How pfnet/plamo-translate-cli works
plamo-translate-cli accepts text through its CLI or MCP interface, prepares a translation request, and dispatches it to an MLX or llama.cpp server for local model inference. The MLX path supports PLaMo 2 and PLaMo 3 checkpoint loading, prompting, and streaming; llama.cpp resolves GGUF models and serves translations through its local process. The README also documents MCP-client use and model selection. Arrows are limited to relationships shown in excerpts, the source index, or README; some backend internals are represented by their exact source paths because their full implementations were not sampled.
Main entry points: main.py CLI entrypoint · mlx/server.py MLX backend · llama_cpp/server.py llama.cpp backend · loader.py MLX checkpoint loader
Components 20
CLI and MCP
- CLI entrypoint
src/plamo_translate/main.py - MCP client
src/plamo_translate/clients/translate.py
Inference servers
- MCP translation server
src/plamo_translate/servers/mlx/server.py - MLX backend
src/plamo_translate/servers/mlx/server.py - llama.cpp backend
src/plamo_translate/servers/llama_cpp/server.py - Mock backend
src/plamo_translate/servers/mock/server.py - Server configuration
src/plamo_translate/servers/utils.py - Optional dependency warnings
src/plamo_translate/servers/warnings.py
MLX models
- MLX checkpoint loader
src/plamo_translate/servers/mlx/loader.py - PLaMo 2 model setup
src/plamo_translate/servers/mlx/model.py - PLaMo 3 prompts
src/plamo_translate/servers/mlx/pfml3.py - PLaMo 3 runtime
src/plamo_translate/servers/mlx/plamo3_runtime.py - PLaMo 3 architecture
src/plamo_translate/servers/mlx/plamo3.py - Metal acceleration
src/plamo_translate/servers/mlx/plamo3_metal.py
Model assets
- MLX release assets
src/plamo_translate/servers/mlx/release.py - GGUF model resolver
src/plamo_translate/servers/llama_cpp/model.py
Other components
- CLI user
- Hugging Face Hub
- llama-server process
- MCP client application
Connections 21
- CLI user → CLI entrypoint: submits text
README.md - CLI entrypoint → MCP client: invokes
src/plamo_translate/main.py - CLI entrypoint → MLX backend: starts backend
src/plamo_translate/main.py - CLI entrypoint → llama.cpp backend: starts backend
src/plamo_translate/main.py - CLI entrypoint → Mock backend: starts backend
src/plamo_translate/main.py - MCP client → MCP translation server: calls MCP tool
src/plamo_translate/clients/translate.py - MCP translation server → MCP client application: serves MCP
README.md - MLX backend → MLX checkpoint loader: loads model
src/plamo_translate/servers/mlx/server.py - MLX backend → PLaMo 3 prompts: builds prompts
src/plamo_translate/servers/mlx/server.py - MLX backend → PLaMo 3 runtime: generates with
src/plamo_translate/servers/mlx/server.py - MLX checkpoint loader → PLaMo 2 model setup: configures model
src/plamo_translate/servers/mlx/loader.py - MLX checkpoint loader → Optional dependency warnings: suppresses warnings
src/plamo_translate/servers/mlx/loader.py - PLaMo 3 runtime → PLaMo 3 architecture: runs model
src/plamo_translate/servers/mlx/plamo3_runtime.py - PLaMo 3 runtime → Metal acceleration: uses kernels
src/plamo_translate/servers/mlx/plamo3_runtime.py - llama.cpp backend → GGUF model resolver: resolves model
src/plamo_translate/servers/llama_cpp/server.py - llama.cpp backend → llama-server process: starts and queries
src/plamo_translate/servers/llama_cpp/server.py - MLX backend → Server configuration: uses configuration
src/plamo_translate/servers/mlx/server.py - llama.cpp backend → Server configuration: uses configuration
src/plamo_translate/servers/llama_cpp/server.py - MLX checkpoint loader → Hugging Face Hub: downloads checkpoints
README.md - GGUF model resolver → Hugging Face Hub: downloads GGUF
README.md - MLX release assets → Hugging Face Hub: packages checkpoints