent knowledge sharing pool with access control.
  • Conversation Memory — Thread-aware chat history management per agent.
  • LLM Provider — Unified interface for Ollama, OpenAI, and Anthropic for tree reasoning.
  • HTTP API Server — RESTful API alongside MCP server.
  • Comprehensive Documentation — Full API reference, architecture docs, integration guides.
  • ATTRIBUTION.md — Clear attribution guidelines for users.

Changed

  • Repositioned from "Agent-Native Vector Database" to "Agent-Native Retrieval Engine"
  • Enhanced MCP server with hybrid retrieval and orchestration tools
  • Improved RAG pipeline with hybrid retrieval support
  • Better CLI with demo, serve, mcp, and bench commands

Architecture

FusionPact 2.0 Architecture:
├── Core: FusionEngine + HNSWIndex
├── Index: TreeIndex (reasoning-based)
├── Retrieval: HybridRetriever (vector + tree + keyword fusion)
├── Memory: AgentMemory (episodic + semantic + procedural + shared)
├── Orchestration: AgentOrchestrator (multi-agent)
├── RAG: RAGPipeline (one-click ingestion)
├── MCP: MCPServer (stdio + HTTP)
└── Embedders: Ollama, OpenAI, Mock

[1.0.0] - 2025-01-15

Initial Release

  • HNSW vector indexing
  • Agent Memory (episodic, semantic, procedural)
  • MCP server
  • RAG pipeline
  • Multi-tenancy
  • Ollama + OpenAI embedders

Built by [FusionPact Technologies Inc.](https://fusionpact.com)

Assets 2
Loading