Launch a Product Hunt Agent with LangGraph

Build a LangGraph agent (TypeScript or Python) with Product Hunt-style tools, stream graph state, and prep it for CometChat.

A compact LangGraph example (TypeScript and Python) that binds a ChatOpenAI model to Product Hunt-inspired tools, loops until tool calls are complete, and streams every graph state. Both servers expose /kickoff as NDJSON with tool call events.

---

What You’ll Build

• A LangGraph with assistant + tool executor nodes.
• Two tools: list_top_posts (sorted by votes) and search_launches (keyword/topic search).
• Streaming runs keyed by thread_id, kept in sync via MemorySaver.
• A template you can expose over HTTP/SSE to connect with CometChat’s AI Agents.

---

Prerequisites

• TypeScript: Node.js 18+ (Node 20 recommended); OPENAI_API_KEY in .env (optional PRODUCT_OPENAI_MODEL, default gpt-4o-mini).
• Python: Python 3.10+; OPENAI_API_KEY in .env (optional MODEL, default gpt-4o-mini).
• CometChat app + AI Agent entry.

---

Quick links

• Repo root: ai-agent-lang-graph-examples
• TypeScript project: typescript/langgraph-product-hunt-agent (src/graph.ts, src/server.ts, .env.example)
• Python project: python/langgraph_product_hunt_agent (agent.py, server.py, .env)

---

How it works

• Tools — list_top_posts and search_launches live in src/graph.ts, backed by helpers in src/data/search.ts. Both return markdown bullets so the assistant can cite results directly.
• Graph — StateGraph(MessagesAnnotation) alternates between the assistant node and the tools node based on shouldCallTools; tool outputs are fed back as ToolMessage objects.
• State — MemorySaver checkpoints per configurable.thread_id let you run multi-turn conversations on the same graph instance.
• Streaming — app.stream emits state snapshots (streamMode: "values"). The console demo prints each message as the model calls tools and drafts the response.

---

Setup (TypeScript)

1. Install

   cd typescript/langgraph-product-hunt-agent && npm install
2. Env

   Copy ../.env.example to .env; set OPENAI_API_KEY (optional PRODUCT_OPENAI_MODEL).

3. Run demo

   npm run demo — “Top Product Hunt style launches right now?”
4. Run server

   npm run server → POST /kickoff on http://localhost:3000.

Setup (Python)

1. Install

   cd python && python -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt
2. Env

   Create .env with OPENAI_API_KEY (optional MODEL).

3. Run server

   python -m langgraph_product_hunt_agent.server → POST /kickoff on http://localhost:8000.

---

Project structure

• TypeScript: Graph src/graph.ts, Demo src/index.ts, Server src/server.ts, Data src/data, Config .env.example + package.json
• Python: Graph agent.py, Server server.py, Data data/, Config .env + requirements.txt

---

Step 1 - Understand the agent tools

buildProductHuntGraph binds both tools to ChatOpenAI (temperature 0 by default). The graph checks each assistant reply for tool_calls, routes to the tools node to execute them, then loops back until there are none left.

---

Streaming API (HTTP)

Event order (TypeScript and Python servers): text_start → text_delta chunks → tool_call_start → tool_call_args → tool_call_end → tool_result → text_end → done (error on failure). Each event includes message_id; echo thread_id/run_id from the client if you want threading.

Example requests:

# TypeScript (localhost:3000/kickoff)
curl -N http://localhost:3000/kickoff \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"What are the top Product Hunt style launches right now?"}]}'

# Python (localhost:8000/kickoff)
curl -N http://localhost:8000/kickoff \
  -H "Content-Type: application/json" \
  -d '{"thread_id":"t1","run_id":"r1","messages":[{"role":"user","content":"What are the top Product Hunt style launches right now?"}]}'