Skip to content

Release v0.10.2

Latest

Choose a tag to compare

@github-actions github-actions released this 28 Jul 15:46
555dbe2

What's New in v0.10.2

Fixed

  • SIGILL on CPUs without AVX2 (#709) β€” Runtime CPU feature detection now dynamically selects between AVX2 and SSE4.2 ONNX Runtime shared libraries. A legacy SSE4.2-only ORT sidecar is included in the Linux legacy bundle for older CPUs (e.g., Intel Celeron J4125/N4020).
  • POST /room/remember returned 500 for invalid memory types (#789) β€” Validation errors now correctly return HTTP 400 instead of 500.
  • Room operations counted deprecated memories (#784) β€” room_stats(), recall_room(), and get_room_memory_ids() now filter deprecated = 0. Previously caused 76% stat inflation (620 vs 148 active).
  • POST /room/recall required query parameter (#785) β€” query is now Option<String> with #[serde(default)]. Empty/missing query falls back to chronological recall instead of returning 400.
  • DELETE /forget rejected short ID prefixes (#794) β€” list and room_recall display 8-char ID prefixes, but forget required full UUIDs. Now resolves short prefixes via SQL LIKE with ambiguous-match detection (400 if >1 match).

Changed

  • ORT loading switched from download-binaries to load-dynamic β€” ORT shared library is now loaded at runtime via dlopen/LoadLibrary instead of being statically linked. Release bundles now include the ORT .so/.dylib/.dll as sidecar files.

Added

  • Auto-generated API reference docs (#786) β€” New crates/docgen binary reads route registry + schemars::JsonSchema derives to generate docs/api-reference.md. CI docs-check job fails if docs are stale. Feature-gated behind docgen flag β€” zero runtime overhead.

Security

  • Bump quinn-proto 0.11.14 β†’ 0.11.15 (GHSA-4w2j-m93h-cj5j) β€” Dependency patch for HTTP/3 stream injection vulnerability.

Docs

  • Hermes integration docs updated β€” Fixed incorrect room_remember example. Added room_document action, "Valid Memory Types" section, and "HTTP API Notes" section (#784, #785, #786).

πŸ“¦ Downloads

Each archive contains three binaries + ONNX Runtime shared library:

Binary Description
uteke CLI tool
uteke-serve HTTP server daemon
uteke-mcp MCP server (stdio + HTTP)
libonnxruntime.so* / libonnxruntime*.dylib / onnxruntime*.dll ONNX Runtime
Platform File
Linux x86_64 (AVX2) uteke-x86_64-unknown-linux-gnu-v0.10.2.tar.gz
Linux x86_64 (Legacy, SSE4.2) uteke-x86_64-unknown-linux-gnu-legacy-v0.10.2.tar.gz
Linux ARM64 uteke-aarch64-unknown-linux-gnu-v0.10.2.tar.gz
macOS Apple Silicon uteke-aarch64-apple-darwin-v0.10.2.tar.gz
Windows x86_64 uteke-x86_64-pc-windows-msvc-v0.10.2.zip

Legacy Bundle (Linux only) β€” Includes a SSE4.2-only ORT sidecar (ort-legacy/) for CPUs without AVX2 (Intel Celeron J4125/N4020). Use this bundle to avoid SIGILL crashes on older hardware.

πŸš€ Quick Start

# Quick install (Linux / macOS)
curl -fsSL https://raw.githubusercontent.com/codecoradev/uteke/main/install.sh | sh

# Pin a specific version
UTEKE_VERSION=0.10.2 curl -fsSL https://raw.githubusercontent.com/codecoradev/uteke/main/install.sh | sh

# Store a memory
uteke remember "Important context" --tags project

# Recall by meaning
uteke recall "what was that context?"

# Start server for fast AI agent access
uteke-serve --port 8767

Full changelog: https://github.com/codecoradev/uteke/blob/main/CHANGELOG.md