harn-vm 0.8.92

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
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74

# ---------- DeepSeek V4 direct (OpenAI-compatible /v1) ------------------------
# V4-Flash + V4-Pro support native tool calls, thinking mode, prompt
# caching, and JSON output. `deepseek-chat`/`deepseek-reasoner` are
# deprecated aliases mapping to V4-Flash non-thinking/thinking modes.

[[provider.deepseek]]
model_match = "deepseek-v4-pro*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
thinking_modes = ["enabled"]
text_tool_wire_format_supported = true
prefers_xml_scaffolding = false
prefers_markdown_scaffolding = true
structured_output_mode = "native_json"
supports_assistant_prefill = false
prefers_role_developer = false
prefers_xml_tools = false
thinking_block_style = "inline"
prompt_caching = true

[[provider.deepseek]]
model_match = "deepseek-v4-flash*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
thinking_modes = ["enabled"]
text_tool_wire_format_supported = true
prefers_xml_scaffolding = false
prefers_markdown_scaffolding = true
structured_output_mode = "native_json"
supports_assistant_prefill = false
prefers_role_developer = false
prefers_xml_tools = false
thinking_block_style = "inline"
prompt_caching = true

[[provider.deepseek]]
model_match = "deepseek-chat*"
native_tools = true
# This row pins the NATIVE channel. `json` (fenced-JSON) is now the GLOBAL
# text-channel default, so no per-model json pin is needed to get json; this
# `native` pin is only kept to select the native channel. If this model is run
# on text instead, fenced-json is the default (json swept a clean 1.0/1.0/1.0
# emission-fidelity bench on the deepseek family). To force text formats:
# `preferred_tool_format = "json"` (fenced) or `"text"` (heredoc).
preferred_tool_format = "native"
structured_output = "native"
text_tool_wire_format_supported = true
prefers_xml_scaffolding = false
prefers_markdown_scaffolding = true
structured_output_mode = "native_json"
supports_assistant_prefill = false
prefers_role_developer = false
prefers_xml_tools = false
thinking_block_style = "none"
prompt_caching = true

[[provider.deepseek]]
model_match = "deepseek-reasoner*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
thinking_modes = ["enabled"]
text_tool_wire_format_supported = true
prefers_xml_scaffolding = false
prefers_markdown_scaffolding = true
structured_output_mode = "native_json"
supports_assistant_prefill = false
prefers_role_developer = false
prefers_xml_tools = false
thinking_block_style = "inline"
prompt_caching = true