codex cli 频繁出现<!-- -->是什么问题啊

c c 2026-07-09 22:16 1

各位大佬,我的codex cli 频繁出现<!-- -->,有遇见过并解决了的么?用的bash脚本方式安装的最新版codex cli


我的配置:


#:schema https://developers.openai.com/codex/config-schema.json
model_provider = "cliproxyapi"
disable_response_storage = true
model = "gpt-5.5"
review_model = "gpt-5.5"
model_reasoning_effort = "xhigh"
plan_mode_reasoning_effort = "xhigh"
model_reasoning_summary = "detailed"
model_supports_reasoning_summaries = true
model_verbosity = "high"
#model_auto_compact_token_limit = 260000
#model_context_window = 270000
tool_output_token_limit = 40000
approval_policy = "on-request"
sandbox_mode = "workspace-write"
web_search = "live"
personality = "pragmatic"
approvals_reviewer = "auto_review"
#approvals_reviewer = "user"
windows_wsl_setup_acknowledged = true
suppress_unstable_features_warning = true
commit_attribution = "Co-authored-by: [email protected]"
project_doc_fallback_filenames = [ "CLAUDE.md", "README.md", "CONTRIBUTING.md" ]
mcp_oauth_credentials_store = "file"
sqlite_home = "~/.codex/sqlite/"
persistent_instructions = "Follow AGENTS.md guidelines. Use available MCP servers and skills when they can help."
service_tier = "default"

[model_providers]
[model_providers.cliproxyapi]
name = "OpenAI"
base_url = "https://xx.xx.xx.xx/v1"
wire_api = "responses"
#supports_websockets = true
requires_openai_auth = true
experimental_bearer_token = "APIKEY"

[mcp_servers.sequential-thinking]
type = "stdio"
startup_timeout_sec = 60
command = "npx"
args = ["-y", "@modelcontextprotocol/server-sequential-thinking"]
default_tools_approval_mode = "approve"

[mcp_servers.chrome-devtools]
type = "stdio"
startup_timeout_sec = 60
command = "npx"
args = ["chrome-devtools-mcp@latest", "--browser-url=http://127.0.0.1:9222"]
default_tools_approval_mode = "approve"
# 禁用chrome-devtools
enabled = false

[mcp_servers.context7]
type = "http"
url = "https://mcp.context7.com/mcp"
http_headers = { CONTEXT7_API_KEY = "APIKEY", Content-Type = "application/json" }
tool_timeout_sec = 30
default_tools_approval_mode = "approve"

[mcp_servers.exa]
type = "http"
url = "https://mcp.exa.ai/mcp?tools=web_search_exa,web_fetch_exa,web_search_advanced_exa"
http_headers = { x-api-key = "APIKEY" }
default_tools_approval_mode = "approve"

[analytics]
enabled = false

[feedback]
enabled = false

[audio]
microphone = "default"
speaker = "default"

[history]
max_bytes = 1000000000

[features]
default_mode_request_user_input = true
memories = true
enable_fanout = true
realtime_conversation = true
exec_permission_approvals = true
request_permissions_tool = true
multi_agent_v2 = true
respect_system_proxy = true
image_generation = false
concurrent_reasoning_summaries = true
terminal_visualization_instructions = true

[features.current_time_reminder]
enabled = true
sleep_tool = true

[sandbox_workspace_write]
network_access = true

[memories]
consolidation_model = "gpt-5.5"
extract_model = "gpt-5.5"
generate_memories = true
use_memories = true
disable_on_external_context = true
max_raw_memories_for_consolidation = 512
max_rollout_age_days = 30
max_rollouts_per_startup = 20
max_unused_days = 180
min_rollout_idle_hours = 12

[tools.web_search]
context_size = "medium"

[agents]
max_depth = 1
job_max_runtime_seconds = 900

[agents.explorer]
description = "Read-only codebase explorer for gathering evidence before changes are proposed."
config_file = "agents/explorer.toml"

[agents.reviewer]
description = "PR reviewer focused on correctness, security, and missing tests."
config_file = "agents/reviewer.toml"

[agents.docs_researcher]
description = "Documentation specialist that verifies APIs, framework behavior, and release notes."
config_file = "agents/docs-researcher.toml"

