A Prompt Adventure / Reference / v0.14.0
Reference release
FastAPI game, browser interface, tests, PostgreSQL leaderboard, and container configuration.
- Package
- 0.14.0 · Latest
- Runtime
- Docker · Python 3.12
- Challenge model
- GPT-OSS-120B
- Files
- 44
Archive SHA-25680e076fd844aa73d92424b8e26274d5d1729841286598260d1486ad958977479
Available releases
Choose a version
Model contract
Specific models, specific behavior
This release is built and tested specifically for openai/gpt-oss-120bas the challenge model. It expects a 131,072-token context and a defined set of Chat Completions and Responses behaviors. It is not presented as a generic OpenAI-compatible application.
The original optional artwork path used a separately served Bonsai Image 4B binary Gemlite model. The archive includes the image-service contract, but neither model, either model’s weights, nor a serving implementation.
Substituting models can change capability, refusal behavior, jailbreak difficulty, structured output, image quality, latency, resource requirements, and game results. External inference and image APIs may charge usage fees; self-hosting can incur compute, storage, bandwidth, and operating costs.
Package scope
Application and tests, not model infrastructure
The package includes the FastAPI application, browser UI, procedural game data, tests, PostgreSQL leaderboard configuration, Docker setup, model contract, and optional art-service contract. It contains no weights, tokenizers, generated artwork, database state, caches, or deployment secrets. Artwork is disabled by default.
v0.14.0 repository snapshot
Files
- A-Prompt-Adventure-reference-0.14.0
- README.md
5.8 KB · MD