Skip to content
CAI
Software that uses CAICheck a score

thedotmack/claude-mem

29.8

Weak · 13 November 2026

278.6k

lines of production code

TypeScript

with Python, JavaScript

6

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Features

Add Claude-Mem plugin with observation feed and memory sync

The OpenClaw plugin now registers a persistent memory service that syncs to MEMORY.md, injects context via the system prompt, and supports an optional SSE-based observation feed to notify channels like Telegram or Discord. It registers commands (e.g., /claude\_mem\_feed) to show feed status and toggle it, and hooks into session lifecycle events (session\_start, after\_compaction, before\_agent\_start, before\_prompt\_build, tool\_result\_persist, agent\_end, gateway\_start) to maintain memory and emit observations. Configuration includes enabling the feed, specifying the target channel and recipient, bot token, and emoji styling for feed messages.

openclaw/src · high confidence

Add OMP (Oh My Pi) hook adapter for shared memory

Introduces a new OMP hook adapter that records OMP coding sessions into the existing claude-mem store, allowing OMP sessions to be viewed alongside Claude Code and Cursor observations under the same project. The adapter translates OMP lifecycle events (session start, prompts, tool results, context injection) into claude-mem's REST API, handling session rotation on branch/switch, caching context responses, and ensuring fire-and-forget, non-blocking communication with a circuit breaker for reliability. Also in this area: 3 commits (3 fixes) fixing omp.

omp/hooks · medium confidence · unverified

Add Observation TV viewer and status icons

The plugin UI now includes a web-based viewer for real-time memory streams, accessible via the new \tv.html\ page which displays observation cards with project and source metadata, supports Picture-in-Picture and full-screen modes, and uses a token for secure remote access. Additionally, four new SVG icons (\icon-thick-completed\, \icon-thick-investigated\, \icon-thick-learned\, \icon-thick-next-steps\) have been added to the \plugin/ui\ directory to support status indicators within the interface.

plugin/ui · high confidence

Add script to generate OpenRouter list-price history for expense reports

A new build script (\scripts/openrouter-price-history/build-dataset.py\) and supporting source data (price ledgers, popularity eras, aliases, and work days) have been added to generate daily OpenRouter list-price datasets. This tool expands vendored price change-points into a daily join table covering July 1 to September 8, 2026, outputting JSON and CSV files to \docs/expense-pricing/\ to support expense reporting and cost analysis.

scripts/openrouter-price-history · high confidence

Automated bug report generation using Claude Agent SDK

A new \scripts/bug-report\ tool has been added to automatically generate structured GitHub issue reports. The CLI collects system diagnostics—including versions, platform details, worker health, database table counts, and logs—and uses the Claude Agent SDK to format these details into a professional Markdown issue, including automatic translation of user descriptions to English. A template-based fallback is provided if the SDK call fails. Also in this area: Fixes to bug-report collection, log following, and export reliability; Fixes for log file naming and hook versioning.

scripts/bug-report · high confidence

Centralized credential management and settings defaults

The \src/shared\ directory now houses the core configuration and credential infrastructure. \EnvManager.ts\ acts as the single source of truth for non-OAuth credentials, reading exclusively from \\~/.claude-mem/.env\ and enforcing a strict whitelist of allowed keys to prevent shell environment leaks. \SettingsDefaultsManager.ts\ centralizes default values, handles legacy migrations (such as OpenRouter model retirements and sync hub URL updates), and normalizes settings types. Supporting modules include \api-key-pool.ts\ for multi-key rotation, \atomic-json.ts\ for crash-safe file writes, and \claude-fallback-readiness.ts\ for managing Anthropic plan fallbacks.

src/shared · high confidence

Introduce Bun runtime support with cross-platform launcher and settings defaults

The plugin now supports running scripts via the Bun runtime. A new \bun-runner.js\ script acts as a cross-platform launcher that detects the Bun executable (handling Windows \bun.exe\/\bun.cmd\ resolution and avoiding \cmd.exe\ shell limits) and forwards stdin to the target script. Additionally, \context-generator.cjs\ has been updated to include a comprehensive \SettingsDefaultsManager\ that defines default values for all plugin configuration keys (such as model selection, Chroma settings, and telemetry options) and handles loading them from the user's settings file.

plugin/scripts · high confidence

Introduce Claude-Mem OpenClaw plugin with installer and E2E testing

The openclaw directory now contains the complete distribution package for the Claude-Mem plugin, enabling persistent memory and real-time observation streaming for OpenClaw agents. This includes the plugin manifest (openclaw.plugin.json) defining configuration options such as worker host/port, project scoping, and observation feed channels (Telegram, Discord, Slack, etc.), alongside a comprehensive one-line installer (install.sh) that handles dependency checks, plugin registration, and interactive setup. To ensure reliability, the release ships with a Docker-based end-to-end test suite (Dockerfile.e2e, test-e2e.sh, e2e-verify.sh) that validates plugin loading, SSE stream connectivity, and gateway integration in a clean environment.

openclaw · high confidence

Introduce Owner Goals skill to surface owner priorities in My Work

Adds the \owner-goals\ skill, which extracts prompts typed personally by the owner (excluding agent-generated text and private content), maps their top priorities and open asks with dated quotes, and saves them to the \owner-goals\ work-state list. These goals appear in a new Goals panel at the top of the My Work interface, allowing users to see their key objectives and open questions with supporting evidence from their own words. The skill supports incremental updates and historical month-by-month goal tracking. Also in this area: 10 commits (2 fixes) modifying plugin.

plugin/skills/owner-goals/scripts · medium confidence · unverified

Introduce canonical content model and two-lane sync architecture

This change introduces a new canonical content model (CanonicalContent.ts) that defines a strict, deterministic wire format for sync operations, including schema versions, stable document IDs, and mutation types (set\_title, set\_prompt\_session, remap\_project). It implements the client-side apply path (SyncApply.ts) for crash-safe, exactly-once application of remote operations to local SQLite, and the pull loop (SyncClient.ts) with Supabase Realtime support. The existing Chroma integration is refactored into a new manager (ChromaMcpManager.ts) with a circuit breaker for prewarm failures and a state tracker (ChromaSyncState.ts) for watermarks, while the push path (CloudSync.ts) is updated to use the new canonical format and mutation outbox.

src/services/sync · high confidence

Introduce dedicated search strategy implementations for Chroma, Hybrid, and SQLite backends

The search worker now uses three distinct strategy classes—ChromaSearchStrategy, HybridSearchStrategy, and SQLiteSearchStrategy—to handle vector, hybrid, and full-text search respectively. ChromaSearchStrategy queries the vector store with project and platform-source filters, then hydrates results from SQLite. HybridSearchStrategy combines SQLite file-path lookups with Chroma ranking for file-specific searches. SQLiteSearchStrategy performs direct full-text queries against the local database. All strategies enforce consistent project scoping, date-range filtering, and result ordering.

src/services/worker/search/strategies · high confidence

Introduce dedicated server-side generation providers for Claude, Gemini, and OpenRouter

The server now includes dedicated provider implementations for Anthropic Claude, Google Gemini, and OpenRouter, replacing previous ad-hoc or shared logic with specialized classes that handle API-specific request formatting, error classification, and model defaults. The Claude provider uses the \claude-sonnet-5\ model by default and supports custom base URLs for Anthropic-compatible gateways. The Gemini provider targets the \v1beta\ API endpoint with the \gemini-flash-latest\ model alias to avoid retired model 404s, and correctly distinguishes between quota exhaustion, rate limits, and region refusals. The OpenRouter provider supports configurable base URLs for OpenAI-compatible endpoints (e.g., DeepSeek, LM Studio), merges extra request body parameters via \CLAUDE\_MEM\_OPENROUTER\_EXTRA\_BODY\, and ensures non-streaming responses by default. All providers share a common error-classification framework and prompt-building utility, ensuring consistent handling of authentication errors, parse failures, and network issues across the generation pipeline.

src/server/generation/providers · high confidence

Introduce npx claude-mem CLI with install, doctor, and server management commands

Users can now manage the Claude Memory service via the \npx claude-mem\ command-line interface. The new \install\ command handles the full setup flow, including runtime provisioning, IDE detection, and provider configuration. A \doctor\ command provides a read-only diagnostic sweep of the environment, checking worker health, Chroma child exits, and tool availability. The \server\ command exposes lifecycle management (start, stop, restart, status) and API key rotation, while \runtime\ allows direct control of the worker and server processes. Additional commands include \prune\ for cleaning up stale plugin cache versions and \teambrain\ for joining and managing team sync memberships.

src/npx-cli/commands · high confidence

Introduce shared server-side generation infrastructure for provider adapters

This change adds the foundational shared components for the server's observation generation pipeline in \src/server/generation/providers/shared\. It introduces \error-classification.ts\ to standardize how HTTP errors from providers (like Gemini and OpenRouter) are categorized into specific types (e.g., \rate\_limit\, \quota\_exhausted\, \transient\) and outcomes (e.g., \refused\_before\_work\), ensuring consistent retry and billing logic. It also adds \prompt-builder.ts\ to construct single-shot generation prompts from agent events, including privacy stripping and payload truncation, and \types.ts\ to define the \ServerGenerationContext\ and \ServerGenerationProvider\ interfaces that decouple the server's generation logic from the worker's internal state.

src/server/generation/providers/shared · high confidence

Introduce smart-file-read service with secure, cross-platform tree-sitter parsing

The new smart-file-read service enables codebase-wide symbol search, outlining, and unfolding across 24 languages (including JavaScript, TypeScript, Python, Go, Rust, Ruby, Java, C/C++, Swift, PHP, Kotlin, Zig, and Markdown) by integrating the tree-sitter CLI. It resolves the tree-sitter executable for the current platform (handling Windows .exe suffixes) and provisions the binary securely when missing from marketplace or script-suppressed installs, verifying the downloaded executable against pinned SHA-256 checksums. The service includes a language map for file extension detection, a parser that generates folded file views and symbol outlines, and a search engine that ranks matches by relevance. Crucially, it enforces workspace containment, preventing MCP calls from reading files outside the project root (e.g., \~/.ssh/id\_rsa). Also in this area: 6 commits (6 fixes) fixing src/services/smart-file-read; Fixes to symbol outline accuracy and Tree-sitter CLI probe robustness; Improved code outline accuracy and docstring display for TypeScript, Python, Ruby, and Lua.

