Diapers 1 - AI Hallucinations 0 🍼

After a 1-hour night session and a quick diaper change, I finally fixed my #LangGraph agent.

The fix was simple but effective:
- Each tool now returns a JOB ID instead of.. nothing.
- The AI checks the status and waits instead of panic-looping.
- Switched outputs from JSON to Markdown. I knew I should've done it sooner, but I was too lazy 😭

Result: No more hallucinations. The agent is actually behaving like a pro now. 👀

#IndieHacker #AIAgent #DadLife

You're building AI agents with #langgraph ? Maybe you can help me.

I have a problem with my AI Agent: when a tool takes time to return a result, e.g., +1min, it starts hallucinating and filling the void by talking to the user and relaunching the tool multiple times.

The worst is that the tool runs in batch mode, so when it receives a response for one out of ten, it thinks the other 9 have failed.

How can I inform the AI about the progress state of a tool without wasting tokens ?

#aiagents

Remote is scaling AI onboarding for thousands of users with LangChain & LangGraph, but warns about token bloat as contexts grow. A look at how open‑source tooling can accelerate adoption while highlighting the hidden costs of prompt length. Curious how they’re handling it? Read the full story. #LangChain #LangGraph #OpenSourceAI #TokenBloat

🔗 https://aidailypost.com/news/remote-uses-langchain-langgraph-aionboard-thousands-notes-token-bloat

✅ Đã xây dựng hệ thống đa‑đại lý mô phỏng chuỗi cung ứng bằng LangGraph, sử dụng SQLite + SQLAlchemy làm nguồn dữ liệu duy nhất. Các đại lý LLM: nhu cầu, tồn kho, rủi ro, mua sắm, logistics; một coordinator tổng hợp quyết định và gate quyết định tự động hay cần duyệt. Mục tiêu: tối giản, dễ hiểu, không phải sản xuất. Mong nhận phản hồi về vai trò đại lý, điểm yếu khi mở rộng, cách triển khai và tính năng cần thêm.

#AI #LLM #LangGraph #SupplyChain #CôngNghệ #ChuỗiCungỨng #Python #OpenSource

h

Build more effective multi-agent systems by mastering agent handoffs. Kenneth Leung teaches you 2 distinct methods in #LangGraph for routing tasks.

https://towardsdatascience.com/how-agent-handoffs-work-in-multi-agent-systems/

AI 에이전트의 3가지 장기 메모리: 경험·지식·스킬을 저장하는 기술

https://aisparkup.com/posts/7854

Финансовый AI-агент на Python: MCP и CodeAct

Продолжаем строить финансового AI-ассистента на базе MCP-сервера Finam. Сначала создадим классического MCP-агента на LangChain, затем эволюционируем его в CodeAct-архитектуру, где AI пишет Python-код вместо прямых вызовов функций. В итоге получим агента, способного анализировать тысячи акций, строить графики и не переполнять контекстное окно.

https://habr.com/ru/articles/980542/

#ai #mcp #aiagent #python #llm #codeact #langchain #langgraph #mcptools #llmприложения

AI Skills 2025: LangChain, RAG & MLOps—The Complete Guide

https://techlife.blog/posts/ai-skills-2025-langchain-rag-mlops-guide/

#LangChain #RAG #MLOps #LangGraph #DeepAgents #Kubernetes #VectorDatabases #ChromaDB #Weaviate #Pinecone #MLflow