maunappl8

maunappl8 / recursive-langgraph-agent

Public

Recursive LangGraph.js agent: split huge texts, answer per chunk, merge results. TypeScript, LangChain, OpenAI-compatible API, MemorySaver checkpoints, Vitest.

agent chunking langchain langgraph llm
20
0
100% credibility
Found Apr 19, 2026 at 20 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

This open-source project enables asking questions about arbitrarily long documents by splitting them into overlapping chunks, processing each with AI in parallel batches, and hierarchically merging the results into a single coherent answer.

How It Works

1
📖 Discover the tool

You find a handy assistant online that helps answer questions about super long books, reports, or any big text file.

2
🛠️ Set it up simply

Download it to your computer and follow easy steps to get everything ready, just like installing a new app.

3
🔑 Connect smart AI

Share a private password with an AI service like OpenAI so the assistant can use its brainpower for deep thinking.

4
Choose your style
🚀
Quick question

Perfect for asking once about a specific file and getting a fast answer.

💬
Interactive chat

Load a document once and keep asking different questions in a conversation.

5
📄 Load your document

Pick any text file, webpage, or even a PDF, and the tool reads it all in.

6
❓ Ask your question

Type what you want to know, like 'Summarize the main ideas' or 'What are the key facts?'

7
✨ Watch it work

It smartly breaks the long text into bite-sized pieces, gets answers for each super fast, then blends them into one perfect response—you see progress every step.

✅ Get your full answer

Celebrate having a complete, spot-on response that understands the whole huge document, saving you hours of reading.

Sign up to see the full architecture

6 more

Sign Up Free

Star Growth

See how this repo grew from 20 to 20 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is recursive-langgraph-agent?

This TypeScript tool builds a langgraph recursive agent using LangGraph.js and LangChain to handle Q&A on huge documents that blow past LLM context limits. Feed it a massive text or PDF, ask a question, and it chunks the content with overlap, processes chunks in parallel via OpenAI-compatible APIs, then merges answers hierarchically into one coherent response. CLI commands like `npm run dev` or REPL make testing instant, with streaming progress and MemorySaver checkpoints for resuming interrupted runs.

Why is it gaining traction?

Unlike basic recursive chunking github scripts or flat RAG setups, it batches parallel LLM calls for speed on giant files, caps merge depth to avoid endless recursion, and exposes events for real-time UIs. Developers dig the drop-in programmatic API for custom agents, PDF extraction out-of-box, and thread-based checkpoints that survive crashes—perfect for production without submodules recursive github actions hassles. It's a clean github recursive llm wrapper that just works with gpt-4o-mini or local Ollama.

Who should use this?

JS/TS devs building document search apps for legal reviews, research papers, or enterprise reports where texts exceed 100k tokens. AI engineers prototyping langgraph.js agents for recursive language models tasks, or backend teams needing quick CLI validation before scaling to APIs. Skip if you're stuck on Python—it's Node-only.

Verdict

Solid prototype for recursive-langgraph-agent needs at 20 stars and 1.0% credibility score; docs are thorough, tests cover edge cases like large docs and checkpoints, but watch for tweaks as it matures. Grab it for proofs-of-concept on huge texts—builds fast, low risk.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.