src/services/smart-file-read · medium confidence · unverified

Introduce web-based memory stream viewer UI

Adds a new React-based web interface for viewing real-time memory streams, featuring a project timeline feed, session list, and session detail views. The viewer supports live updates via Server-Sent Events (SSE), paginated loading of observations, summaries, and prompts, and includes a context settings modal with terminal preview. It also provides session management capabilities such as filtering by project, deleting sessions, and displaying session counts, along with theme toggling and a welcome card.

src/ui/viewer · high confidence

Introduce web-based viewer UI for real-time memory stream

The viewer now features a complete web-based UI for browsing and managing memory data. Users can filter observations by category, delete individual observations or entire sessions, and view detailed session summaries with sections for investigated, learned, completed, and next steps. The interface includes a context settings modal with a terminal preview, a logs drawer for debugging, and a header with project filters, theme toggling, and a real-time queue depth indicator. A 7-day Pro trial promotion is displayed prominently in the header, and the UI supports infinite scrolling with pagination for both session lists and observation feeds.

src/ui/viewer/components · high confidence

Introduces BullMQ observation queue configuration and health reporting

Adds new TypeScript definitions and configuration logic to support the BullMQ-based observation queue in the server-beta runtime. The \redis-config.ts\ module now handles environment-specific Redis connection settings (including IPv6 host normalization, URL parsing, and mode detection) and exposes the selected engine (BullMQ vs. SQLite). The \queue-health-types.ts\ file defines the schema for the \/api/health\ endpoint, enabling deploy probes to monitor queue saturation and engine status via per-lane snapshots.

src/server/queue · high confidence

Introduces durable tool\_uses backup index and get\_tool\_uses disclosure

The SQLite plugin now includes a dedicated \tool\_uses\ table with a robust schema and multiple indexes (including \idx\_tool\_uses\_project\_nocase\_created\) to support durable backup and efficient querying of tool usage data. This change enables the \get\_tool\_uses\ disclosure capability, allowing users to retrieve detailed records of tool interactions stored in the local SQLite database. Also in this area: 10 commits (2 fixes) modifying plugin.

plugin/sqlite · medium confidence · unverified

Introduces embedded Process Supervisor for unified lifecycle management

Adds a new embedded Process Supervisor in src/supervisor that centralizes worker and child-process lifecycle management. It includes an environment sanitizer to prevent sensitive variables (like CLAUDE\_CODE\_EFFORT\_LEVEL) from leaking into SDK subprocesses, a process registry with identity validation to handle PID reuse, a health checker to prune dead entries, and a boot-time sweep to reap orphaned chroma-mcp process trees. The supervisor also manages signal handling and a structured shutdown cascade to ensure clean process termination.

src/supervisor · high confidence

Introduction of a configurable Mode system with inheritance support

The domain service now includes a new ModeManager that loads and manages operational modes defined by JSON configuration files. Users can define custom modes with specific observation types, concepts, and prompt templates. The system supports a single level of inheritance (e.g., 'parent--override'), allowing new modes to extend existing ones by merging configurations. Modes are loaded from user data directories, plugin directories, or development paths, with a fallback to a default 'code' mode if a specified mode is not found.

src/services/domain · high confidence

Near-duplicate detection with Tier-0 auto-merge and Tier-1 candidate review

The deduplication service now implements a two-tier system for identifying near-duplicate observations. Tier-0 performs exact matching on normalized titles (case, whitespace, and punctuation insensitive) to enable safe, silent auto-merges. Tier-1 identifies candidate duplicates using IDF-weighted TF-IDF cosine similarity, but introduces an 'IDF veto' that prevents merging if the titles differ by a rare, discriminating token (e.g., 'rdlp-api' vs 'rdlp-plugin'), ensuring such pairs are flagged for review rather than auto-merged. This logic is supported by new utilities for title normalization, tokenization, and IDF calculation.

src/services/dedup · high confidence

New Agent Cost Report skill for weekly agent spend analysis

A new \agent-cost-report\ skill has been added, providing a self-contained Python pipeline to generate manager-readable cost and failure reports for agent activity. The tool collects measured token usage from local Claude Code transcripts, prices them against OpenRouter public list prices, and calculates estimated costs, waste, and behavior metrics. It outputs a Timing-style HTML report, a PDF (via headless Chrome), a JSON summary, a CSV of line items, and supporting evidence files. The skill includes a read-only source for GitHub merged-PR wins, a heuristic behavior pass to detect frustration and waste patterns, and an optional low-cost LLM classifier for unsettled candidates. It is designed to run with Python 3.9+ standard library only, with specific handling for Windows timezone data and portable date/UTF-8 exports. Also in this area: 1 commit (1 fix) fixing (repo-wide); 6 commits (2 fixes) modifying claude-mem-cursor, claude-mem-grok-bot, openclaw.

(repo-wide) · medium confidence · unverified

New CLI commands for CLAUDE.md management and hook execution

The CLI now includes dedicated commands for managing CLAUDE.md files and executing hooks. The \claude-md-commands.ts\ module introduces \generate\ and \clean\ commands to create and maintain CLAUDE.md files, including logic to track folders, handle file observations, and ensure BMP-safe content to prevent session issues with astral characters. The \hook-command.ts\ module provides a robust entry point for running hooks, featuring improved error handling (fail-open on malformed input or worker unavailability), support for environment-based hook opt-outs, and proper stdin/stdout discipline to prevent blocking or leaking output. Additional utilities in \hook-original-cwd.ts\ and \spool-hook-event.ts\ ensure hooks run in the correct directory and reliably spool events to remote workers.

src/cli · high confidence

New HTTP routes for advisor calls, cloud sync status, and corpus management

The worker now exposes dedicated HTTP endpoints for several capabilities: advisor tool calls can be ingested and queried via /api/advisor-calls; cloud sync configuration and status are available at /api/sync/status; and full corpus lifecycle management (build, rebuild, prime, query, delete) is handled through /api/corpus, including long-running operations that stream progress via Server-Sent Events. Additional new routes include /api/chroma/status for vector store health, /api/dedup/candidates and /api/dedup/scan for near-duplicate observation management, /api/spool/ingest for authenticated remote hook spooling, /api/logs for log tailing and clearing, /api/memory/ingest for importing native memory files, /api/memory/save for manual memory entries, and /api/dedup/candidates for dedup review.

src/services/worker/http/routes · high confidence

New IDE integrations and memory alignment features

This change introduces dedicated installers and adapters for the Antigravity CLI, Codex CLI, Cursor, and DeepSeek Harness, enabling users to install and manage these integrations directly. It also adds the CCS Align subsystem, which provides a seat-owned middle cache for observations and a rules walker to detect and report conflicts across house, project, and seat layers. Additionally, it implements Grok Bot awareness features, including a pusher for writing facts to agent logs, a brainbeat webhook for real-time observation notifications, and an index writer for maintaining a live memory index. Also in this area: 4 commits (3 fixes) fixing src/services/integrations.

src/services/integrations · medium confidence · unverified

New README translation script with multilingual support and caching

A new CLI tool and library (\scripts/translate-readme\) has been added to automatically translate README.md files into multiple languages using the Claude Agent SDK. Users can now generate localized documentation via command-line arguments or programmatically, supporting over 30 languages including Chinese, Urdu, and others. The tool features parallel translation execution, a local cache to avoid re-translating unchanged content, budget controls, and the ability to reference existing translations for improved consistency. It supports various authentication methods, including local Claude Code sessions, API keys, AWS Bedrock, and Google Vertex AI.

scripts/translate-readme · high confidence

New agent processing and SSE broadcasting infrastructure

The worker agent layer now includes dedicated modules for handling agent responses and broadcasting events via Server-Sent Events (SSE). A new ResponseProcessor extracts file evidence (reads and modifications) from tool usage observations, while an ObservationBroadcaster sends new observations and summaries to connected clients, respecting project tracking settings. Supporting utilities include an AbortError checker and type definitions for worker references and SSE payloads. Also in this area: 7 commits (4 fixes) fixing src/services/worker; 2 commits (1 fix) modifying src/services/worker/agents; Improved detection and logging of truncated agent replies; Claude quota fallback and search scope fixes; Fixes to response processing logic and metadata handling.

src/services/worker/agents · medium confidence · unverified

New anti-pattern detection script for error handling

A new script, \scripts/anti-pattern-test/detect-error-handling-antipatterns.ts\, has been added to automatically scan the codebase for fragile error handling practices. This tool identifies specific anti-patterns such as detecting error types via string matching (e.g., \error.message.includes(...)\), logging only partial error messages instead of the full error object, guessing error types through multiple string checks, and using empty promise catch handlers. The script supports an override mechanism using \\[ANTI-PATTERN IGNORED\]\ comments for approved exceptions, particularly in critical paths like provider implementations and session stores.

scripts/anti-pattern-test · high confidence

New build, validation, and operational scripts for hooks, workers, and UI

The scripts directory now includes a comprehensive set of new tooling: build-hooks.js generates hook and MCP server bundles with strict integrity checks to prevent build-path leaks; build-viewer.js compiles the React-based memory viewer UI; and build-worker-binary.js produces a standalone Windows executable for the worker service. Operational discipline is enforced by check-hook-io-discipline.cjs (ensuring handlers don't write directly to streams) and check-spawn-env-discipline.cjs (preventing credential leakage in subprocesses). User-facing management is improved with check-pending-queue.ts and clear-pending-queue.ts for monitoring and clearing the worker's message queue, while export-memories.ts allows users to export their memory data. Additional scripts include claude-mem-sync for cross-machine database synchronization, check-postinstall-allowlist.js to prevent CI hangs from unexpected dependency scripts, discord-release-notify.js for automated release announcements, and e2e-server-docker.sh for end-to-end testing of the new Docker-based server architecture. Also in this area: 1 commit (0 fixes) modifying scripts; 5 commits (2 fixes) modifying scripts; Test suite split into fast and integration lanes; subagent stop hook added; Fix macOS Intel search failure and ensure BMP-only output in regenerated markdown; My Work app build process and request-progress CLI integration; and 4 more.

