v0.1.0 — Rooms, Intelligence & Pluggability

🎉 v0.1.0 — Rooms, Intelligence & Pluggability

The biggest release yet — uteke evolves from a simple memory store to a full intelligence layer for AI agents.

🏠 Rooms

Group memories by context (meetings, projects, discussions) with author attribution.

• uteke room create/list/add/recall/remove/delete
• Semantic room recall (--query)
• Structured document generation (uteke room document)
• LLM-free room summaries via tag clustering

⏳ Time-Travel Queries

Recall memories as they existed at any point in time.

• uteke recall "query" --at 2026-06-01T12:00:00Z
• uteke list --at 2026-06-01T12:00:00Z
• Filters by created_at, valid_from/valid_until, deprecated

🔌 Pluggable Embedding Models

New Embedder trait enables different backends.

• ONNX backend (default, EmbeddingGemma Q4 768d)
• Config: [embedding] backend = "onnx"
• Future: OpenAI, Ollama

📊 Benchmarks

• uteke bench — built-in performance benchmarking
• LongMemEval retrieval harness (benchmarks/longmemeval/)

🔗 Relationship Graph

Link memories with typed edges and traverse the graph.

• uteke recall "query" --related --depth 2
• Relationship types: supersedes, contradicts, references, part_of

📉 Smart Decay & Importance

Composite importance scoring with pin/unpin.

• uteke pin <id> / uteke unpin <id>
• uteke importance — recalculate scores

⚡ Recall Cache

LRU cache (256 entries, 5min TTL) eliminates redundant embedding.

🏷️ Tag Junction Table (Schema v5)

Normalized memory_tags table for O(log n) lookups. Dual-write with backward-compatible JSON column.

🔒 Configurable Recall Threshold

• uteke recall --min 0.7 or --strict
• [recall] min_score = 0.3 in config

🔧 JSON Output Fix

recall --json now always outputs [] on empty results (was {"results":[]}). Ensures reliable parsing by cora-cli and other consumers.

---

Full Changelog: https://github.com/codecoradev/uteke/blob/develop/CHANGELOG.md
Docs: https://github.com/codecoradev/uteke/tree/develop/docs
Milestone: v0.1.0 — 0 open / 13 closed ✅