ai

• chat tool

• chatgpt

  • model

  • API

    • get authorization key

    • model

    • model_limits

    • rate_limits

  • Administration API

    • admin key

    • organization id

    • group id

  • using chatgpt to generate git commits

  • using chatgpt to review git diff

• RAG

• tips

  • chatgpt

[!NOTE|label:references:]

• Introduction to Generative AI

chat tool

[!NOTE|label:references:]

• 如何在 Chatbox 中接入 SiliconCloud - 超简单完整教程

chatgpt

model

[!NOTE|label:references:]

• token limits

• models

ABBREVIATION	DESCRIPTION
RPM	Requests Per Minute
TPM	Tokens Per Minute
IPM	Images Per Minute
AMPM	Audio Megabytes Per Minute
RPD	Requests Per Day
TPD	Tokens Per Day
-	Batch Tokens Per Day

MODEL NAME	MAX CONTEXT	MAX OUTPUT TOKENS
gpt-3.5-turbo	16,385	4,096
gpt-3.5-turbo-16k	16,385	-
gpt-4	8,192	8,192
gpt-4-32k	32,768	-
gpt-4-turbo	128,000	4,096
gpt-4-32k	32,768	-
gpt-4-32k-0613	32,768	-
gpt-4 turbo	128,000	4,096
gpt-4o	128,000	16,384
gpt-4o-mini	128,000	100,000
gpt-4.1	1,047,576	32,768
gpt-o3	200,000	100,000
gpt-4.1 nano	1,047,576	32,768

MODEL	TOKEN LIMITS	REQUEST AND OTHER LIMITS	BATCH QUEUE LIMITS
gpt-3.5-turbo	200,000 TPM	500 RPM 10,000 RPD	2,000,000 TPD
gpt-3.5-turbo-0125	200,000 TPM	500 RPM 10,000 RPD	2,000,000 TPD
gpt-3.5-turbo-1106	200,000 TPM	500 RPM 10,000 RPD	2,000,000 TPD
gpt-3.5-turbo-16k	200,000 TPM	500 RPM 10,000 RPD	2,000,000 TPD
gpt-3.5-turbo-instruct	90,000 TPM	3,500 RPM	200,000 TPD
gpt-3.5-turbo-instruct-0914	90,000 TPM	3,500 RPM	200,000 TPD
gpt-4	10,000 TPM	500 RPM 10,000 RPD	100,000 TPD
gpt-4-0613	10,000 TPM	500 RPM 10,000 RPD	100,000 TPD
gpt-4-turbo ↪︎ gpt-4-turbo-2024-04-09 ↪︎ gpt-4-turbo-preview ↪︎ gpt-4-0125-preview ↪︎ gpt-4-1106-preview	30,000 TPM	500 RPM	90,000 TPD
gpt-4.1 ↪︎ gpt-4.1-2025-04-14	30,000 TPM	500 RPM	900,000 TPD
gpt-4.1 (long context)	200,000 TPM	100 RPM	2,000,000 TPD
gpt-4.1-mini ↪︎ gpt-4.1-mini-2025-04-14	200,000 TPM	500 RPM	2,000,000 TPD
gpt-4.1-mini (long context)	400,000 TPM	200 RPM	4,000,000 TPD
gpt-4.1-nano ↪︎ gpt-4.1-nano-2025-04-14	200,000 TPM	500 RPM	2,000,000 TPD
gpt-4.1-nano (long context)	400,000 TPM	200 RPM	4,000,000 TPD
gpt-4.5-preview ↪︎ gpt-4.5-preview-2025-02-27	125,000 TPM	1,000 RPM	50,000 TPD
gpt-4o ↪︎ gpt-4o-2024-05-13 ↪︎ gpt-4o-2024-08-06 ↪︎ gpt-4o-2024-11-20 ↪︎ gpt-4o-audio-preview ↪︎ gpt-4o-audio-preview-2024-10-01 ↪︎ gpt-4o-audio-preview-2024-12-17	30,000 TPM	500 RPM	90,000 TPD
gpt-4o-mini ↪︎ gpt-4o-mini-2024-07-18 ↪︎ gpt-4o-mini-audio-preview ↪︎ gpt-4o-mini-audio-preview-2024-12-17	200,000 TPM	500 RPM 10,000 RPD	2,000,000 TPD

API

[!NOTE|label:references:]

