M.liaichi

Agno Multi-Agent Examples packages three Python reference systems for coordinated AI work: technical content generation, natural-language agent creation, and candidate screening.

MuLIAICHI/n8n-vs-paydanticai compares a Python, code-first route with an n8n visual-workflow route for an assistant that answers questions from Pydantic AI documentation.

OptiWHoobii is a TypeScript starter kit for teams that need authentication, subscriptions, AI chat, file uploads, and analytics in a single Next.js application.

GroqVoice-Assistant is a Python project for turning spoken input into an AI response and returning that response as audio. Its compact design is a useful reference for builders exploring browser-based conversational voice interfaces.

Hya-ultimate-voice-assistant is a Python LiveKit agent that pairs real-time voice interaction with system actions, web search, GitHub, Google services, and local memory.

mcp-prompts-library packages 20+ parameterized prompt templates as a Python Model Context Protocol server. It gives Claude Desktop, Cursor, and other compatible clients a structured menu of reusable tasks that builders can extend.

Reportify is a Python and React application for processing source documents, searching them by meaning, generating section content with GPT-4, and exporting reports as PDF or DOCX.

Second Brain is an Obsidian vault and Claude Code configuration for turning natural-language captures into linked notes, tasks, projects, and an auditable Git history.

Federated Agent Protocol is an alpha-stage Python monorepo for coordinating work across private domains without centralizing raw private data. It includes a protocol, reference runtime, clients, participants, examples, and specifications.

Mubawab-Housing-Scraper is an Apify Actor that turns mubawab.ma property listings into structured CSV or JSON records, with fields suited to price modeling and market analysis.

second-brain-skills-ml collects Claude Code skills for documentation, MCP integrations, presentations, video work, and skill design. Its central pattern is progressive disclosure: load detailed guidance only when a task calls for it.

Marv is a local web command center that overlays an Obsidian vault’s folders, files, and tasks on a webcam feed. It combines gesture input with Gemini Live interactions for exploring notes in real time.

pydanticai-tla-guard is a small PydanticAI wrapper that proposes checking an agent workflow's state transitions against TLA+ invariants before it moves to the next step.

QwenAir- pairs a Streamlit chat interface and a command-line script with AirLLM to run Qwen1.5-7B-Chat on CPU. It is a compact example for builders exploring local inference without a discrete GPU.

MuLIAICHI/jira-mcp-server connects MCP-compatible AI assistants to Jira Cloud for issue search, project work, sprint visibility, comments, and workflow transitions.

Dataco pairs a batch validation workflow for Moroccan housing listings with an incremental training path that can process newly scraped records without rebuilding its preprocessing step.

MuLIAICHI/rekrute-jobs-scraper is an Apify Actor for collecting Rekrute.com job listings with configurable entry URLs and limits. Its documented output is structured around the fields teams commonly need for job-market monitoring and downstream analysis.

OpenDataLoader PDF extracts Markdown, HTML, and JSON with element-level coordinates from PDFs. Its documented local and hybrid modes target the tradeoff between fast deterministic parsing and harder pages such as scans and complex tables.

Career-Ops organizes job discovery, fit evaluation, tailored CV generation, and application tracking around local files and AI coding CLIs. Its design emphasizes screening opportunities before spending time applying.

Resume Story Finder uses Claude and AG-UI to turn pasted resume text into structured STAR story cards as they are discovered. It is a small example of an agent loop driving a React interface through typed state events.

MuLIAICHI/The_scraper_need_friends frames web scraping as a collaborative Python learning project, with room to practice processing, file formats, cleanup, and code modularity.

AY Framework is an npm-distributed workflow layer for Claude Code and several other AI coding runtimes. It combines task planning, scoped agent ownership, file-based coordination, and two explicit human review gates.

CogDB is a Python project for agent builders who want to store interaction history, evolving facts, and reusable workflows behind one memory API. Its documented retrieval flow also works within a caller-set token budget.

MuLIAICHI/sentinel is a paper-only pump.fun pipeline published with its negative result: stop-loss triggers could fill far below their intended price. Its value is the system design and the evidence for knowing when not to deploy one.

MuLIAICHI/avito-maroc-scraper targets structured extraction from Avito.ma search and category pages by reading the site's documented Next.js data payload rather than relying on CSS selectors.

MuLIAICHI/MuLIAICHI is a GitHub profile repository that maps Mustapha Liaichi’s work across agent memory, retrieval, AI-ready data, and automation. Its value is in connecting those areas to concrete linked projects rather than presenting a single codebase.

TopoRAG is a Python library for inspecting RAG context with topological data analysis, then describing conceptual gaps before that context reaches an LLM. It combines embeddings, persistent homology, and optional LLM-generated suggestions.

Claude Code Lens is a local-first viewer for Claude Code session JSONL, turning scattered transcripts and tool payloads into browsable conversations, collapsed tool calls, diffs, and display-layer secret masking.

This Python project analyzes thousands of n8n marketplace workflows, turns the results into training data, and documents a path to natural-language workflow generation.