scripts · medium confidence · unverified

New changelog generation script for release notes

A new Node.js script, generate\_changelog.js, has been added to the version-bump plugin to automate the creation of project changelogs. This tool reads release data from standard input (expecting a JSON array of release objects) and outputs a formatted Markdown changelog file, including the release tag, publication date, and body text for up to the 50 most recent releases.

plugin/skills/version-bump/scripts · high confidence

New claude-mem Docker harness for local end-to-end testing

Introduces a new Docker-based environment for running claude-mem locally, including a Dockerfile that bundles the Claude Code CLI, Bun, and a baked-in embedding model, along with shell scripts (build.sh, run.sh, entrypoint.sh) to manage authentication and container modes (server, worker, shell). This location also adds an E2E test driver (docker/e2e/server-e2e.mjs) that validates the server-beta runtime's health, API endpoints, and authentication flows against a Postgres and Valkey backend.

docker/claude-mem · high confidence

New core data schemas and worker lifecycle management services

This change introduces a new set of Zod-based data schemas in src/core/schemas/ (covering agent events, authentication, memory items, projects, sessions, and teams) to standardize data validation and typing. It also adds new service modules in src/services to manage the worker lifecycle: worker-spawner.ts handles daemon spawning and health checks, worker-shutdown.ts manages graceful shutdowns and restart handoffs, restart-verify.ts validates successful worker restarts, and context-generator.ts provides context injection utilities. These additions support a more robust and verifiable worker process management system.

src/services · high confidence

New core data schemas for agents, auth, memory, projects, sessions, and teams

The core module now includes a comprehensive set of Zod validation schemas that define the structure and constraints for key domain entities. These new files establish the data contracts for agent events (including platform source tracking), API keys and audit logs, memory items (with specific handling for PATCH operations to preserve omitted fields), projects, server sessions, and teams (including member roles). This provides the foundational type safety and validation logic for these resources within the application.

src/core · high confidence

New database schema types and shell-quote type definitions

Added new TypeScript type definitions for the database layer in src/types/database.ts, introducing interfaces for table column info, indexes, schema versions, and specific record types for observations, session summaries, and user prompts (including platform source and memory session ID fields). Additionally, added type declarations for the shell-quote module in src/types/shell-quote.d.ts to provide proper typing for command parsing.

src/types · high confidence

The CLI now includes dedicated handlers for file context injection, file edits, and memory search. The file context handler enforces a read gate that blocks full-file reads for files 32 KB or larger when prior observations exist, and it deduplicates file-context injections per session using a SQLite gate. The file-edit handler records write\_file observations, routing them to the server in server runtime or spooling them in worker runtime. The memory search handler bridges native shell searches (grep, rg) and memory tools to the Smart search index, supplementing results with observation data. These handlers are registered in the CLI's event dispatch system.

src/cli/handlers · high confidence

New platform adapters for Antigravity, Claude Code, Codex, Cursor, Kimi, and Windsurf

The CLI now includes dedicated adapters for several AI coding platforms, enabling native hook integration. The new \src/cli/adapters\ module provides \antigravity-cli\, \claude-code\, \codex\, \cursor\, \kimi\, and \windsurf\ adapters, along with a \raw\ adapter for generic input. These adapters normalize platform-specific hook payloads (stdin) into a common internal format and format output responses back to the platform's expected protocol (stdout). The \index.ts\ file registers these adapters, allowing the CLI to dispatch to the correct handler based on the platform identifier. Specific implementations handle platform nuances, such as Antigravity's protojson format, Claude Code's \submitted\_prompt\ handling, Codex's file path extraction from Bash commands, Cursor's transcript path derivation, Kimi's attachment scheme handling, and Windsurf's action-based routing. Also in this area: 6 commits (3 fixes) modifying src/cli.

src/cli/adapters · medium confidence · unverified

New remote recall MCP server for hosted access

A new \createRecallMcpServer\ factory has been added to expose read-only recall tools (\search\, \context\, \recent\) via the Model Context Protocol. This allows hosted environments to provide the same memory retrieval capabilities as the local CLI, ensuring that pasted recall links and hosted interactions access identical data. The implementation is designed to be injected with a scoped backend, keeping the server layer pure and testable while handling tool dispatch and error reporting.

src/server/mcp · high confidence

New server-beta middleware suite for authentication, rate limiting, and usage tracking

The server-beta runtime introduces a new set of middleware components in src/server/middleware to handle core operational concerns. Authentication is now managed by dedicated middlewares (auth.ts and postgres-auth.ts) that support both SQLite and Postgres backends, accepting Bearer tokens with X-Api-Key as a fallback and allowing local-dev bypass under specific conditions. Rate limiting and monthly usage quotas are enforced via new middleware (rate-limit.ts) that tracks per-key request counts and per-team monthly caps, failing open to avoid service disruption. Additionally, request metering (usage-metering.ts) records request counts for usage tracking, and request ID generation (request-id.ts) ensures traceability across logs and services.

src/server/middleware · high confidence

New standalone transcript watcher CLI and configuration system

A new standalone transcript watcher is now available via the \claude-mem transcript\ CLI (init, watch, validate), allowing users to configure and run file watchers independently of the main worker. The system introduces a structured configuration format (\TranscriptWatchConfig\) with schema-based event matching, field resolution, and durable state management (atomic writes, device/inode identity checks, and zstd frame boundary handling). It specifically handles Codex subagent rollouts by scoping or suppressing watches based on native hook capture, and supports DeepSeek Harness zstd session logs.

src/services/transcripts · high confidence

New standup skill for multi-branch reconciliation

A new 'standup' skill has been added to facilitate read-only group chats between AI agents representing different git worktrees or pull requests. This feature allows multiple branches to discuss their changes, resolve conflicts, and agree on a single consolidation plan without performing any merges. The implementation includes a Node.js CLI tool (\standup.mjs\) that manages a shared markdown file as a chat room, supporting commands to list active worktrees/PRs, post messages, track consensus, and generate a final summation for the user to execute.

plugin/skills/standup · high confidence

New web-based viewer UI and Observation TV broadcast interface

This change introduces two new web-based user interfaces in the \src/ui\ directory. The \viewer-template.html\ file provides a full-featured viewer with light, dark, and system-preference themes, displaying observations, prompts, and summaries with distinct color-coded badges. Additionally, \tv.html\ implements a read-only 'Observation TV' broadcast interface designed for large displays, featuring a minimalist card-based layout, live stream status indicators, and support for Picture-in-Picture and full-screen modes. Also in this area: 2 commits (0 fixes) modifying src/ui; Refined UI design and accessibility for the Memory viewer.

src/ui · medium confidence · unverified

OpenCode plugin now supports both V1 and V2 plugin contracts

The OpenCode integration has been rewritten to support both the legacy V1 plugin contract (OpenCode 1.3.4+) and the newer V2 contract (OpenCode 2.x). The plugin entry module now exports a unified definition that detects the host version and registers hooks via the appropriate API—returning a hooks object for V1 or imperatively registering them via \ctx.\*.hook()\ for V2. This ensures the plugin loads correctly regardless of the OpenCode version, while sharing a common core for session mapping, memory context injection, and worker communication. Also in this area: 1 commit (0 fixes) modifying (repo-wide); 3 commits (2 fixes) fixing src/integrations; Memory search defaults to a three-step Smart workflow and adds explicit timeline/observation tools.

src/integrations/opencode-plugin · low confidence · unverified

Opt-in ACT-R reinforcement ranking for context injection

Observations are now ranked by a strength-weighted score based on ACT-R base-level activation, allowing frequently re-confirmed notes to remain visible even if they are older. This feature is opt-in via the \CLAUDE\_MEM\_REINFORCE\_ALPHA\ setting (defaulting to 0, which preserves the legacy recency-only behavior). The implementation tracks reinforcement history in the \observations\ table and uses a power-law decay formula to calculate scores, ensuring that durable knowledge is not buried by pure recency.

src/services/reinforcement · high confidence

Postgres image now includes pgvector and pg\_textsearch extensions

The local Docker image for Postgres (built from docker/postgres) now installs the pgvector and pg\_textsearch extensions, enabling semantic search by meaning and keyword ranking via bm25. To handle potential C library mismatches (e.g., when switching between Alpine and Debian-based images), a new reindex-collation script is included to safely rebuild only the affected text indexes without touching vector or JSONB indexes.

docker/postgres · high confidence

Postgres storage backend for the server runtime

The server now uses Postgres as its primary storage layer, replacing the previous implementation. This change introduces repositories for managing agent events, observations, generation jobs, sessions, and projects, along with support for vector-based semantic search via pgvector and keyword search ranking via pg\_textsearch (BM25). It also adds data deletion capabilities for compliance, usage metering, per-key rate limiting, and audit logging, while ensuring sensitive content is redacted and memory tags are stripped before storage.

src/storage/postgres · high confidence

Repository infrastructure and documentation scaffolding

The repository now includes essential configuration and documentation files to support development, packaging, and security practices. A \.dockerignore\ and \.npmignore\ are added to exclude source, test, and build artifacts from Docker images and npm packages, ensuring cleaner distributions. A \.gitattributes\ file enforces consistent line endings (LF) for text files and marks binary assets. A \.markdownlint.json\ configures linting rules, and a \.npmrc\ sets \legacy-peer-deps=true\ to resolve dependency conflicts. A \.translation-cache.json\ tracks the status and cost of localized README translations. Additionally, a comprehensive \CHANGELOG.md\ documents the project history, a \CLAUDE.md\ provides build and usage instructions for AI assistants, and \SECURITY.md\ outlines the security policy and vulnerability reporting process. Docker configuration files (\Dockerfile.test-installer\, \docker-compose.yml\, \docker-compose.e2e.yml\) are added to support testing and server deployment.

