Building a Coherent AI Dungeon Master with DeepSeek: 1000+ Turns for Under $2
zacurryy · reddit · 2026-08-14
A developer successfully built a coherent AI Dungeon Master using the DeepSeek model.
Typically, LLMs struggle with long-term memory in multi-turn scenarios due to context window limits. This developer overcame the issue by using multiple API calls for different tasks within a single turn. While this architecture is prohibitively expensive with OpenAI or Anthropic APIs, DeepSeek's low pricing made it highly practical.
In real-world testing, despite requiring about three API calls per turn to maintain memory and logic, the total cost for a campaign lasting over 1,000 turns was less than $2, with no noticeable memory leaks or severe hallucinations.
Related event: Dev Builds Long-Memory AI Dungeon Master with DeepSeek(2 posts)→
More from coding & agent
- A vector DB is not LLM memory: long thread maps write/read paths, failure modes and evals — techNmak · 2026-10-16
- OpenUI 1.0 launches open source, turning agent text answers into live interactive UIs — heyshrutimishra · 2026-10-16
- 10 GitHub repos to see what your AI agent really did, from Langfuse to claude-code-log — victor_explore · 2026-10-16
- Mosaic: multi-session agentic RL with hierarchical advantage estimation — LARK-Lab · 2026-10-16
- Recursive Environment Scaling tanks six models' success rates by 57–77 points — Lingxiao Jing · 2026-10-16
- Harvard's MPT: decentralized agent swarms learn self-orchestration via bidding — Harvard · 2026-10-16