harn-vm 0.10.34

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
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176

# OpenAI-compatible providers whose per-model catalogs are refreshed live when
# callers provide API keys. These wildcard rules are conservative transport
# defaults: model-family overrides can still be added as empirical probes find
# provider-specific reasoning/tool quirks.

[[provider.nebius]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Nebius documents function calling and structured JSON output; Harn has not yet run model-family probes for every Token Factory route."

[[provider.flexai]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "FlexAI exposes model-level support flags through discovery; use live refresh/probes to tighten route-specific tool preferences."

[[provider.hyperbolic]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Hyperbolic documents tool calling for selected Qwen/Llama routes; model-specific probes should override this wildcard if needed."

[[provider.siliconflow]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "SiliconFlow is OpenAI-compatible, but thinking controls are provider-specific on some model families; prefer live model probes before pinning reasoning-heavy agents."

[[provider.parasail]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Parasail exposes an OpenAI-compatible serverless/batch API; use model-specific probes before relying on native tools for long agent loops."

[[provider.atlas]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Atlas is a coding-plan OpenAI-compatible route; Harn has no route-specific tool parity probes yet."

[[provider.github_models]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "GitHub Models publishes capability flags per catalog model; Harn needs route-specific probes before assuming native/text tool parity."

[[provider.vercel_ai_gateway]]
model_match = "google/gemini-3.1*"
extends = true
native_tools = true
preferred_tool_format = "native"
thinking_modes = ["enabled", "adaptive"]
prompt_caching = true
cache_breakpoint_style = "last_block"
vision_supported = true
vision = true
audio_supported = true
pdf_supported = true
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
prefers_xml_scaffolding = false
prefers_markdown_scaffolding = true
supports_assistant_prefill = false
prefers_role_developer = false
prefers_xml_tools = false
thinking_block_style = "reasoning_summary"

[[provider.vercel_ai_gateway]]
model_match = "*"
extends = true
responses_api = true

[[provider.cloudflare_ai_gateway]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Cloudflare exposes OpenAI-compatible Workers AI and AI Gateway routes, but capabilities depend on the routed model."

[[provider.inception]]
model_match = "mercury-2"
native_tools = true
preferred_tool_format = "native"
thinking_modes = ["effort"]
reasoning_effort_supported = true
reasoning_effort_levels = ["low", "medium", "high"]
structured_output = "native"
structured_output_mode = "native_json"
thinking_block_style = "reasoning_summary"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Inception documents OpenAI-compatible tool use for Mercury 2; no Harn parity probe has run yet."

[[provider.inception]]
model_match = "mercury-coder-*"
native_tools = true
preferred_tool_format = "native"
structured_output = "delimited"
structured_output_mode = "delimited"
thinking_block_style = "none"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Mercury Coder Small supports tool use on the OpenAI-compatible API; structured object generation was not documented as GA in the source pass."

[[provider.friendli]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Friendli serverless endpoints are OpenAI-compatible; model-specific capability probes should narrow this wildcard."

[[provider.volcengine_ark]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Volcengine Ark exposes OpenAI-compatible chat completions; routed model families still need native-vs-text tool probes."

[[provider.hunyuan]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Tencent Hunyuan documents OpenAI-compatible function calling; Harn has not yet probed every Hunyuan model family."

[[provider.qianfan]]
model_match = "*"
native_tools = true
preferred_tool_format = "native"
structured_output = "native"
structured_output_mode = "native_json"
text_tool_wire_format_supported = true
tool_mode_parity = "unknown"
tool_mode_parity_notes = "Baidu Qianfan v2 documents OpenAI-compatible tool calls; model-specific parity remains unprobed."