(repo-wide) · high confidence

SQLite observation file tracking and logging infrastructure

The SQLite observations plugin now includes a new module that manages file path tracking for observations, allowing the system to record and replace the list of files associated with a specific observation in the database. This change also bundles a comprehensive logging utility that writes timestamped, structured logs to daily files in the user's data directory, supporting configurable log levels and detailed formatting for debugging and operational visibility.

plugin/sqlite/observations · high confidence

Server beta introduces Postgres-backed event ingestion, MCP recall, and usage controls

The server runtime now supports a Postgres storage backend alongside the existing SQLite mode, enabling new capabilities in the v1 API. Legacy clients can continue using the translated \/api/sessions/observations\ and \/api/sessions/summarize\ endpoints, which now map to the new event model. A new remote MCP recall endpoint (\POST /v1/mcp\) exposes Smart search tools (search, timeline, get\_observations, context, recent) for external integrations. Additionally, the server now supports optional usage metering, per-key rate limiting, and monthly request/token caps via environment variables, providing granular control over API consumption. Also in this area: 1 commit (1 fix) fixing src/server, src/ui; 7 commits (7 fixes) fixing src/server.

src/server/routes · low confidence · unverified

Server runtime introduces BullMQ-based generation job queues with concurrency controls and health metrics

The server runtime now supports a BullMQ-backed job queue system for generation tasks (event and summary lanes), replacing the previous in-memory or disabled state. This change introduces explicit concurrency tuning via the CLAUDE\_MEM\_SERVER\_GENERATION\_CONCURRENCY environment variable (capped at 64 per lane), deterministic job ID generation to prevent Redis collisions, and robust stale-job reclamation to recover work left by dead workers. Health endpoints (/api/health and /v1/info) now expose per-lane metrics (waiting, active, completed, failed, delayed, stalled) and Redis connectivity status, allowing operators to monitor queue saturation and detect stalled jobs. The system also includes audit logging for stalled jobs and validates job payloads at the queue boundary to prevent malformed jobs from entering the transport. Also in this area: 1 commit (1 fix) fixing src/server, src/ui; 7 commits (7 fixes) fixing src/server.

src/server/runtime · low confidence · unverified

Viewer UI enhancements: theme toggle, spinning favicon, and real-time streaming

The viewer now supports a theme toggle (light, dark, and system preferences) persisted in localStorage, and displays a spinning favicon while the system is processing. Real-time updates are handled via a new Server-Sent Events (SSE) hook that streams observations, summaries, and prompts, and includes a queue depth indicator. Context previewing has been improved with a hook that isolates Claude and Codex session sources, and pagination logic has been refactored to handle filter changes and loading states more robustly. Additionally, settings are now saved with improved error handling, ensuring sensitive fields like CLAUDE\_CODE\_PATH are not sent to the backend. Also in this area: 1 commit (1 fix) fixing src/server, src/ui; Refined UI design and accessibility for the Memory viewer; Improved reliability and error handling in settings and session management; Viewer stability, logging, and data handling improvements.

src/ui/viewer/hooks · low confidence · unverified

Removals

Removal of JavaScript-based hook scripts and shared utilities

The JavaScript hook implementations located in the hooks directory have been removed. This includes the pre-compact, session-end, and session-start hooks, along with their shared configuration loader and helper utilities. Users relying on these specific JavaScript-based hooks for session context loading, pre-compaction validation, or session-end memory archiving will no longer have this functionality available in this location.

hooks · high confidence

Security

Hardened HTTP security and error handling in the worker service

The worker's HTTP layer now enforces strict security policies to prevent cross-origin attacks and DNS rebinding. It refuses framing for all pages except the Observation TV endpoint, restricts CORS to localhost origins or explicitly allowed origins, and blocks DELETE requests from any browser page other than the worker's own viewer. A Host header check prevents DNS rebinding by rejecting requests from untrusted hostnames. Additionally, a new BaseRouteHandler standardizes error handling, ensuring client errors (4xx) are logged as warnings while server errors (5xx) are routed to the error sink, and safely parses request parameters to prevent stack overflows from deeply nested input.

src/services/worker/http · high confidence

Observer security lockdown and output classification improvements

The SDK now enforces a defense-in-depth security model for Observer and KnowledgeAgent sessions to prevent unauthorized tool access, using a new hardened-options module that disables all built-in tools, restricts filesystem and MCP access, and audits every tool attempt. Additionally, a new output classifier distinguishes between valid XML, idle, and prose responses to prevent silent data loss, while the parser has been updated to preserve unsupported observation types, handle nested XML tags, and normalize concept tags to improve search accuracy. Also in this area: 2 commits (1 fix) modifying src/sdk; 1 commit (1 fix) fixing src/sdk; Improved detection and logging of truncated agent replies; Claude quota fallback and search scope fixes; Fix parser handling of whitespace-only array entries and notes-only summaries.

src/sdk · medium confidence · unverified

Architecture

Worker service restructured into modular provider and service components

The worker service has been refactored from a monolithic structure into a modular architecture with dedicated files for each provider and core service. This introduces new provider implementations for Claude, Codex, Gemini, and OpenAI-compatible endpoints, each with specialized error classification and request handling. The refactoring also includes new services for database management, context caching, and formatting, along with improved process management for the Codex app-server pool. These changes improve code maintainability and allow for better isolation of provider-specific logic while maintaining the same external behavior for users.

src/services/worker · high confidence

Behavioural changes

1 commit (0 fixes) modifying examples/elizaos

A change to existing behaviour in examples/elizaos — 1 commit, 1 file: “feat(elizaos): add verified explicit memory example (\#6035)” (refs \#6035).

examples/elizaos · medium confidence · unverified

1 commit (0 fixes) modifying examples/pipecat

A change to existing behaviour in examples/pipecat — 1 commit, 2 files: “feat(pipecat): add explicit project-memory pipeline example (\#6050)” (refs \#6050).

examples/pipecat · medium confidence · unverified

1 commit (0 fixes) modifying examples/vercel-ai-sdk

A change to existing behaviour in examples/vercel-ai-sdk — 1 commit, 2 files: “docs(integrations): add a verified Vercel AI SDK memory workflow (\#5951)” (refs \#5951).

(repo-wide) · medium confidence · unverified

1 commit (0 fixes) modifying integrations/eve/provider, integrations/eve/template/apps/web

A change to existing behaviour in integrations/eve/provider, integrations/eve/template/apps/web — 1 commit, 89 files: “feat(eve): claude-mem memory for eve agents + one-click personal agent template (\#5863)” (refs \#5863).

integrations/eve/template/apps/web · medium confidence · unverified

1 commit (0 fixes) modifying scripts/media

A change to existing behaviour in scripts/media — 1 commit, 6 files: “Add bounded media contracts and native decoder validation (\#4341)” (refs \#4341).

scripts/media · medium confidence · unverified

1 commit (0 fixes) modifying src/mods

A change to existing behaviour in src/mods — 1 commit, 4 files: “Auto-clear at 33% context: save work state, /clear, "continue" (\#5486)” (refs \#5486).

src/mods · medium confidence · unverified

10 commits (2 fixes) modifying src/ui/i18n

A change to existing behaviour in src/ui/i18n — 10 commits (2 fixs), 13 files: “feat: move install choices into web onboarding (\#5967)”; “fix(server): bound the viewer feed offset on the server runtime (\#5485)”; “feat(context): one Rules + Tools line on top, close the terminal banner on the viewer line (\#5654)”; and 7 more (refs \#5967, \#5485, \#5654, \#5652, \#5827, …).

src/ui/i18n · medium confidence · unverified

10 commits (4 fixes) modifying plugin

A change to existing behaviour in plugin — 10 commits (4 fixs), 84 files: “fix(cost-report): require explicit repository-wide wins lookup (\#6081)”; “chore: bump version to 15.18.3”; “feat(security): opt-in local API token for the worker API and MCP HTTP transport (\#5141) (\#6056)”; and 7 more (refs \#6081, \#5141, \#6056, \#6058, \#6001, …). Also in this area: Add Screenpipe skill for searching and saving desktop history.

(repo-wide) · medium confidence · unverified

11 commits (5 fixes) modifying services/viewer-host

A change to existing behaviour in services/viewer-host — 11 commits (5 fixs), 13 files: “feat(viewer): synchronize guarded reviews and owner decisions”; “fix(viewer-host): keep My Work sharing on for the cloud viewer (\#5433)”; “Fix cloud viewer wake, image pinning and image size found on staging (\#5988)”; and 8 more (refs \#5433, \#5988, \#5986, \#5193, \#5500, …).

services/viewer-host · medium confidence · unverified

18 commits (9 fixes) modifying src/ui/viewer/utils

A change to existing behaviour in src/ui/viewer/utils — 18 commits (9 fixs), 13 files: “Session ID Display, Filter, and Delete (\#3458)”; “fix(viewer): preserve live session counts across catalog refreshes (\#4492)”; “chore: remove over-engineering findings from repo-wide ponytail audit (wave 1)”; and 15 more (refs \#3458, \#4492, \#4885, \#4598, \#4695, …).

src/ui/viewer/utils · medium confidence · unverified

2 commits (0 fixes) modifying docker/e2e

A change to existing behaviour in docker/e2e — 2 commits, 2 files: “refactor(server): rename server-beta → server runtime (plan phase 1)”; “chore: clean out the repo root (\#5259)” (refs \#5259).

docker/e2e · medium confidence · unverified

3 commits (0 fixes) modifying ods

A change to existing behaviour in ods — 3 commits, 12 files: “feat: claude-mem updates itself to the release the owner promotes (\#5500)”; “test: trim the suite and stop replaying rejected capture (\#5663)”; “feat: add native ODS memory integration (\#4980)” (refs \#5500, \#5663, \#4980).

ods · medium confidence · unverified

4 commits (1 fix) modifying examples/llamaindex

A change to existing behaviour in examples/llamaindex — 4 commits (1 fix), 2 files: “feat(llamaindex): add a project-scoped memory workflow”; “fix(llamaindex): preserve exact project scope and isolate verification”; “docs(integrations): group the agent framework guides and drop a setting that doesn't exist (\#6030)”; and 1 more (refs \#6030).

