Codex Config

 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

#:schema https://developers.openai.com/codex/config-schema.json

# ================ Model, Reasoning & Provider ================
model = "gpt-5.5"
review_model = "gpt-5.5"
personality = "pragmatic"

model_reasoning_effort = "high"
plan_mode_reasoning_effort = "xhigh"
model_reasoning_summary = "detailed"
model_verbosity = "high"
model_supports_reasoning_summaries = true

# ================ Approval, Sandbox, Permissions & Project Trust ================
approval_policy = "on-request"
approvals_reviewer = "auto_review"

sandbox_mode = "workspace-write"
sandbox_workspace_write.network_access = true
sandbox_workspace_write.exclude_slash_tmp = false
sandbox_workspace_write.exclude_tmpdir_env_var = false

windows.sandbox = "elevated"
windows.sandbox_private_desktop = true

default_permissions = ":workspace"

# ================ Network & Web Search ================
web_search = "live"
tools.web_search.context_size = "high"

# ================ Shell, Execution Environment & Local Tools ================

# ================ Apps, Connectors, Plugins & Skills ================
features.apps = false
features.plugins = false

# ================ MCP Servers & OAuth ================
features.enable_mcp_apps = false

# ================ Hooks Lifecycle ================
features.hooks = false

# ================ Multi-Agent ================
features.multi_agent = true

# ================ Memories ================
features.memories = true

# ================ General Behavior, Project Docs & Feature Toggles ================
check_for_update_on_startup = false
features.codex_git_commit = true
features.js_repl = true
features.terminal_resize_reflow = true
features.mentions_v2 = true
features.undo = true
features.fast_mode = false
features.prevent_idle_sleep = true

# ================ Logs, History, Feedback & Telemetry ================
analytics.enabled = false
feedback.enabled = false
history.persistence = "save-all"

# ================ TUI, Notifications, Appearance & Prompt Status ================
tui.status_line = [
    "model-with-reasoning",
    "current-dir",
    "permissions",
    "context-used",
]
tui.status_line_use_colors = true
notice.hide_full_access_warning = false
notice.hide_world_writable_warning = false
notice.hide_rate_limit_model_nudge = false
show_raw_agent_reasoning = true
hide_agent_reasoning = false
tui.show_tooltips = false
tui.model_availability_nux."gpt-5.5" = 4

 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

#:schema https://developers.openai.com/codex/config-schema.json

# ================ Model, Reasoning & Provider ================
model_provider = "openrouter"
model_catalog_json = "~/.codex/model-catalog.openrouter.json"

model_providers."openrouter".name = "OpenRouter"
model_providers."openrouter".base_url = "https://openrouter.ai/api/v1"
model_providers."openrouter".env_key = "OPENROUTER_API_KEY"
model_providers."openrouter".wire_api = "responses"

personality = "pragmatic"
model_reasoning_effort = "xhigh"
plan_mode_reasoning_effort = "xhigh"
model_reasoning_summary = "detailed"
model_verbosity = "high"
model_supports_reasoning_summaries = true

# ================ Approval, Sandbox, Permissions & Project Trust ================
approval_policy = "on-request"
approvals_reviewer = "auto_review"

sandbox_mode = "workspace-write"
sandbox_workspace_write.network_access = true
sandbox_workspace_write.exclude_slash_tmp = false
sandbox_workspace_write.exclude_tmpdir_env_var = false

windows.sandbox = "elevated"
windows.sandbox_private_desktop = true

default_permissions = ":workspace"

# ================ Network & Web Search ================
web_search = "live"
tools.web_search.context_size = "high"

# ================ Shell, Execution Environment & Local Tools ================

# ================ Apps, Connectors, Plugins & Skills ================
features.apps = false
features.plugins = false

# ================ MCP Servers & OAuth ================
features.enable_mcp_apps = false

# ================ Hooks Lifecycle ================
features.hooks = false

# ================ Multi-Agent ================
features.multi_agent = true

# ================ Memories ================
features.memories = true

# ================ General Behavior, Project Docs & Feature Toggles ================
check_for_update_on_startup = false
features.codex_git_commit = true
features.js_repl = true
features.terminal_resize_reflow = true
features.mentions_v2 = true
features.undo = true
features.fast_mode = false
features.prevent_idle_sleep = true

# ================ Logs, History, Feedback & Telemetry ================
analytics.enabled = false
feedback.enabled = false
history.persistence = "save-all"

# ================ TUI, Notifications, Appearance & Prompt Status ================
tui.status_line = [
    "model-with-reasoning",
    "current-dir",
    "permissions",
    "context-used",
]
tui.status_line_use_colors = true
tui.show_tooltips = false
notice.hide_full_access_warning = false
notice.hide_world_writable_warning = false
notice.hide_rate_limit_model_nudge = false
show_raw_agent_reasoning = true
hide_agent_reasoning = false

  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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300

#!/usr/bin/env python3
"""
Convert OpenRouter /api/v1/models into a Codex model_catalog_json file.