• Data controls in the OpenAI platform

• Models

• Error codes

• api type

  [!TIP|label:curl commands:]

  $ curl -s '${scheme}://${authority}${path}' \
         -H 'Authorization: Bearer ${Authorization}' \
         -H 'User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64)' \
         -H 'Origin: https://platform.openai.com' \
         -H 'Referer: https://platform.openai.com/settings/organization/limits' \
         -H 'Accept: application/json' \
         -H 'Content-Type: application/json'

TYPE	EXAMPLE	PUBLICLY DOCUMENTED	STABLE	RECOMMENDED FOR AUTOMATION
Public API	/v1/chat/completions	✅ Yes	✅ Stable	✅ Yes
Private API	/dashboard/rate_limits	❌ No	⚠️ May change	⚠️ For personal/internal use only
Internal API	/api/internal/user/settings	❌ No	❌ Unstable	❌ Not recommended
Browser-bound API	Requires session + Origin + Referer headers	❌ No	⚠️ Risky	❌ Manual use only

get authorization key

1. visit https://platform.openai.com/settings/organization/limits

2. ⌥ + ⌘ + i to open the devtools

3. open the tab Network → Fetch/XHR, and search with keyword limit

4. check the item one by one, untile find the one with Authorization: Bearer sess-**************************************** header in the Request Headers

5. double confirm the Response tab has the model limits data

chrom devtool

as long as found the correct item, check the Request Headers tab, it will be ( example ):

KEY	VALUE
:authority	api.openai.com
:path	/dashboard/rate_limits
:method	GET
:scheme	https
Authorization	Bearer sess-xxxxxx

[!TIP|label:other APIs]

AUTHORITY	Endpoint	SCHEME	METHOD
api.openai.com	/dashboard/billing/subscription	https	GET
api.openai.com	/dashboard/rate_limits	https	GET
api.openai.com	/dashboard/user/api_keys	https	GET
api.openai.com	/dashboard/feature_flags	https	GET
api.openai.com	/v1/dashboard/organizations/<ORGANIZATION_ID>/projects/<PROJECT_ID>/users	https	GET

• curl API call example:

$ curl -s '${scheme}://${authority}${path}' \
       -H 'Authorization: Bearer ${Authorization}' \
       -H 'User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64)' \
       -H 'Origin: https://platform.openai.com' \
       -H 'Referer: https://platform.openai.com/settings/organization/limits' \
       -H 'Accept: application/json' \
       -H 'Content-Type: application/json'

# i.e.:
$ curl -s 'https://api.openai.com/dashboard/rate_limits' \
       -H 'Authorization: Bearer sess-xxxxxx' \
       -H 'User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64)' \
       -H 'Origin: https://platform.openai.com' \
       -H 'Referer: https://platform.openai.com/settings/organization/limits' \
       -H 'Accept: application/json' \
       -H 'Content-Type: application/json'

model

$ curl -s https://api.openai.com/v1/models \
       -H "Authorization: Bearer $OPENAI_API_KEY"

# or show in table format
$ curl -s https://api.openai.com/v1/models \
       -H "Authorization: Bearer $OPENAI_API_KEY" |
  jq -r '["CREATED AT", "MODEL ID"],
         (.data | sort_by(.created) | reverse[] | [(.created | strftime("%Y-%m-%d %H:%M:%S")), .id])
         | @tsv' |
  column -t -s$'\t'
CREATED AT           MODEL ID
2025-06-02 23:54:58  gpt-4o-audio-preview-2025-06-03
2025-06-02 23:43:58  gpt-4o-realtime-preview-2025-06-03
2025-05-08 03:00:57  codex-mini-latest
2025-04-24 17:50:30  gpt-image-1
...

model_limits

$ curl -s https://api.openai.com/v1/fine_tuning/model_limits \
       -H "Authorization: Bearer ${OPENAI_API_KEY}"

# with table format
$ curl -s https://api.openai.com/v1/fine_tuning/model_limits \
       -H "Authorization: Bearer $OPENAI_API_KEY" |
  jq -r '["ID", "PENDING JOBS", "MAX PENDING JOBS", "JOBS", "MAX JOBS/DAY"],
         (.data[] | [.id, .curr_pending_jobs, .max_pending_jobs, .jobs_so_far_today, .max_jobs_per_day])
         | @tsv' |
  column -t -s$'\t'
ID                       PENDING JOBS  MAX PENDING JOBS  JOBS  MAX JOBS/DAY
gpt-3.5-turbo-0125       0             3                 0     48
gpt-3.5-turbo-1106       0             3                 0     48
gpt-4.1-2025-04-14       0             3                 0     8
gpt-4.1-mini-2025-04-14  0             6                 0     24
gpt-4.1-nano-2025-04-14  0             6                 0     24
gpt-4o-2024-08-06        0             3                 0     8
gpt-4o-mini-2024-07-18   0             5                 0     24
o4-mini-2025-04-16       0             2                 0     4