(repo-wide) · medium confidence · unverified

4 commits (1 fix) modifying src/services/vector

A change to existing behaviour in src/services/vector — 4 commits (1 fix), 3 files: “fix(vector): reject unusable norms before database writes (\#5798)”; “Reject incompatible hidden tensors before pooling text embeddings (\#5539)”; “feat(server): search by meaning with pgvector (\#5001)”; and 1 more (refs \#5798, \#5539, \#5001, \#5002).

src/services/vector · medium confidence · unverified

5 commits (2 fixes) modifying plugin/skills/mode-creator

A change to existing behaviour in plugin/skills/mode-creator — 5 commits (2 fixs), 6 files: “fix(modes): validate mode prompts and never let a user file shadow a bundled mode (\#4326)”; “feat: add custom mode creator skill (\#3487)”; “fix(security): exclusive-create random temp names for atomic writes (\#5151)”; and 2 more (refs \#4326, \#3487, \#5151, \#4073, \#3817).

plugin/skills/mode-creator · medium confidence · unverified

50 commits (10 fixes) modifying plugin/skills/request-progress/scripts

A change to existing behaviour in plugin/skills/request-progress/scripts — 50 commits (10 fixs), 1 file: “chore: bump version to 15.10.1”; “chore: bump version to 15.11.1”; “feat(work-view): group work by overall request and show progress from memory”; and 46 more (refs \#5444, \#5986, \#5110, \#5109, \#5054, …).

plugin/skills/request-progress/scripts · medium confidence · unverified

51 commits (10 fixes) modifying plugin/ui/work-view

A change to existing behaviour in plugin/ui/work-view — 51 commits (10 fixs), 39 files: “chore: bump version to 15.10.1”; “fix(setup): lead consent with the benefit and simplify copy (\#5270)”; “feat(work-view): group work by overall request and show progress from memory”; and 47 more (refs \#5270, \#5443, \#5187, \#4845, \#5261, …).

plugin/ui/work-view · medium confidence · unverified

6 commits (3 fixes) modifying src/server/compat

A change to existing behaviour in src/server/compat — 6 commits (3 fixs), 2 files: “fix(server): apply usage guards to legacy ingestion routes (\#5641)”; “fix(server): retain final answers from legacy summaries (\#5759)”; “feat(recovery): scope memories by platform source”; and 3 more (refs \#5641, \#5759, \#3119, \#2383).

src/server/compat · medium confidence · unverified

6 commits (3 fixes) modifying src/server/jobs

A change to existing behaviour in src/server/jobs — 6 commits (3 fixs), 3 files: “fix(server): reclaim generation jobs a dead worker left in processing (\#4560)”; “chore: remove over-engineering findings from repo-wide ponytail audit (wave 1)”; “Server-beta: Postgres storage + independent runtime + BullMQ queue (Phases 1–3) (\#2351)”; and 3 more (refs \#4560, \#2351, \#3119, \#5616, \#2383).

src/server/jobs · medium confidence · unverified

65 commits (23 fixes) modifying src/ui/my-work

A change to existing behaviour in src/ui/my-work — 65 commits (23 fixs), 206 files: “feat(viewer): synchronize guarded reviews and owner decisions”; “fix(setup): lead consent with the benefit and simplify copy (\#5270)”; “fix(my-work): reconcile a memory review with older item statuses”; and 62 more (refs \#5270, \#5967, \#5443, \#5141, \#6056, …).

src/ui/my-work · medium confidence · unverified

7 commits (2 fixes) modifying src/server/services

A change to existing behaviour in src/server/services — 7 commits (2 fixs), 3 files: “chore: remove over-engineering findings from repo-wide ponytail audit (wave 1)”; “feat(server): search by meaning with pgvector (\#5001)”; “feat(vector): the server's embedding model as a setting, with multilingual models (\#5002)”; and 4 more (refs \#5001, \#5002, \#2383, \#3021, \#5693, …).

src/server/services · medium confidence · unverified

Centralized privacy validation with improved handling of missing prompts

A new PrivacyCheckValidator has been introduced to centralize user prompt privacy checks, replacing the previous boolean-based logic. This change specifically addresses an issue where missing prompt rows (due to race conditions during session initialization) were incorrectly treated as private, which previously caused observations to be silently dropped. The new validator now distinguishes between genuinely redacted prompts (which are suppressed) and absent prompt rows (which are allowed with a warning), ensuring that observations and summaries are no longer lost when a prompt row is merely missing.

src/services/worker/validation · high confidence

Installer replaced with npx-based installation

The previous curl-pipe-bash and bundled JavaScript installers have been removed and replaced with a redirect to the new npx-based installation method. Users are now instructed to run 'npx claude-mem install' (requiring Node.js \>= 20) instead of the old installation scripts. The install.cmem.ai domain now serves these replacement messages via install.sh and installer.js, configured with Vercel rewrites and appropriate content-type headers.

install · high confidence

Introduce distinct Agent and Human context formatters with optimized ID display

The context formatter module has been restructured into two specialized components: AgentFormatter, which outputs plain-text context for machine consumption, and HumanFormatter, which renders the same data with ANSI color codes and enhanced readability for users. A new id-display utility automatically abbreviates Postgres UUIDs to 8-character prefixes when by-id fetching is unsupported (server runtime), significantly reducing tokenizer costs while preserving full identifiers in search results. This change also introduces a 'Previously' section in the context output to display prior assistant messages, improving continuity in multi-turn interactions.

src/services/context/formatters · high confidence

Introduces structured error handling and security hardening for the server runtime

The server now includes a dedicated error handling module that standardizes API error responses with structured codes and messages, replacing ad-hoc error logic. Security is strengthened by applying hardening headers (such as X-Content-Type-Options and X-Frame-Options) to all responses and validating allowed operations and topics for the /api/instructions endpoint to prevent path traversal and injection. Additionally, the server startup and shutdown processes are hardened to prevent zombie processes and ensure clean teardown, while a new utility ensures responses are flushed before process exit. Also in this area: 1 commit (0 fixes) modifying (repo-wide); 2 commits (1 fix) modifying src/services/server; Support for idle-based worker exit.

src/services/server · low confidence · unverified

MCP server runtime scoping and protocol protection

The MCP server now intercepts console output to prevent JSON-RPC protocol interference and introduces runtime-based tool visibility: when the runtime is set to 'server', all tools are available, but for other runtimes (e.g., 'worker'), tools previously restricted to 'server-beta' (such as observation and memory management tools) are hidden from the advertised tool list. Also in this area: 2 commits (1 fix) modifying src/servers; 2 commits (0 fixes) modifying src/servers; Opt-in loopback HTTP transport for MCP server; Enhanced marketplace detection and configurable semantic search settings; Improved search error handling and date validation; and 3 more.

src/servers · medium confidence · unverified

The context service now uses dedicated renderers for each section (Header, Footer, Summary, Timeline, WorkState) to improve modularity and control output size. The FooterRenderer conditionally displays token economics (discovery and read tokens) only when enabled and positive, using separate Agent and Human formatters. The HeaderRenderer adds environment labels to project names and includes legends, column keys, and context economics. The SummaryRenderer ensures the last summary is shown if it contains content and is newer than the latest observation. The TimelineRenderer groups observations by day and supports full or table-row views based on configuration. The WorkStateRenderer introduces a persistent to-do list and working state system, rendering open tasks and state fields with relative timestamps, while enforcing a character limit to preserve the context window budget. Also in this area: Accurate subagent lifecycle tracking and finish detection in Work View; Keyword search performance and My Work context filtering.

src/services/context/sections · high confidence

Native Claude plugin commands replace custom installer

The npx CLI now uses native Claude plugin commands instead of a custom installer. This change introduces a new harness worker mechanism that manages the Claude memory service lifecycle, including automatic startup, health checks, and connection handling via a dedicated worker process. The CLI also provides commands for installing, updating, and uninstalling these plugins, with improved help text for action verbs and better handling of non-interactive installs.

src/npx-cli · high confidence

Prevention of duplicate user prompts via recent-duplicate detection

The system now includes logic to detect and prevent duplicate user prompts within a specified time window. A new service function queries the SQLite database to find recent prompts with identical text associated with the current session or content session, ensuring that accidental repeated submissions are identified. This change introduces a specific check mechanism to maintain prompt uniqueness during user interactions.

src/services/sqlite/prompts · high confidence

Telemetry overhaul: new privacy controls, buffering, and scrubbing

The telemetry service has been rewritten to introduce a default opt-out policy (telemetry is off unless the user explicitly enables it or sets DO\_NOT\_TRACK), along with a granular error-tracking kill-switch via the CLAUDE\_MEM\_TELEMETRY\_ERRORS environment variable. To reduce PostHog costs and improve performance, high-volume events are now aggregated into 5-minute time-window rollups or per-session compressed rollups before being sent. The system also includes a one-time historical backfill to ingest anonymized daily activity rollups from the local SQLite database into PostHog, and a dedicated CLI transport for short-lived installer events. All outgoing data is strictly sanitized using a whitelist scrubber for structured properties and an aggressive allow-then-redact scrubber for error messages and stacks to prevent PII leakage.

