{"id":28153749,"url":"https://github.com/sarita-joshi/multi-agent-llm-article-generation","last_synced_at":"2026-04-11T04:33:16.672Z","repository":{"id":249446911,"uuid":"831522150","full_name":"Sarita-Joshi/multi-agent-llm-article-generation","owner":"Sarita-Joshi","description":"Explore AI writing systems with agents, graphs, adapters, RAG, and multi-model generation workflow.","archived":false,"fork":false,"pushed_at":"2025-05-03T04:17:44.000Z","size":39,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-05-15T06:11:22.574Z","etag":null,"topics":["agentic-ai","ai-tools","article-geenration","generative-ai","graph-ai","langchain","langgraph","llm","llm-agents","multi-agent","openai","prompt-engineering","rag","streamlit"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Sarita-Joshi.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2024-07-20T20:01:19.000Z","updated_at":"2025-05-03T04:41:57.000Z","dependencies_parsed_at":"2025-05-15T06:20:24.500Z","dependency_job_id":null,"html_url":"https://github.com/Sarita-Joshi/multi-agent-llm-article-generation","commit_stats":null,"previous_names":["sarita-joshi/openai-article-generator-using-langchain"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/Sarita-Joshi/multi-agent-llm-article-generation","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Sarita-Joshi%2Fmulti-agent-llm-article-generation","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Sarita-Joshi%2Fmulti-agent-llm-article-generation/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Sarita-Joshi%2Fmulti-agent-llm-article-generation/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Sarita-Joshi%2Fmulti-agent-llm-article-generation/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Sarita-Joshi","download_url":"https://codeload.github.com/Sarita-Joshi/multi-agent-llm-article-generation/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Sarita-Joshi%2Fmulti-agent-llm-article-generation/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":264502387,"owners_count":23618587,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agentic-ai","ai-tools","article-geenration","generative-ai","graph-ai","langchain","langgraph","llm","llm-agents","multi-agent","openai","prompt-engineering","rag","streamlit"],"created_at":"2025-05-15T06:11:00.252Z","updated_at":"2025-12-30T22:12:47.443Z","avatar_url":"https://github.com/Sarita-Joshi.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# 🧠 LLM Article Generator Hub\n\n\u003e A modular, multi-agent, multi-LLM system to explore and benchmark different architectures for AI-powered article generation — from simple prompt pipelines to deep LangGraph reasoning workflows.\n\n\n## 📚 Table of Contents\n\n- [Project Motivation](#-project-motivation)\n- [System Capabilities](#-system-capabilities)\n- [Evolution of the Architecture](#-evolution-of-the-architecture)\n- [Achievements So Far](#-achievements-so-far)\n- [Benchmark Goals](#-benchmark-goals)\n- [Current Implementations](#-current-implementations)\n- [Comparison of Approaches](#-comparison-of-approaches)\n- [Directory Structure](#-directory-structure)\n- [Getting Started](#-getting-started)\n- [Roadmap \u0026 Next Phases](#-features-to-explore)\n\n\n\n## 🧩 Project Motivation\n\nThis project started as a simple OpenAI-based prompt wrapper for generating blog-style articles. As I explored more real-world use cases and enterprise LLM patterns, the system evolved into a **research platform** to test and compare:\n\n- Prompt-only vs Tool-using vs Multi-agent designs  \n- One-shot text vs modular section-based writing  \n- Static generation vs interactive reasoning graphs  \n- Single-model dependence vs LLM portability  \n\nThis repo now acts as a sandbox for testing different approaches to LLM-driven generation systems — not just for the final output, but also for how we structure and scale the process itself.\n\nThe goal is not just to generate content — but to **analyze and benchmark reliability, explainability, scalability, and extensibility** using a simple usecase!\n\n\n\n## 🧠 System Capabilities\n\nThis hub currently supports:\n\n✅ **Multi-Agent Design**  \n- Research agents (Web, Academic, RAG, Validator)  \n- Writer agents (Outline, Section Writer, Editor)  \n- Quality agents (Fact Checker, Plagiarism, SEO)  \n- Controller \u0026 Planner models in graph mode\n\n✅ **Multi-Model Support (LLM Factory)**  \n- OpenAI (GPT-3.5, GPT-4, GPT-4o)  \n- Anthropic (Claude 3 family)  \n- Cohere, Gemini (planned)  \n- Adapter-based config system with shared prompt schema\n\n✅ **Multi-Tool Retrieval**  \n- Wikipedia (cached)  \n- Tavily, Serper, Exa, PubMed, Arxiv  \n- Local file RAG with chunked vector storage (FAISS)\n\n✅ **Execution Modes**  \n- **Quick mode (V1):** Chain-of-agents with optional RAG  \n- **Deep mode (V2):** LangGraph-based reasoning and retry DAG\n\n✅ **Streamlit UI with Mode Switcher**  \n- Form-based UI that adapts between modes  \n- Optional flags (summary, outline, research toggle)  \n- Customizable LLM behavior: temperature, model, etc.