[tui]
status_line = ["model-with-reasoning", "current-dir", "git-branch", "context-used", "context-window-size", "fast-mode"]
status_line_use_colors = true

[tui.model_availability_nux]
"gpt-5.5" = 4

[notice]
fast_default_opt_out = true

[marketplaces.ponytail]
last_updated = "2026-07-09T10:53:09Z"
last_revision = "8f32ae0f6eb0fe73189aa505e00da170e8d1f2fc"
source_type = "git"
source = "https://github.com/DietrichGebert/ponytail.git"

[hooks.state]

[hooks.state."ponytail@ponytail:hooks/claude-codex-hooks.json:session_start:0:0"]
trusted_hash = "sha256:10e0ac0497512b279dc2b073d0d559fb9fb252bc18b9c943d2f0a6d58152b1f4"

[hooks.state."ponytail@ponytail:hooks/claude-codex-hooks.json:user_prompt_submit:0:0"]
trusted_hash = "sha256:2a3c9015a1fe10f0809ffec8da3d67c326e27c38ebf720725099283b1e2085a8"

[hooks.state."ponytail@ponytail:hooks/claude-codex-hooks.json:subagent_start:0:0"]
trusted_hash = "sha256:4765354e31cc783c9cc7379f8beac7ac7ff68743c3cd7dad333d1f6a64e7a6f3"

[plugins."ponytail@ponytail"]
enabled = true
最新回复 (19)
  • amdyes 07-09 22:17
    1

    看到过,似乎是思考过程出现的,不过用的是opencode

  • nianzhibai 07-09 22:19
    2

    回退上一版本0.142.5

    0.143.0版本bug太多

  • 泰迪 07-09 22:20
    3

    灰度到5.6了把这是。听说5.6就会有这个问题

  • inndeex 07-09 22:20
    4

    哪都有 我在爱马仕/小龙虾/CC 里都见到了,感觉至少不只是cli的问题,模型端感觉问题大些

  • pluto233 07-09 22:22
    5

    w? 据说是灰度到新模型导致的bug喵?~(呆


    w 你怎么发了两个帖子喵~(呆

  • kalikali 07-09 22:22
    6

    ooenai负责总结思考的那个小模型罢工了,应该不是灰度到了5.6,估计是新模型快发布了,在维护吧。

  • peng chuan 07-09 22:23
    7

    我上午也有,晚上重新build最新的版本的codex就没有了

  • pzyyl 07-09 22:24
    8

    和客户端没有关系,我在 pi 中使用也是这样,可能真的和准备发布 gpt-5.6 有关系把…

  • Waterfox 07-09 22:25
    9

    我都快看吐了,原来不只有我一个人有问题


    肯定100%不是5.6,压根不是那智商

  • Dark_iaji 07-09 22:25
    10

    那我这是用上真5.5了?

    反正中转站就这样。

  • 无敌暴龙战士 07-09 22:26
    11

    和使用工具没有关系,好像是模型的问题。

  • 董炸淘 07-09 22:28
    12

    我用pi也这样,一大早就这样了,只有思考的标题,没有思考内容

  • 岛歌 07-09 22:30
    13



    没想到还是个傲娇,颜文字 瀑布汗,有点文艺复兴了哦 ^-^

  • 还想成为88VIP 07-09 22:31
    14

    大家发的都是cli的图,可能是新版本cli的问题?我用桌面端没见过这东西

  • c c 楼主 07-09 23:18
    15

    是思考过程,思考总结没了,就给个这。。

  • Waterfox 07-10 09:38
    16



    彻底疯狂

    怎么5.6还是这样

    这个也太恶心了

  • Waterfox 07-10 09:47
    17

    cpa+newapi

    也不知道哪出了问题

    codex cli和opencode都会这样

  • c c 楼主 07-10 09:54
    18

    我试过不用中专也是这样的问题。。。要么模型问题,要么codex cli问题。

  • ligen131 07-10 13:01
    19

    今天 gpt-5.6 发布了,试了下只有 5.6 系列模型会出现这种情况,而且是稳定复现。昨天 5.6 还没发布的时候部分人 5.5 也出现了这种现象,估计是被灰度到了 5.6 模型。



* 帖子来源Linux.do
返回