src/services/telemetry · high confidence

Viewer UI constants and configuration defaults consolidated

The viewer's UI layer now uses a dedicated set of constants files to centralize configuration and display logic. This includes API endpoint definitions, UI timing and pagination settings, and a comprehensive list of OpenAI-compatible provider presets (such as NVIDIA NIM, Groq, and Ollama) for the settings modal. Additionally, default model settings (defaulting to claude-sonnet-5), trial promotion details (30-day free trial), and specific feature flags (like the file read gate and subagent observation filtering) are now explicitly defined here, ensuring the viewer's behavior and appearance are driven by these local constants.

src/ui/viewer/constants · high confidence

Fixes

10 commits (6 fixes) fixing services/sync-hub

A fix in services/sync-hub — 10 commits (6 fixs), 32 files: “fix(sync-hub): cut Durable Object row writes per request on the DO path (\#4424)”; “chore: remove old files and dead code found by a repo-wide audit (\#5261)”; “fix(sync): release the projection lease when nothing is in flight; Retry-After on projection\_busy (\#3488)”; and 7 more (refs \#4424, \#5261, \#3488, \#5663, \#5332, …).

services/sync-hub · medium confidence · unverified

10 commits (8 fixes) fixing src/services/media

A fix in src/services/media — 10 commits (8 fixs), 11 files: “fix(media): keep every captured screenshot locally and store it small (lossy WebP q80, 1 MiB canonical, 20 s waiting conversion, fit instead of reject) (\#5431)”; “feat(media): keep images by default and link them without inference (\#5426)”; “fix(media): keep transient cloud read failures retryable (\#5614)”; and 7 more (refs \#5431, \#5426, \#5614, \#5601, \#5653, …).

src/services/media · medium confidence · unverified

103 commits (70 fixes) fixing plugin/hooks

A fix in plugin/hooks — 103 commits (70 fixs), 6 files: “fix: replace GNU sort -V with POSIX-portable version sort”; “fix: patch 7 critical bugs affecting all non-dev-machine users and Windows”; “feat: add smart-explore AST-based code navigation (\#1244)”; and 100 more (refs \#1244, \#3470, \#4760, \#1330, \#614, …).

plugin/hooks · medium confidence · unverified

11 commits (8 fixes) fixing hermes

A fix in hermes — 11 commits (8 fixs), 5 files: “fix(hermes): honor skipped prompt acknowledgements from the worker (\#5825)”; “Add Hermes Agent memory support (\#4905)”; “fix(hermes): strip memory tags with a linear scan (\#5320)”; and 8 more (refs \#5825, \#4905, \#5320, \#5678, \#5425, …).

hermes · medium confidence · unverified

11 commits (9 fixes) fixing src/integrations/pi-extension

A fix in src/integrations/pi-extension — 11 commits (9 fixs), 1 file: “fix(pi): give the session-start context request the hooks' API budget (\#5115) (\#5118)”; “fix(pi): find the system message by role so a leading custom message keeps capture on (\#4968) (\#4975)”; “fix(pi): remove private output before truncating capture (\#5793)”; and 8 more (refs \#5115, \#5118, \#4968, \#4975, \#5793, …).

src/integrations/pi-extension · medium confidence · unverified

12 commits (7 fixes) fixing src/services/backup

A fix in src/services/backup — 12 commits (7 fixs), 4 files: “fix(backup): create SQLite snapshots with owner-only permissions (\#5642)”; “fix(restore): report committed swaps despite sidecar cleanup errors (\#5624)”; “Honor backup intervals beyond the timer limit (\#5784)”; and 9 more (refs \#5642, \#5624, \#5784, \#5781, \#5746, …).

src/services/backup · medium confidence · unverified

132 commits (78 fixes) fixing src/services/work-view

A fix in src/services/work-view — 132 commits (78 fixs), 65 files: “feat(viewer): synchronize guarded reviews and owner decisions”; “fix(sync): schedule fair uploads with scoped prerequisites”; “fix(work-view): a locked database at startup no longer fails the worker (\#6041)”; and 129 more (refs \#6041, \#5782, \#5711, \#5245, \#5991, …).

src/services/work-view · medium confidence · unverified

14 commits (9 fixes) fixing src/integrations/dsh-plugin

A fix in src/integrations/dsh-plugin — 14 commits (9 fixs), 12 files: “feat(security): opt-in local API token for the worker API and MCP HTTP transport (\#5141) (\#6056)”; “fix(dsh): keep requested records, honor exclusions with an override, unblock uninstall”; “docs: state the token claim as a per-result size ratio (\#4138) (\#5889)”; and 11 more (refs \#5141, \#6056, \#4138, \#5889, \#5548, …).

src/integrations/dsh-plugin · medium confidence · unverified

15 commits (10 fixes) fixing src/storage/sqlite

A fix in src/storage/sqlite — 15 commits (10 fixs), 8 files: “fix(server): sanitize SQLite memory content before storage (\#5800)”; “chore: remove over-engineering findings from repo-wide ponytail audit (wave 1)”; “feat(server): plan-07 boot correctness + auth hardening + rename”; and 12 more (refs \#5800, \#2351, \#3119, \#5774, \#4663, …).

src/storage/sqlite · medium confidence · unverified

168 commits (115 fixes) fixing src/utils

A fix in src/utils — 168 commits (115 fixs), 14 files: “Refactor SettingsDefaultsManager: Move to shared directory and update imports”; “feat(grok-bot): Phase 0+1 awareness breathing for LFG and Orifice”; “fix(project-name): use parent/worktree composite so observations don't cross worktrees”; and 164 more (refs \#4791, \#5174, \#974, \#1820, \#5224, …).

src/utils · medium confidence · unverified

18 commits (14 fixes) fixing src/services/memory

A fix in src/services/memory — 18 commits (14 fixs), 5 files: “fix(memory): preserve Unicode at native note preview limits (\#5707)”; “feat(integrations): export new observations to a Notion database (\#4799)”; “fix(projects): preserve whitespace in checkout identity (\#5368)”; and 15 more (refs \#5707, \#4799, \#5368, \#5469, \#5687, …).

src/services/memory · medium confidence · unverified

19 commits (11 fixes) fixing services/sync-api

A fix in services/sync-api — 19 commits (11 fixs), 17 files: “fix(sync-api): build on bun 1.4 image (bun.lock v2 unreadable by 1.2) (\#4233)”; “Move CMEM Pro sync hub off Cloudflare to Fly + Neon (\#4232)”; “fix(sync): get memory pushes past Supabase's Cloudflare WAF (\#4547)”; and 16 more (refs \#4233, \#4232, \#4547, \#5193, \#5573, …).

services/sync-api · medium confidence · unverified

21 commits (14 fixes) fixing src/npx-cli/utils

A fix in src/npx-cli/utils — 21 commits (14 fixs), 5 files: “fix(install): never abort before sign-in over a port rebound after the worker stopped; keep the working cache version (\#4289)”; “feat: move install choices into web onboarding (\#5967)”; “fix: harden sqlite migrations and settings writes”; and 17 more (refs \#4289, \#5967, \#3119, \#5500, \#3080, …).

src/npx-cli/utils · medium confidence · unverified

21 commits (18 fixes) fixing src/launcher

A fix in src/launcher — 21 commits (18 fixs), 6 files: “fix(launcher): clear the old signature before the ad-hoc re-sign on darwin (\#4947) (\#4988)”; “fix(security): keep the current directory out of executable discovery (\#5148)”; “feat(skills): allow opting out of bundled plugin skills (\#5837)”; and 18 more (refs \#4947, \#4988, \#5148, \#5837, \#4894, …).

src/launcher · medium confidence · unverified

24 commits (15 fixes) fixing src/services/agent-runner

A fix in src/services/agent-runner — 24 commits (15 fixs), 4 files: “fix(work-view): keep the My Work refresh on the configured agent instead of falling back to another (\#5245)”; “fix(agent-runner): run Codex and Gemini refreshes with no MCP tools”; “fix(agent-runner): wait for each agent CLI instead of timing out”; and 21 more (refs \#5245, \#5140, \#3607, \#5372, \#5777, …).

src/services/agent-runner · medium confidence · unverified

27 commits (22 fixes) fixing src/services/hooks

A fix in src/services/hooks — 27 commits (22 fixs), 3 files: “fix(server-beta): accept contentSessionId on /v1/memories, mirroring /v1/events”; “fix: harden sqlite migrations and settings writes”; “fix(server): provision bootstrap keys in one transaction (\#5815)”; and 23 more (refs \#5815, \#3119, \#5774, \#3080, \#3472, …).

src/services/hooks · medium confidence · unverified

28 commits (18 fixes) fixing cowork

A fix in cowork — 28 commits (18 fixs), 37 files: “fix(cowork): redact the key shapes the shim still sent verbatim (\#5543)”; “fix(report): find installed Windows Chrome for PDF export (\#4404)”; “feat(agent-cost-report): read-only GitHub merged-PR wins source (\#4241)”; and 25 more (refs \#5543, \#4404, \#4241, \#6081, \#6085, …).

cowork · medium confidence · unverified

29 commits (23 fixes) fixing src/services/worker/knowledge

A fix in src/services/worker/knowledge — 29 commits (23 fixs), 7 files: “fix(corpus): preserve observation type filters (fixes \#3892)”; “Preserve source evidence and identities in knowledge corpora (\#5528)”; “fix(knowledge): retain complete multipart answers (\#6002)”; and 26 more (refs \#3892, \#5528, \#6002, \#4874, \#2219, …).

src/services/worker/knowledge · medium confidence · unverified

34 commits (31 fixes) fixing src/services/prompt-import

A fix in src/services/prompt-import — 34 commits (31 fixs), 6 files: “fix(import): exclude copied OpenCode fork activity (\#6004)”; “fix(import): read current OpenCode history without duplicate legacy rows (\#5788)”; “fix(import): honor Gemini JSONL history mutations (\#5865)”; and 31 more (refs \#6004, \#5788, \#5865, \#5886, \#5989, …).

