MiOS/usr/share/mios/ai/v1/models.json at main · mios-dev/MiOS · GitHub

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
{
  "object": "list",
  "data": [
    {
      "id": "granite4.1:8b",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-llm-light",
      "permission": [],
      "root": "granite4.1:8b",
      "parent": null,
      "x-mios": {
        "role": "default-chat",
        "family": "IBM Granite (dense 'granite' arch; loads on mainline llama.cpp)",
        "context_window": 131072,
        "ssot": "[ai].model in /usr/share/mios/mios.toml",
        "served_by": "mios-llm-light.service (primary lane; llama.cpp behind the upstream llama-swap proxy)",
        "endpoint": "http://localhost:11450/v1",
        "openai_api_compat": ["chat/completions", "completions"]
      }
    },
    {
      "id": "lfm2:700m",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-llm-light",
      "permission": [],
      "root": "lfm2:700m",
      "parent": null,
      "x-mios": {
        "role": "micro / classify-expand-gate",
        "family": "Liquid AI LFM2 (dense 'lfm2' arch; merged mainline llama.cpp)",
        "context_window": 32768,
        "ssot": "[ai].micro_model in /usr/share/mios/mios.toml",
        "served_by": "mios-llm-light.service (always-warm micro lane); also the mios-cpu-node CPU fallback (:11451)",
        "endpoint": "http://localhost:11450/v1",
        "openai_api_compat": ["chat/completions"]
      }
    },
    {
      "id": "nomic-embed-text",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-llm-light",
      "permission": [],
      "root": "nomic-embed-text",
      "parent": null,
      "x-mios": {
        "role": "default-embed",
        "family": "Google EmbeddingGemma-300m QAT (served NAME kept 'nomic-embed-text'; 768d native = zero pgvector migration)",
        "dimensions": 768,
        "ssot": "[ai].embed_model in /usr/share/mios/mios.toml",
        "served_by": "mios-llm-light.service (llama-server --embedding)",
        "endpoint": "http://localhost:11450/v1",
        "openai_api_compat": ["embeddings"]
      }
    },
    {
      "id": "mios-heavy",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-llm-heavy",
      "permission": [],
      "root": "mios-heavy",
      "parent": null,
      "x-mios": {
        "role": "heavy-reasoner (gated / off-by-default on VRAM)",
        "family": "Mistral Magistral Small 2509 AWQ (intended); incumbent Qwen2.5-14B-AWQ is the rollback",
        "context_window": 131072,
        "ssot": "[ai.sglang]/[ai.vllm] in /usr/share/mios/mios.toml",
        "served_by": "mios-llm-heavy.service (vLLM :11441)",
        "endpoint": "http://localhost:11441/v1",
        "openai_api_compat": ["chat/completions"]
      }
    },
    {
      "id": "hermes-agent",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-agents",
      "permission": [],
      "root": "hermes-agent",
      "parent": null,
      "x-mios": {
        "role": "general-orchestration-sub-agent",
        "served_by": "mios-gateway-agent.service",
        "endpoint": "http://localhost:8642/v1",
        "registered_as_sub_agent_of": "mios-agent-pipe",
        "registry": "/usr/share/mios/ai/v1/agents.json"
      }
    },
    {
      "id": "mios-opencode:latest",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-agents",
      "permission": [],
      "root": "mios-opencode:latest",
      "parent": null,
      "x-mios": {
        "role": "code-specialist-sub-agent (FROM granite4.1:8b)",
        "transport": "OpenAI /v1 (mios-opencode-gateway.service :8633)",
        "endpoint": "http://localhost:8633/v1",
        "registered_as_sub_agent_of": "mios-agent-pipe",
        "registry": "/usr/share/mios/ai/v1/agents.json"
      }
    },
    {
      "id": "mios-daemon-agent",
      "object": "model",
      "created": 1747000000,
      "owned_by": "mios-agents",
      "permission": [],
      "root": "mios-daemon-agent",
      "parent": null,
      "x-mios": {
        "role": "log-and-journal-follow-up-sub-agent",
        "served_by": "mios-daemon.service",
        "cpu_pinned_cores": 2,
        "registered_as_sub_agent_of": "mios-agent-pipe",
        "registry": "/usr/share/mios/ai/v1/agents.json",
        "produces": [
          "/var/lib/mios/daemon/launch_failures.json",
          "/var/lib/mios/scratch/agent-nudges.md",
          "satisfaction verdicts (event.type=user_query_satisfied)"
        ]
      }
    }
  ],
  "x-mios": {
    "schema": "OpenAI /v1/models response shape",
    "schema_ref": "https://platform.openai.com/docs/api-reference/models/list",
    "manifest_path": "/usr/share/mios/ai/v1/models.json",
    "endpoint": "http://localhost:8640/v1/models",
    "served_by": "mios-agent-pipe.service",
    "fleet_note": "2026-06-13 family-diverse fleet: granite4.1:8b (IBM) brain+coder, lfm2:700m (Liquid AI) micro, nomic-embed-text NAME = EmbeddingGemma-300m QAT (Google) embeddings, mios-heavy = Mistral Magistral Small 2509 (gated). All everyday inference is served by mios-llm-light.service (:11450, llama.cpp behind the upstream llama-swap proxy); 128k ctx on every chat/reasoning lane.",
    "note": "agent-pipe (:8640) is the canonical /v1 entry. It refines the user prompt, routes to a sub-agent (default: hermes-agent), wraps the sub-agent's raw output in <details type=\"reasoning\">, and polishes the final reply. Strict OpenAI clients should ignore x-mios extensions."
  }
}