BlueFish AIAQI AI Intelligence
⌕
Official documentationNot tested by BlueFish

AI prototypes in your own game

Distinguish NPC dialogue from reinforcement learning and build a dialogue scene with fallback.

Last checked:
A quick glossary

Models generate and reason. Agents use tools to work on tasks. A harness manages an agent’s environment. MCP connects tools and data. Skills are reusable task guidance.

Current task

AI prototypes in your own game

Estimated time

Plan 2–4 practice sessions; an editorial estimate, not measured duration.

Limits & boundaries

Wrong port, unloaded model, insufficient resources or incorrect model ID.

Connection/permission errors, model limitations, timeouts and version differences.

Validation

Build an owned scene with one NPC and demonstrate a reply, timeout and fixed-dialogue fallback.

Next steps

Complete the linked recipes’ verification items before adding one capability; repeat source checks when dependencies change.

Cost

Not verified

Price not verified; check software, model, API, hardware and hosting costs separately.

Implementation steps

  1. 01
    Choose a specifically licensed model fitting your hardware and record its ID and quantization.
  2. 02
    Prepare an isolated Godot test scene and running local model; pin versions.
  3. 03
    Verify model responses with a local HTTP client, then send bounded JSON through HTTPRequest.
  4. 04
    Implement timeouts, error handling, cancellation and fixed-dialogue fallback; treat model text as data.
  5. 05
    Test normal, disconnected and timed-out cases; add a server boundary protecting credentials before distribution.

Implementation steps

  1. Choose a specifically licensed model fitting your hardware and record its ID and quantization.

  2. Prepare an isolated Godot test scene and running local model; pin versions.

  3. Verify model responses with a local HTTP client, then send bounded JSON through HTTPRequest.

  4. Implement timeouts, error handling, cancellation and fixed-dialogue fallback; treat model text as data.

  5. Test normal, disconnected and timed-out cases; add a server boundary protecting credentials before distribution.

Original sources

  1. Ollama official documentation ↗

    docs.ollama.com · 2026-10-15 · Official documentation

  2. LM Studio official documentation ↗

    lmstudio.ai · 2026-10-15 · Official documentation

  3. Godot HTTPRequest official documentation ↗

    docs.godotengine.org · 2026-10-15 · Official documentation

Connected resources

Local runtimeFree software; check license and compute costs

Ollama

A local model runtime and API tool supporting a subset of OpenAI endpoints.

Official documentation
Local runtimeFree software; check license and compute costs

LM Studio

A desktop local-model tool with compatible Chat Completions and Responses endpoints.

Official documentation
Integration platformFree software; check license and compute costs

Godot HTTPRequest

Godot’s official HTTPRequest interface connects external services; AI integration needs custom implementation.

Official documentation