Usage:
  python3 openrouter_to_codex_catalog.py \
    --output ~/.codex/model-catalog.openrouter.json

Then in ~/.codex/config.toml:
  model_catalog_json = "/Users/yourname/.codex/model-catalog.openrouter.json"
"""

from __future__ import annotations

import argparse
import json
import os
import sys
import urllib.parse
import urllib.request
from pathlib import Path
from typing import Any


OPENROUTER_MODELS_URL = "https://openrouter.ai/api/v1/models"

BASE_INSTRUCTIONS = (
    "You are Codex, a coding agent. Help the user modify, understand, test, "
    "and improve code in the current workspace."
)

REASONING_LEVELS = [
    {"effort": "low", "description": "Fast responses with lighter reasoning"},
    {"effort": "medium", "description": "Balanced reasoning"},
    {"effort": "high", "description": "Deeper reasoning"},
    {"effort": "xhigh", "description": "Maximum reasoning"},
]


def fetch_json(url: str) -> dict[str, Any]:
    headers = {
        "User-Agent": "openrouter-to-codex-catalog/1.0",
        "Accept": "application/json",
    }

    # OpenRouter's models endpoint is generally public, but this supports
    # private/provider-filtered results if OPENROUTER_API_KEY is present.
    api_key = os.environ.get("OPENROUTER_API_KEY")
    if api_key:
        headers["Authorization"] = f"Bearer {api_key}"

    req = urllib.request.Request(url, headers=headers)

    try:
        with urllib.request.urlopen(req, timeout=60) as resp:
            body = resp.read().decode("utf-8")
            return json.loads(body)
    except Exception as exc:
        raise SystemExit(f"Failed to fetch {url}: {exc}") from exc


def with_query(url: str, params: dict[str, str]) -> str:
    parsed = urllib.parse.urlparse(url)
    existing = dict(urllib.parse.parse_qsl(parsed.query, keep_blank_values=True))
    existing.update(params)
    query = urllib.parse.urlencode(existing)
    return urllib.parse.urlunparse(parsed._replace(query=query))


def clean_text(value: Any, default: str = "") -> str:
    if value is None:
        return default
    text = str(value).replace("\r\n", "\n").strip()
    return text if text else default


def int_or_none(value: Any) -> int | None:
    try:
        if value is None:
            return None
        return int(value)
    except (TypeError, ValueError):
        return None


def model_context_length(model: dict[str, Any]) -> int | None:
    top_provider = model.get("top_provider") or {}
    return (
        int_or_none(top_provider.get("context_length"))
        or int_or_none(model.get("context_length"))
    )


def codex_input_modalities(model: dict[str, Any]) -> list[str]:
    architecture = model.get("architecture") or {}
    raw = architecture.get("input_modalities") or ["text"]

    allowed = []
    for item in raw:
        if item == "text":
            allowed.append("text")
        elif item == "image":
            allowed.append("image")

    # Codex only has text/image input modality enum. If OpenRouter says only
    # file/audio/etc., keep text as a conservative fallback for chat usage.
    return allowed or ["text"]


def has_text_output(model: dict[str, Any]) -> bool:
    architecture = model.get("architecture") or {}
    outputs = architecture.get("output_modalities") or ["text"]
    return "text" in outputs


def supports_reasoning(model: dict[str, Any]) -> bool:
    params = set(model.get("supported_parameters") or [])
    return bool({"reasoning", "include_reasoning"} & params)


def supports_tools(model: dict[str, Any]) -> bool:
    params = set(model.get("supported_parameters") or [])
    return "tools" in params


def make_codex_model(
    model: dict[str, Any],
    priority: int,
    enable_parallel_tools: bool,
) -> dict[str, Any] | None:
    model_id = clean_text(model.get("id"))
    if not model_id:
        return None

    ctx = model_context_length(model)

    reasoning = supports_reasoning(model)
    tool_support = supports_tools(model)

    item: dict[str, Any] = {
        "slug": model_id,
        "display_name": clean_text(model.get("name"), model_id),
        "description": clean_text(model.get("description"), f"OpenRouter model: {model_id}"),

        # Reasoning metadata. Leave null/empty for models that do not advertise reasoning.
        "default_reasoning_level": "medium" if reasoning else None,
        "supported_reasoning_levels": REASONING_LEVELS if reasoning else [],

