v2.0.0: Personal Intelligence Layer

What's New

HyperCore is no longer just an inference server. It's a personal intelligence system.

Personal Intelligence Layer

• Memory extraction from your documents (journals, notes, ADRs, meeting docs)
• 7 CLI commands: \show, \ imeline,
  ecall, \patterns, \explain, \insight, \sync\
• LLM synthesis with prompt compression and token budgeting
• Feedback collection — rate every insight 1-4, persisted to SQLite

Knowledge Store

• SQLite with hybrid FTS5 + cosine vector similarity
• Content-hash deduplication
• Tree-sitter parsing for code files
• Streaming ingestion with batch embedding

Evaluation

• Real retrieval eval pipeline (no hardcoded scores)
• Benchmark questions with actual theme matching

Infrastructure

• TitanMem KV congestion controller (real, benchmarked)
• 300s timeout for CPU inference
• Memory synthesis uses all CPU threads
• Zero compiler warnings

Sample Documents

• \examples/sample_journal.md\ — engineering journal
• \examples/sample_decisions.md\ — architecture decision records
• \examples/sample_projects.md\ — project notes with relationships

Quick Start

\\�ash
cargo build --release
hypercore ingest --path ./my-notes
hypercore memory show
hypercore memory --model ./model.gguf explain
\\

Full Changelog: v1.0.0...v2.0.0

v1.0.0

fix: rename binary to asset_name before upload