harn-vm 0.8.79

Async bytecode virtual machine for the Harn programming language
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45

# Local runtime lifecycle metadata for `harn local`.
# These rows describe provider mechanics, not machine-specific model paths.

[providers.ollama.local_runtime]
kind = "daemon_api"
command = "ollama"
default_port = 11434
stop = "keep_alive_zero"
source_url = "https://github.com/ollama/ollama/blob/main/docs/api.md"
last_verified = "2026-06-05"
notes = "Load via Ollama generate/chat warmup; unload by posting an empty prompt with keep_alive=0."

[providers.llamacpp.local_runtime]
kind = "managed_process"
command = "llama-server"
model_source_env = "LLAMACPP_MODEL"
default_port = 8001
model_arg = "--model"
served_model_arg = "--alias"
host_arg = "--host"
port_arg = "--port"
ctx_arg = "--ctx-size"
parallel_arg = "--parallel"
gpu_layers_arg = "--n-gpu-layers"
cache_type_k_arg = "--cache-type-k"
cache_type_v_arg = "--cache-type-v"
cache_ram_arg = "--cache-ram"
default_args = ["--jinja", "--reasoning", "off", "--reasoning-format", "deepseek", "--metrics", "--flash-attn", "on"]
stop = "pid"
source_url = "https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md"
last_verified = "2026-06-05"
notes = "OpenAI-compatible HTTP server. Use --model-source or LLAMACPP_MODEL for the GGUF path; Harn records the launched PID for local stop."

[providers.mlx.local_runtime]
kind = "managed_process"
command = "mlx_lm.server"
model_source_env = "MLX_MODEL"
default_port = 8002
model_arg = "--model"
host_arg = "--host"
port_arg = "--port"
stop = "pid"
source_url = "https://github.com/ml-explore/mlx-lm/blob/main/mlx_lm/SERVER.md"
last_verified = "2026-06-05"
notes = "OpenAI-like MLX-LM server. Use --model-source or MLX_MODEL for an MLX-compatible path or Hugging Face repo id."