Your code activity, made readable.
Install Nom in one click via GitHub → every push, PR, and release turns into a live feed you can follow, share, and reference.

Upgraded the ImageCarousel component with framer-motion animations, including swipe-to-navigate gesture support and mouse interaction handling. Added adaptive arrow coloring that samples image luminance for better visibility. Success Kid

This release adds first-class PHP parsing and compatibility tools, including semantic mapping and validation gating. It also introduces precise declaration-level term-aware retrieval and ensures consistent performance with new accuracy benchmarks. Additionally, we've implemented notarized, signed macOS package releases for improved security and distribution. PHP developer

This release brings deep PHP support, including tree-sitter parsing for declarations, Composer autoloading, and framework relationships. We have introduced declaration-aware terms-v3 retrieval, complete with exact answer-line evaluation, hard-negative filtering, and robust train/validation/holdout splits for improved context accuracy. Additionally, binaries for all platforms are now checksum-protected, and the CI pipeline rigorously validates asset consistency before publishing. PHP retrieval validation with Drake

Updated the release workflow and test suite to mandate precise validation checks for macOS release assets. The CI pipeline now explicitly verifies that all expected artifacts (including optional signed and notarized packages) are present and checksum-validated before finalizing a release. CI pipeline status

This update makes macOS signing and notarization optional for server releases, controlled by an MACOS_NOTARIZATION_ENABLED repository variable. Raw macOS binaries are now consistently built for all release versions, while .pkg installers are generated only when notarization is enabled. This provides greater flexibility for development and local testing workflows while maintaining secure production standards.

We have officially prepared the release for server v0.10.0. This update involves bumping the version across the plugin manifest, marketplace configuration, and the launcher script. Testing configurations were also updated to ensure proper validation of the new release version. RELEASE v0.10.0

Updated the PHP compatibility evaluation suite to incorporate more robust retrieval validation, including declaration-level checks and new training data. These changes promote retrieval misses into the terms v3 training set and introduce a post-freeze retrieval holdout set to ensure better model generalization. Quality Control

We've added a deterministic, term-aware lexical scorer to improve repo_context focus queries without requiring an expensive model runtime. This update supports natural-language retrieval by splitting identifiers, applying conservative inflections, and rewarding concept coverage, while leaving literal repo_search exact-substring behavior untouched. To ensure long-term retrieval accuracy, we've gated 19 PHP retrieval judgments across multiple language versions as part of our standard blocking CI and test suite. Search Accuracy

We've added robust, native PHP parsing capabilities to improve code analysis and indexing workflows. By integrating a dedicated PHP parser, the system now offers better understanding of PHP semantics, framework structures, and dependency graphs. This significantly enhances the precision of tools interacting with legacy and modern PHP projects. Finally understand my own codebases

We have integrated automated code signing and notarization into our release pipeline for macOS. This ensures that our distributed binaries meet Apple's security requirements, preventing gatekeeper warnings and improving user trust. The new workflow includes a specific script to handle the signing process and updates to our CI configurations. Apple notarization experience

We've integrated a pure-Go tree-sitter runtime to provide consistent, language-aware structural parsing across the codebase. This replaces fragmented language-specific indexers and regex scanners with a unified, robust approach that correctly identifies declaration boundaries in Go, JavaScript/TypeScript, Python, and Rust. As part of this change, we've updated repository chunking and note-anchoring to leverage these precise parsing results, ensuring more reliable context retrieval and memory management while maintaining our CGO-free cross-compilation support.

Updated the project documentation to unify client configuration guidance, ensuring a consistent setup experience across supported MCP clients. This change centralizes configuration details and advice for Claude Code, Claude Desktop, and other tools, reducing ambiguity and improving ease of use.

This update introduces 'memento-workflows', a new lightweight companion plugin designed for environments where running the full Memento MCP server isn't feasible. These skills—covering repository priming, change review, and handoffs—operate entirely using built-in Claude Code tools without requiring native binary downloads or external execution permissions.

Implemented structural tests for the VS Code extension's configuration merging logic and binary selection behavior. Key functions were extracted to new modules to facilitate testing, and new GitHub Actions workflows were added to ensure type-checking and tests run reliably on pull requests and main pushes.

The repo_diff_context tool now supports automatic detection of staged, unstaged, and untracked changes when called without a specific paths argument. It generates a bounded, redacted unified diff summary and intelligently evicts deleted or renamed files from the index to ensure context remains clean and relevant. This update streamlines code reviews and debugging sessions by automatically focusing on the current work-in-progress. Git worktree status

We've implemented the repo_diff_context tool, which provides focused, file-centered git diff chunks without needing graph expansion. This tool enforces strict limits on output, file count, and byte size to maintain deterministic, high-quality context for review workflows, while hardening against stale or invalid index data. It successfully passed all retrieval baseline evaluations, significantly improving efficiency for targeted debugging and code analysis. Context retrieval fixed!

We replaced fixed-line chunking with syntax-aware boundaries for Go and JavaScript/TypeScript. This approach align chunks to top-level declarations and associated documentation instead of arbitrary byte counts, significantly improving context relevance for AI retrieval. Structural chunking transition

We've added a secure self-update flow, allowing standalone installations to easily check for and download binary updates. The new implementation features robust verification using SHA-256 sidecars and handles OS-specific replacement safely, ensuring no corruption during the process. Users can now run memento-mcp update or check availability with --check, while automatic background checks keep you informed of newer versions without disrupting your workflow. Safety first update check

The Docker healthcheck for EchOS previously relied on the web interface's /health endpoint, meaning deployments with ENABLE_WEB=false were incorrectly reported as unhealthy. This release replaces that dependency with a new internal file-based heartbeat system, ensuring accurate status reporting regardless of your deployment configuration.

This release fixes an issue where Docker containers appeared as unhealthy in deployments where the web interface was disabled. We introduced a new, independent heartbeat mechanism in src/heartbeat.ts that updates a local filesystem indicator instead of relying on the web server, ensuring accurate container status reporting regardless of your deployment configuration. Consistent container health reporting