roadmap.md

V0 Roadmap - Fall'23

• Draft implementation with CGO llama.cpp backend
• Simple REST API to allow text generation
• Inference with Apple Silicon GPU using Metal framework
• Parallel inference both with CPU and GPU
• Support both AMD64 and ARM64 platforms
• CUDA support and fast inference with Nvidia cards
• Retain dialog history by Session ID parameter
• Support moderm GGUF V3 model format
• Inference for most popular LLM architectures
• Janus Sampling for better non-English text generation

V1 Roadmap - Winter'23

• Rebrand project: LLaMAZoo => Large Model Collider
• Is it 2023, 30th of November? First birthday of ChatGPT! Celebrate ...
• ... then release Collider V1 after half a year of honing it :)

V2 Roadmap - Spring'24

• Full LLaMA v2 support
• Freeze JSON / YAML config format for Native API

V3 Roadmap - Summer'24

• Rebrand project again :) Collider => Booster
• Complete LLaMA v3 support
• OpenAI API Chat Completion compatible endpoints
• Ollama compatible endpoints
• Interactive mode for chatting from command line
• Update Janus Sampling for LLaMA-3
• Broader integration with Ollama ecosystem
• Smarter context shrinking when reaching its limits chatting with model
• Embedded web UI with no external dependencies
• Allow native Windows support
• Prebuilt binaries for all platforms
• Support LLaVA multi-modal models inference
• Better code test coverage
• Perplexity computation useful for benchmarking