        # Codex agent/tool metadata.
        "shell_type": "shell_command",
        "visibility": "list",
        "supported_in_api": True,
        "priority": priority,

        # Newer Codex versions may expect these collection fields.
        "additional_speed_tiers": [],
        "service_tiers": [],
        "default_service_tier": None,

        "availability_nux": None,
        "upgrade": None,
        "base_instructions": BASE_INSTRUCTIONS,
        "model_messages": None,

        # Be conservative for non-OpenAI models unless OpenRouter exposes
        # exact compatibility. You can override these manually per model.
        "supports_reasoning_summaries": False,
        "default_reasoning_summary": "none",
        "support_verbosity": False,
        "default_verbosity": None,

        # Keep patch/shell tools available for Codex.
        "apply_patch_tool_type": "freeform",
        "web_search_tool_type": "text",
        "truncation_policy": {"mode": "tokens", "limit": 10000},

        # OpenRouter advertises "tools", but not every backend handles parallel
        # tool calls cleanly. Default is off unless --enable-parallel-tools is used.
        "supports_parallel_tool_calls": bool(tool_support and enable_parallel_tools),

        "supports_image_detail_original": "image" in codex_input_modalities(model),
        "context_window": ctx,
        "max_context_window": ctx,

        # Let Codex derive 90% from context_window when null.
        "auto_compact_token_limit": None,
        "effective_context_window_percent": 95,

        "experimental_supported_tools": [],
        "input_modalities": codex_input_modalities(model),
        "supports_search_tool": False,
        "tool_mode": None,
    }

    return item


def main() -> int:
    parser = argparse.ArgumentParser(
        description="Convert OpenRouter models to Codex model-catalog JSON."
    )
    parser.add_argument(
        "--url",
        default=OPENROUTER_MODELS_URL,
        help=f"OpenRouter models URL. Default: {OPENROUTER_MODELS_URL}",
    )
    parser.add_argument(
        "--output",
        default=str(Path.home() / ".codex" / "model-catalog.openrouter.json"),
        help="Output path. Default: ~/.codex/model-catalog.openrouter.json",
    )
    parser.add_argument(
        "--modalities",
        default="all",
        help='Value for output_modalities query parameter. Default: "all". '
             'Use "text" if you only want text-output models.',
    )
    parser.add_argument(
        "--include-non-text-output",
        action="store_true",
        help="Include image/audio/embedding-only models. Usually not useful for Codex.",
    )
    parser.add_argument(
        "--include-expired",
        action="store_true",
        help="Include models with expiration_date set.",
    )
    parser.add_argument(
        "--enable-parallel-tools",
        action="store_true",
        help="Set supports_parallel_tool_calls=true for models advertising tools.",
    )
    parser.add_argument(
        "--pretty",
        action="store_true",
        help="Pretty-print JSON with indentation.",
    )

    args = parser.parse_args()

    url = args.url
    if args.modalities:
        url = with_query(url, {"output_modalities": args.modalities})

    payload = fetch_json(url)
    models = payload.get("data")
    if not isinstance(models, list):
        raise SystemExit("Unexpected OpenRouter response: missing top-level data list")

    codex_models: list[dict[str, Any]] = []
    skipped_non_text = 0
    skipped_expired = 0

    for raw in models:
        if not isinstance(raw, dict):
            continue

        if raw.get("expiration_date") and not args.include_expired:
            skipped_expired += 1
            continue

        if not args.include_non_text_output and not has_text_output(raw):
            skipped_non_text += 1
            continue

        converted = make_codex_model(
            raw,
            priority=len(codex_models),
            enable_parallel_tools=args.enable_parallel_tools,
        )
        if converted:
            codex_models.append(converted)

    catalog = {"models": codex_models}

    output = Path(args.output).expanduser()
    output.parent.mkdir(parents=True, exist_ok=True)

    with output.open("w", encoding="utf-8") as f:
        if args.pretty:
            json.dump(catalog, f, ensure_ascii=False, indent=2)
            f.write("\n")
        else:
            json.dump(catalog, f, ensure_ascii=False, separators=(",", ":"))
            f.write("\n")

    print(f"Wrote {len(codex_models)} models to {output}")
    if skipped_expired:
        print(f"Skipped expired/deprecated models: {skipped_expired}")
    if skipped_non_text:
        print(f"Skipped non-text-output models: {skipped_non_text}")
    print()
    print("Add this to ~/.codex/config.toml:")
    print(f'model_catalog_json = "{output}"')

    return 0


if __name__ == "__main__":
    raise SystemExit(main())