src/services/prompt-import · medium confidence · unverified

42 commits (28 fixes) fixing src/npx-cli/install

A fix in src/npx-cli/install — 42 commits (28 fixs), 8 files: “fix(installer): prepare setup before opening the browser (\#5631)”; “feat: move install choices into web onboarding (\#5967)”; “Honor a compatible PATH tree-sitter during repair (\#5576)”; and 37 more (refs \#5631, \#5967, \#5576, \#5753, \#4551, …).

src/npx-cli/install · medium confidence · unverified

5 commits (3 fixes) fixing python

A fix in python — 5 commits (3 fixs), 2 files: “fix: preserve exact project scope and CLI failure reasons”; “fix(crewai): preserve exact project scope and isolate verification”; “docs(integrations): add Microsoft Agent Framework memory workflow”; and 2 more.

python · medium confidence · unverified

5 commits (3 fixes) fixing src/integrations

A fix in src/integrations — 5 commits (3 fixs), 1 file: “fix(pi): give the session-start context request the hooks' API budget (\#5115) (\#5118)”; “feat(security): opt-in local API token for the worker API and MCP HTTP transport (\#5141) (\#6056)”; “fix(integrations): honor numeric harness settings (\#5756)”; and 2 more (refs \#5115, \#5118, \#5141, \#6056, \#5756, …).

src/integrations · medium confidence · unverified

7 commits (4 fixes) fixing src/services/setup

A fix in src/services/setup — 7 commits (4 fixs), 3 files: “feat: move install choices into web onboarding (\#5967)”; “Measure the actual browser installer funnel (\#5443)”; “fix(setup): report unavailable history source without crashing (\#5757)”; and 4 more (refs \#5967, \#5443, \#5757, \#5661, \#5031, …).

src/services/setup · medium confidence · unverified

8 commits (6 fixes) fixing src/server/auth

A fix in src/server/auth — 8 commits (6 fixs), 4 files: “feat(server): plan-07 boot correctness + auth hardening + rename”; “fix: use health oracle and configured worker endpoint”; “Server-beta: Postgres storage + independent runtime + BullMQ queue (Phases 1–3) (\#2351)”; and 4 more (refs \#2351, \#3119, \#2541, \#5160, \#5455, …).

src/server/auth · medium confidence · unverified

Context injection now fits within the 10,000-character hook output limit

The context service now guarantees that the injected context block stays under the 10,000-character limit imposed by Claude Code's hook output. Previously, exceeding this limit caused the block to be silently replaced by a small preview stub, leaving the model without context. The new \ContextBudget\ system measures the rendered output and, if it exceeds the limit, iteratively drops the most expensive content—full observation narratives, last-session summaries, prior messages, and session history—until the block fits. Additionally, the service now fetches context rows from the shared server store for server-runtime sessions, ensuring the local corpus is supplemented by the broader server-side data, and includes token economics calculations to show users how much context was saved.

src/services/context · high confidence

Fixes observation retrieval and storage robustness in SQLite backend

The SQLite observation service now correctly handles file paths stored as bare strings in the \files\_read\ and \files\_modified\ columns, ensuring that context injection works consistently even when path formats (absolute vs. relative) differ between events. It also introduces a null-byte delimiter in the observation content hash to prevent collisions and adds logic to scope observations by platform source and project, improving data isolation and retrieval accuracy.

src/services/sqlite/observations · high confidence

Improved worker shutdown reliability and diagnostics

The installer now uses a new shutdown helper that determines worker liveness based on verified PID file ownership rather than relying solely on network connection states. This prevents false aborts during installation when a port is rebound by another process after the worker stops, and correctly handles scenarios where the worker process has exited but the port remains held. Users will experience fewer installation failures due to ambiguous connection errors and receive clearer diagnostics if a shutdown is blocked by a still-running worker or a port rebound.

src/services/install · high confidence

Infrastructure hardening: graceful shutdown, health probes, and v12.4.3 cleanup

The infrastructure layer now includes a new GracefulShutdown module that ensures a complete teardown sequence (HTTP server, session manager, MCP client, Chroma MCP, database, and process supervisor) and specifically tolerates the ERR\_SERVER\_NOT\_RUNNING error to prevent aborted shutdowns. Health monitoring has been improved with bounded HTTP probes to /api/health (replacing /api/readiness) and socket-based port checks that are timed out to prevent hanging on ghost listeners, particularly on Windows. Additionally, a one-time v12.4.3 cleanup script has been added to handle database pollution by backing up and purging orphaned observer sessions and stuck pending messages, with safeguards for disk space and Bun runtime compatibility issues.

src/services/infrastructure · high confidence

New observation generation pipeline with reliability and safety improvements

The server now uses a new \ProviderObservationGenerator\ to handle observation generation jobs, introducing several key reliability and safety features. Provider calls are now bounded by a configurable timeout (defaulting to 10 minutes) to prevent hung requests from blocking the generation lane. To avoid context window overflows, session summary inputs are now capped by byte size, preserving the head and tail of the session for better context. The system also handles empty provider responses by failing fast instead of retrying, and includes a mechanism to reclaim jobs left in a 'processing' state by dead workers, ensuring they are either requeued or marked as failed. Additionally, usage metering is now recorded after the main transaction commits to prevent metering failures from rolling back observation data.

src/server/generation · high confidence

Refactored session generator lifecycle with bounded retries and pause/resume logic

The session worker's generator lifecycle has been restructured into dedicated modules (GeneratorRunner, GeneratorExitHandler, SessionCompletionHandler) to improve reliability and prevent unbounded retry storms. The new logic ensures that sessions are preserved and buffered work is retained during specific failure states (quota, auth, rate limits, transport pauses) rather than being immediately finalized. It introduces bounded automatic resumption for response stalls, rate limits, and unattended gateway retries, and implements a conversation recycling mechanism that drops the in-memory history to reset context budget while preserving observations. Additionally, a 'summarize rescue' pass allows a final summary to be processed even if the generator fails, and idle teardowns now correctly spawn a fresh generator if new work arrives during the stop sequence.

src/services/worker/session · high confidence

Search now falls back to SQLite when Chroma is unavailable or returns empty results

The search worker now ensures results are never silently lost due to Chroma issues. If Chroma is disabled, unreachable, or returns zero matches for requested categories (such as with CJK queries or specific filters), the system automatically falls back to SQLite keyword search. This prevents 400 errors for filter-only searches and ensures users still receive relevant observations, sessions, or prompts even when semantic search is offline or incomplete.

src/services/worker/search · high confidence

Unified shell template for plugin hook resolution

The plugin hook resolution logic is now generated from a single source of truth (\src/build/hook-shell-template.ts\) instead of being hand-maintained in separate JSON config files. This ensures that the fallback chain for locating plugin roots—checking the host-injected root, development directories, versioned cache directories, and marketplace installs—is consistent across Claude Code setup, Codex CLI, and MCP hosts. The change also fixes Windows-specific spawn issues by avoiding broken pipes in Cygwin/MSYS shells and ensures that host-injected plugin roots are only accepted if their dependencies (node\_modules) are present, preventing runtime crashes.

src/build · high confidence

Test coverage

1 commit adding/updating tests in tests/evals/read-gate/fixture, tests/evals/read-gate/fixture-large; 1 commit adding/updating tests in tests/fixtures/auto-clear-mod; 1 commit adding/updating tests in tests/fixtures/context; 1 commit adding/updating tests in tests/fixtures/hooks; 1 commit adding/updating tests in tests/fixtures/memory; 1 commit adding/updating tests in tests/fixtures/telemetry; 10 commits adding/updating tests in tests/hermes; 109 commits adding/updating tests in tests/integrations; 11 commits adding/updating tests in services/viewer-host/test; 11 commits adding/updating tests in tests/fixtures/worker; 13 commits adding/updating tests in tests/backup; 13 commits adding/updating tests in tests/ui/my-work; 136 commits adding/updating tests in tests/infrastructure; 16 commits adding/updating tests in tests/context/formatters; 16 commits adding/updating tests in tests/media; 17 commits adding/updating tests in services/sync-api/test; 18 commits adding/updating tests in tests/helpers; 2 commits adding/updating tests in tests/fixtures/windsurf; 2 commits adding/updating tests in tests/services/hooks; 2 commits adding/updating tests in tests/services/transcripts; 20 commits adding/updating tests in tests/memory; 209 commits adding/updating tests in tests/worker/http/routes; 21 commits adding/updating tests in tests/worker/knowledge; 23 commits adding/updating tests in tests/launcher; 23 commits adding/updating tests in tests/storage/postgres; 25 commits adding/updating tests in tests/supervisor; 25 commits adding/updating tests in tests/worker/search; 26 commits adding/updating tests in tests/services/agent-runner; 29 commits adding/updating tests in tests/cli/adapters; 295 commits adding/updating tests in tests/worker; 3 commits adding/updating tests in examples/crewai; 3 commits adding/updating tests in tests/integration/fixtures; 32 commits adding/updating tests in tests/services/prompt-import; 33 commits adding/updating tests in tests/scripts; 37 commits adding/updating tests in tests/ui/viewer; 39 commits adding/updating tests in tests/server/generation; 39 commits adding/updating tests in tests/servers; 39 commits adding/updating tests in tests/services; 46 commits adding/updating tests in tests/cli; 5 commits adding/updating tests in tests/compat; 5 commits adding/updating tests in tests/fixtures; 5 commits adding/updating tests in tests/services/setup; 5 commits adding/updating tests in tests/services/vector; 5 commits adding/updating tests in tests/services/worker/session; 5 commits adding/updating tests in tests/ui/work-view; 51 commits adding/updating tests in tests/services/sqlite; 56 commits adding/updating tests in tests/transcripts; 6 commits adding/updating tests in tests/integration/helpers; 6 commits adding/updating tests in tests/server/jobs; 60 commits adding/updating tests in tests/worker/work-view; 67 commits adding/updating tests in tests/context; 68 commits adding/updating tests in tests/worker/sync; 7 commits adding/updating tests in tests/evals; 7 commits adding/updating tests in tests/fixtures/openclaw; 7 commits adding/updating tests in tests/storage/sqlite; 74 commits adding/updating tests in tests/cli/handlers; 78 commits adding/updating tests in tests/npx-cli; 81 commits adding/updating tests in tests/work-view; 89 commits adding/updating tests in tests/sqlite; 9 commits adding/updating tests in tests/context/sections; 9 commits adding/updating tests in tests/cowork; 9 commits adding/updating tests in tests/services/worker; 9 commits adding/updating tests in tests/worker/http; Added comprehensive test coverage for telemetry subsystem; Added server test suite covering API keys, context, deletion, and readiness; Added test coverage for search strategy implementations; Added test coverage for the server runtime and job queue infrastructure; Added test coverage for viewer utility functions and components; Added test fixtures for custom provider module shapes; Added test fixtures for opencode tool-file evidence and worker endpoint behavior; Added tests for CLAUDE.md context utilities, logger formatting, and project filtering; Added tests for CORS middleware origin restrictions and preflight headers; Added tests for FTS maintenance, project merging, and worktree adoption; Added tests for Observer session tool enforcement and hardened SDK options; Added tests for Redis queue configuration validation; Added tests for Windows spawn contract and macOS Codex bundle detection; Added tests for hooks file-context, runtime-selector, and server-client; Added tests for shared infrastructure components; Added tests for smart-file-read parser and search improvements; Added tests for the remote recall MCP server; Added tests for worker agent error handling and response processing; Expanded SDK test coverage for observer behavior, parsing, and configuration; Expanded integration test coverage for installers, lifecycle, and hooks; Expanded test coverage for Chroma sync and MCP manager behaviors; Expanded test coverage for core integration and worker logic.

