arthurcolle/agents.erl

This system is an Erlang-based distributed framework for orchestrating autonomous AI agents, featuring a React web interface for real-time monitoring and interaction. It provides infrastructure for multi-agent collaboration, tool composition, and scheduling, supported by integrations with OpenAI and Anthropic models. The platform also includes utilities for dynamic API generation, code scaffolding, and comprehensive system testing.

The Erlang JSON parsing and encoding library jsx (version 3.1.0) has been added to the build. This includes the core modules for decoding, encoding, and parsing JSON, along with configuration and utility modules, making the library available for use in Erlang applications.

_\build/default · high confidence

The \apps/agent\_web/priv/data\ directory now includes a \conversations.json\ file containing sample chat history for various agent personas (such as Educational Assistant, Task Orchestrator, and Code Assistant). This static data file serves as a seed or example dataset for the web-based agent interface, allowing users to view pre-populated conversation threads upon initialization.

_apps/agent\web/priv/data · high confidence

This commit introduces the core Erlang application structure, including the \agents\ and \agent\_web\ applications, along with a React/TypeScript frontend. It establishes the infrastructure for autonomous agent orchestration, featuring a real-time scheduling engine, temporal awareness, and WebSocket-based event streaming. The release includes a comprehensive Makefile and build scripts for compiling the backend and frontend, as well as a suite of demo and utility scripts (e.g., \agent\_demo.erl\, \create\_diverse\_agent\_fleet.erl\) to initialize and manage agent fleets. Additionally, it adds an MIT license, a \.gitignore\ to exclude sensitive files like API keys, and documentation for the new scheduling capabilities.

(repo-wide) · high confidence

The sandbox now includes a complete Erlang/OTP application for managing autonomous agents with tool-calling capabilities and inter-agent communication. This release introduces core modules for agent discovery, messaging, and protocol handling, along with a simplified chat interface that integrates with the OpenAI API. The project is structured with a standard OTP supervision tree, includes a Makefile for build and test automation, and is licensed under Apache 2.0.

sandbox, sandbox/myapp/apps/myapp · high confidence

The \apps/openai/src\ directory now includes a complete set of new modules to support Anthropic Claude API interactions and broader AI model management. This introduces \anthropic\_api\_structure\ and \anthropic\_client\ to handle API endpoint definitions and client logic with specific support for Model Context Protocol (MCP) connectors, including validation and beta header handling. Additionally, \anthropic\_client\_template\ provides code generation capabilities for these API modules. The update also adds \model\_registry\ and \model\_comparison\ to manage a catalog of AI models (including OpenAI and Anthropic) with capabilities, pricing tiers, and ranking utilities, alongside \cost\_tracker\ for monitoring usage costs across different agents and models. Supporting infrastructure includes \json\_serializer\ for consistent JSON handling with \jiffy\, and \openai\_background\ for managing long-running asynchronous tasks.

apps/openai/src · high confidence

Introduced the \openapi\_scaffold\ application, which enables the automatic generation of Erlang API servers, client libraries, and documentation from OpenAPI 3.0+ specifications. This new capability includes a dynamic router for runtime route management, request/response validation, and optional API proxying. It also provides a multi-language code generator (supporting TypeScript, Python, Go, and Rust) and an AI-powered transformation module for intelligent request handling and error explanation. A web-based playground interface is included for interactive API testing, alongside introspection endpoints for exploring loaded specs and a health check handler for system monitoring.

(repo-wide) · high confidence

A modern React frontend using shadcn/ui, Tailwind CSS, and Vite has been added to the agent web application. This new UI provides a comprehensive dashboard for creating, monitoring, and controlling agents, featuring real-time updates via WebSocket, interactive network topology visualization, and a chat interface for communicating with AI agents. It includes specialized components for managing MCP servers, structured outputs, and agent consensus (quorum), along with an adaptive AI interface that personalizes the UI density and layout based on user behavior. Users can switch to this new interface by building the frontend and updating the Erlang handler configuration.