\n\n\n\n## 🔄 Evolution of the Architecture\n\n| Stage | Description | Status |\n|-------|-------------|--------|\n| Stage 0 | Prompt wrapper with OpenAI | Completed |\n| Stage 1 | LangChain agent tools (Wikipedia, RAG, etc.) | Completed |\n| Stage 2 | Multi-agent flow with LangGraph | Completed |\n| Stage 3 | Graph-based planner (Minigraph-style) | Planned |\n| Stage 4 | Multimodal + voice-based interaction | In Research |\n\n\n\n## 🔧 Achievements So Far\n\n- Built a working **multi-agent research → writer → editor** system with optional summary\n- Added **MultiPrompt + RAG** pipelines using LangChain tools\n- Developed a **LangGraph DAG** with built-in retries, QA gates, and modular agent nodes\n- Created a **LLM Adapter Factory** for model switching at runtime\n- Designed a **versioned architecture**: `v1/`, `v2/` with flexible UI routing\n- Implemented simple caching, structured prompts, vector search fallback\n- Developed `pytest`-based testing suite for modular nodes\n- Ready to scale with Minigraph, external APIs, Notion/Slack exports\n\n\n\n## 🧪 Benchmark Goals\n\nNot a benchmarking tool (yet), but this project can evolve to help measure:\n\n| Aspect            | Intention                     |\n|-------------------|-------------------------------|\n| Repeatability     | How consistent is generation? |\n| Cost efficiency   | Token usage across workflows  |\n| Output quality    | Clarity, tone, factuality     |\n| Tool effectiveness| When RAG or search helps most |\n| Model flexibility | LLM portability in real flows |\n\n\n\n\n## 🚀 Current Implementations\n\n### ✅ [V1 — Lightweight Agentic System](v1/README.md)\n\n- Tools: Wikipedia, RAG, WebSearch  \n- Agents: Research, Outline, Writer, Summary  \n- Control: Sequential routing (if/else logic)  \n- LLM: OpenAI GPT-3.5 / GPT-4  \n\n### ✅ [V2 — LangGraph Multi-Agent Pipeline](v2/README.md)\n\n- Pods: Research → Writing → Quality  \n- Tools: + Validators (Fact, SEO, Plagiarism)  \n- Architecture: StateGraph with retries + node state  \n- LLM: OpenAI, Claude, etc. via Adapter Factory\n\n\n\n## ⚖️ Comparison of Approaches\n\n| Feature              | V1: RouterChain Agents | V2: LangGraph Pipeline |\n|----------------------|------------------------|------------------------|\n| Flow Control         | Sequential             | Graph DAG              |\n| Modularity           | Medium                 | High                   |\n| Tools                | RAG, WebSearch, Wiki   | All V1 + QA pods       |\n| Revisions / Retry    | Manual                 | Built-in via edges     |\n| LLM Portability      | OpenAI only            | Multi-provider Adapter |\n| Token Budgeting      | None                   | Per-agent budgeting    |\n| Best For             | Fast blogs             | Long-form, formal docs |\n| Planning Layer       | None                   | Planned                |\n| Multimodal Input     | N/A                    | In progress            |\n\n\n\n## 📦 Directory Structure\n\n```\n.\n├── app.py               # Unified Streamlit UI\n├── generate.py          # Delegates to V1 / V2\n├── requirements.txt\n│\n├── v1/                  # RouterChain agentic workflow\n│   └── README.md\n│\n├── v2/                  # LangGraph-based agent system\n│   └── README.md\n```\n\n\n\n## 📌 Getting Started\n\n### 1. Clone the Repo\n\n```bash\ngit clone https://github.com/yourname/llm-article-generator-hub.git\ncd llm-article-generator-hub\n```\n\n### 2. Create Environment\n\n```bash\npython -m venv .venv\nsource .venv/bin/activate  # or .venv\\Scripts\\activate on Windows\n```\n\n### 3. Install Dependencies\n\n```bash\npip install -r requirements.txt\n```\n\n### 4. Configure Keys\n\n```\nOPENAI_API_KEY=\nTAVILY_API_KEY=\n...\n```\n\n\n## 🧭 Features to explore...\n\n- ✅ Multi-agent workflow\n- ✅ LLM Adapter pattern\n- ✅ Tool integration (RAG)\n- Section-wise RAG + adaptive search queries  \n- Minigraph-based planning before generation  \n- Quality control layers: fact-checking, plagiarism, tone analysis  \n- Export support: PDF, Markdown, Notion  \n- Community prompts and evaluation modes  \n- LLM feedback + edit suggestions\n- Guardrails + EvalSuite\n- Multimodal/vision LLM \n\n## 📜 License\n\nMIT License.  \nFeel free to fork, extend, experiment, or collaborate!","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsarita-joshi%2Fmulti-agent-llm-article-generation","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsarita-joshi%2Fmulti-agent-llm-article-generation","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsarita-joshi%2Fmulti-agent-llm-article-generation/lists"}