Dependencies

Initial dependency manifests for claude-mem, plugin, and sync-hub

This change introduces the foundational package.json files for the project's core components. The main \claude-mem\ package (v13.25.3) defines its runtime dependency on \better-auth\ and lists build-time dependencies like \@anthropic-ai/claude-agent-sdk\, \bullmq\, and \express\ which are bundled into the worker. The \plugin\ package (v13.25.3) specifies the runtime dependencies for bundled hooks, including \zod\ and a comprehensive set of \tree-sitter\ grammars for code parsing. Additionally, a new \sync-hub\ worker package is added with dependencies for Cloudflare Workers deployment (\wrangler\, \miniflare\).

(dependencies) · high confidence

Refactor database integration to use bun:sqlite instead of better-sqlite3

The SQLite service layer has been migrated from the \better-sqlite3\ library to Bun's native \bun:sqlite\ module. This change updates the core database connection, session storage, search, and schema management files to use the new driver, ensuring compatibility with the Bun runtime environment while maintaining existing database functionality and schema versions.

src/services/sqlite · high confidence

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

Score

  • CAI 27 → 30 (+2.9)
  • Rubric changed (rubric-2026.10.5 → rubric-2026.11.3) — scores are not directly comparable.

Lenses

  • Code Health 58 → 61 (+2.1)
  • Architecture 85 → 84 (-1.7)
  • Maturity 84 → 91 (+6.7)
  • Readiness 27 → 33 (+5.9)
  • Security 44 → 51 (+7.0)
  • Domain Modelling 10 → 10 (+0.0)
  • Accessibility 52 → 57 (+5.3)
  • Performance 60 → 60 (+0.0)

Resolved (367)

  • Boundary-crossing change coupling: SessionRoutes.ts ↔ observer-health.ts (src/services/worker/http/routes/SessionRoutes.ts)
  • Boundary-crossing change coupling: file-context.ts ↔ get.ts (src/cli/handlers/file-context.ts)
  • Change coupling: CodexCliInstaller.ts ↔ config.ts (src/services/integrations/CodexCliInstaller.ts)
  • Change coupling: OpenCodeInstaller.ts ↔ WindsurfHooksInstaller.ts (src/services/integrations/OpenCodeInstaller.ts)
  • Change coupling: SessionRoutes.ts ↔ GeneratorExitHandler.ts (src/services/worker/http/routes/SessionRoutes.ts)
  • Change coupling: SessionStore.ts ↔ SyncApply.ts (src/services/sqlite/SessionStore.ts)
  • Change coupling: ide-detection.ts ↔ uninstall.ts (src/npx-cli/commands/ide-detection.ts)
  • ClassTooLong: SessionRoutes (src/services/worker/http/routes/SessionRoutes.ts)
  • CloudSync.drainContentOutbox (cognitive 20) (src/services/sync/CloudSync.ts)
  • CloudSync.drainMutations (cognitive 64) (src/services/sync/CloudSync.ts)
  • CloudSync.drainMutations (cyclomatic 29) (src/services/sync/CloudSync.ts)
  • CodexCliInstaller.tomlStructuralLines (cognitive 25) (src/services/integrations/CodexCliInstaller.ts)
  • CodexCliInstaller.tomlStructuralLines (cyclomatic 18) (src/services/integrations/CodexCliInstaller.ts)
  • Documentation: contradicts the code (docs/anti-pattern-cleanup-plan.md)
  • Duplicated block (10 lines × 5) (claude-mem-cursor/skills/agent-cost-report/scripts/acr/mistakes.py)
  • Duplicated block (10 lines × 5) (claude-mem-cursor/skills/agent-cost-report/scripts/acr/mistakes.py)
  • Duplicated block (10 lines × 5) (claude-mem-cursor/skills/agent-cost-report/scripts/acr/rollup.py)
  • Duplicated block (10 lines × 5) (claude-mem-cursor/skills/agent-cost-report/scripts/acr/transcripts.py)
  • Duplicated block (11 lines × 5) (claude-mem-cursor/skills/agent-cost-report/scripts/acr/patterns.py)
  • Duplicated block (12 lines × 5) (claude-mem-cursor/skills/agent-cost-report/scripts/acr.py)
  • …and 347 more

New (1103)

  • AgentTextIndex.add (cognitive 18) (src/services/prompt-import/owner-prompt-filter.ts)
  • ApiCorpusAgent.scoreObservation (cognitive 20) (src/services/worker/knowledge/ApiCorpusAgent.ts)
  • AutoClearController.onTurnEnd (cognitive 18) (src/mods/auto-clear/controller.ts)
  • AutoClearController.onTurnEnd (cyclomatic 17) (src/mods/auto-clear/controller.ts)
  • Banned license: @img/sharp-libvips-darwin-arm64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-darwin-x64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linux-arm (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linux-arm64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linux-ppc64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linux-riscv64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linux-s390x (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linux-x64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linuxmusl-arm64 (plugin/bun.lock)
  • Banned license: @img/sharp-libvips-linuxmusl-x64 (plugin/bun.lock)
  • Banned license: @img/sharp-wasm32 (plugin/bun.lock)
  • Banned license: @img/sharp-win32-arm64 (plugin/bun.lock)
  • Banned license: @img/sharp-win32-ia32 (plugin/bun.lock)
  • Banned license: @img/sharp-win32-x64 (plugin/bun.lock)
  • Boundary-crossing change coupling: OpenRouterObservationProvider.ts ↔ OpenRouterProvider.ts (src/server/generation/providers/OpenRouterObservationProvider.ts)
  • Boundary-crossing change coupling: observation.ts ↔ shared.ts (src/cli/handlers/observation.ts)
  • …and 1083 more

Changes since last survey

  • 300 commits — 111 feature/other, 189 fixes

By area

  • src/services — 90 commits
  • docs/public — 41 commits
  • docs/CHANGELOG.md — 9 commits
  • src/npx-cli — 8 commits
  • src/ui — 8 commits
  • releases/promoted-release.json — 7 commits
  • src/server — 7 commits
  • plugin/scripts — 6 commits
  • src/shared — 6 commits
  • tests/worker — 6 commits
  • src/integrations — 5 commits
  • tests/services — 5 commits
  • examples/haystack — 4 commits
  • examples/smolagents — 4 commits
  • hermes/claude-mem — 4 commits
  • integrations/eve — 4 commits
  • plugin/skills — 4 commits
  • plugin/ui — 4 commits
  • src/cli — 4 commits
  • .github/workflows — 3 commits

Notable commits

  • fix: fix(work-view): cap human-copy validation retries and surface failure reason (#6037) (#6039)
  • fix: Fix cloud viewer wake, image pinning and image size found on staging (#5988)
  • fix: fix(advisor): preserve project case variants in call listings (#6009)
  • fix: fix(ag2): keep worktree memory reads in the selected project
  • fix: fix(ag2): preserve literal project keys across MCP reads
  • fix: fix(agent-runner): reject content-filtered API answers (#5878)
  • fix: fix(auth): keep SQLite key derivation off the request loop (#5808)
  • fix: fix(autogen): keep worktree memory reads in the selected project
  • fix: fix(autogen): preserve literal project keys across MCP reads
  • fix: fix(capture): suppress advisor records from private prompts (#5747)
  • fix: fix(chroma): bootstrap large projects without spreading row IDs (#5862)
  • fix: fix(chroma): load backfill bodies inside row writer turns (#5803)
  • fix: fix(chroma): preserve live metadata during project merge (#5881)
  • fix: fix(ci): warm Chrome before the integration lane (#6070)
  • fix: fix(cli): make live observer handoff timeout tunable and log skips at warn (#5676) (#5831)
  • fix: fix(cli): preserve completed setup during updates (#6045)
  • fix: fix(codex): count cache and reasoning usage only once (#5720)
  • fix: fix(codex): preserve cancellation while awaiting turn interruption (#5805)
  • fix: fix(codex): register hooks in the user's hooks.json so one approval lasts across updates (#6058)
  • fix: fix(codex): validate forbidden capabilities in terminal turn items (#5811)
  • …and 280 more

Architecture

  • Unchanged — 0 containers · 1 contexts · 0 edges

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

thedotmack/claude-mem was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 13 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit 0c488b54e761d7f6f4091b3a6e1958712f13b19d — the exact code this score is about.
  • Scored under rubric-2026.11.3 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.63+preprod-4a8cc97a9484.