_apps/agent\web/src · high confidence

Added a comprehensive suite of Erlang examples demonstrating advanced capabilities including distributed agent clusters with mesh networking, complex tool composition pipelines (code analysis, autonomous debugging, security audits), and real-time streaming/async processing. The examples also showcase an OpenAPI scaffold demo with multi-language code generation and AI transformations, a quantum-inspired multi-agent system with lock-free coordination, and temporal awareness features for scheduled agent tasks.

examples · high confidence

The config directory now includes a comprehensive set of new configuration files to support the agent system's deployment, logging, and clustering. These include example files for API keys (OpenAI, Anthropic, Slack, Linear, Plaid, Graphlit, GitHub, Jina, Pipedream) and Pipedream MCP integration, deployment configurations for the agent web server and agent system (including chat settings, JSON serialization, CORS, and tool definitions), and various logging configurations (development, focused, streaming tokens). Additionally, there are VM argument files for clustering, performance analysis, and production environments, along with SSL and system configuration files.

config · high confidence

The web application now includes a new CSS stylesheet and a compiled distribution bundle. The stylesheet introduces a modern design system with CSS variables for theming, a responsive grid layout for the agent list and content areas, and interactive hover effects. The distribution bundle integrates KaTeX (version 0.16.22) for rendering mathematical expressions and includes the GitHub Dark syntax highlighting theme for code blocks, alongside a comprehensive set of utility classes for layout and styling.

_apps/agent\web/priv/static · high confidence

The \agent\_web\ documentation has been regenerated for version 0.2.0, now including API references for new modules such as \agent\_initializer\, \agent\_monitoring\_handler\, \agent\_templates\_handler\, and \agent\_web\_sup\. The documentation is built using ExDoc v0.38.2 and provides updated source links to the \agents.erl\ repository.

_apps/agent\web/doc, apps/openai/doc · high confidence

Added a new set of test scripts in the tests directory to verify the startup and basic functionality of various system components. These include tests for the color library and colored logger, autonomous agent infrastructure (including fleet creation and initialization), deep reflection and metacognitive analysis engines, runtime component startup (agency, knowledge graph, goal planner), and HTTPS connectivity for MCP servers.

tests · high confidence

The \apps/agent\_web/frontend\ directory now includes a \package.json\ and \package-lock.json\ for the \agent-orchestration-ui\ application. This establishes the frontend's dependency tree, specifying React 18, Vite 5, TypeScript 5, Tailwind CSS 3, and various UI libraries including Radix UI, Headless UI, Framer Motion, and Recharts.

(dependencies) · high confidence

The \apps/agents\ location now includes the generated ExDoc documentation for version 0.3.0. This update adds the full HTML documentation site, including the API reference and module pages for core components such as \agent\, \advanced\_agent\_system\, \active\_exploration\_engine\, and \agent\_collaboration\, along with the necessary static assets and build artifacts.

apps/agents · high confidence

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

• CAI 34 → 34 (+0.2)
• Rubric changed (rubric-2026.09.12 → rubric-2026.09.18) — scores are not directly comparable.

• Code Health 39 → 39 (+0.0)
• Architecture 100 → 89 (-11.2)
• Maturity 82 → 82 (+0.0)
• Readiness 20 → 20 (-0.2)
• Security 76 → 73 (-3.6)
• Event-Driven 61 → 61 (+0.0)
• Event Sourcing 74 → 74 (+0.0)
• Accessibility 33 → 34 (+0.8)
• Performance 100 (new)

• Coverage not measured — no coverage collector is wired up
• Dependency hygiene PARTLY measured — npm pinning read, dependency currency not (no pnpm-resolved versions to grade)
• Documentation: written for insiders (docs/STREAMING_FUNCTION_CALLS_DOCUMENTATION.md)
• Documentation: written for insiders (docs/STREAMING_IMPLEMENTATION_SUMMARY.md)
• Documentation: written for insiders (docs/UI_REVIEW_FINDINGS.md)

• Dependency not covered by the lockfile: jsx
• Documentation: no project overview (docs/DISTRIBUTION_STRATEGY.md)
• Floating source dependency: jesse
• High CVE: [GHSA redacted] (rebar.lock)
• High CVE: [GHSA redacted] (rebar.lock)
• High CVE: [GHSA redacted] (rebar.lock)
• High CVE: [GHSA redacted] (rebar.lock)
• Off the main sequence: openai
• Outdated: accept
• Outdated: base64url
• Outdated: cowboy
• Outdated: exometer_core
• Outdated: folsom
• Outdated: gproc
• Outdated: gun
• Outdated: hackney
• Outdated: jiffy
• Outdated: observer_cli
• Outdated: prometheus
• Outdated: prometheus_cowboy
• …and 4 more

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

arthurcolle/agents.erl 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.