# or
$ (
    echo "model | curr pending | max pending | today jobs | max daily jobs" | tr '[:lower:]' '[:upper:]';
    curl -s https://api.openai.com/v1/fine_tuning/model_limits -H "Authorization: Bearer ${OPENAI_API_KEY}" |
      jq -r '.data[] | [.id, .curr_pending_jobs, .max_pending_jobs, .jobs_so_far_today, .max_jobs_per_day] | @tsv' |
      while IFS=$'\t' read -r model curr max today maxday; do
        echo "${model} | ${curr} | ${max} | ${today} | ${maxday}";
      done
  ) | column -t -s '|'

rate_limits

[!TIP|label:references:]

• rate limits in headers

$ curl -s https://api.openai.com/dashboard/rate_limits \
       -H 'Authorization: Bearer sess-s**************************************P' \
       -H 'User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64)' \
       -H 'Accept: application/json' \
       -H 'Content-Type: application/json' |
  jq -r 'to_entries |
         (["MODEL", "RPM", "TPM", "IMAGES/MIN", "AUDIO_MB/MIN", "RPD", "BATCH_TOKENS"]),
         (.[] | [
           .key,
           (.value.max_requests_per_1_minute        // "-" | if . == "-" then . else (. | tostring + " RPM") end),
           (.value.max_tokens_per_1_minute          // "-" | if . == "-" then . else (. | tostring + " TPM") end),
           (.value.max_images_per_1_minute          // "-" | if . == "-" then . else (. | tostring + " IPM") end),
           (.value.max_audio_megabytes_per_1_minute // "-" | if . == "-" then . else (. | tostring + " MB/min") end),
           (.value.max_requests_per_1_day           // "-" | if . == "-" then . else (. | tostring + " req/day") end),
           (.value.batch_1_day_max_input_tokens     // "-" | if . == "-" then . else (. | tostring + " tokens") end)
         ]) | @tsv' |
  column -t -s $'\t'

Administration API

admin key

[!TIP|label:OPENAI_ADMIN_KEY:]

• openai admin key

• Admin API Keys

• List all organization and project API keys

# get user id, can be used for `-H "OpenAI-User: user-xxxxxxxxxxxxxxxx"`
$ curl -s https://api.openai.com/v1/organization/users \
       -H "Authorization: Bearer $OPENAI_ADMIN_KEY" \
       -H "Content-Type: application/json"

# get details of admin key
$ curl -s https://api.openai.com/v1/organization/admin_api_keys \
       -H "Authorization: Bearer ${OPENAI_ADMIN_KEY}" \
       -H "Content-Type: application/json"

organization id

# get organization id, can be used for `-H "OpenAI-Organization: org-xxxxxxxxxxxxxxxx"`
$ curl -s https://api.openai.com/v1/organizations \
       -H "Authorization: Bearer sess-s**************************************P" \
       -H "Content-Type: application/json" |
  jq -r '.data[].id'
org-G**********************o

# or with ADMIN_KEY
$ curl -s https://api.openai.com/v1/organizations \
       -H "Authorization: Bearer ${OPENAI_ADMIN_KEY}" \
       -H "Content-Type: application/json" |
  jq -r '.data[].id'
org-GUjEKO81ydwmSGORg4boXRso

group id

[!TIP|label:group id:]

• to get group id, can be used for -H "OpenAI-Group: group-xxxxxxxxxxxxxxxx"

# or show in table format
$ curl -s https://api.openai.com/v1/organization/projects \
       -H "Authorization: Bearer ${OPENAI_ADMIN_KEY}" \
       -H "Content-Type: application/json" |
  jq -r '["NAME", "GROUP ID"], (.data[] | [.name, .id]) | @tsv' |
  column -t -s$'\t'
NAME             GROUP ID
Default project  proj_f**********************4
marslo           proj_E**********************m

# or via organization API
$ curl -s https://api.openai.com/v1/organizations \
       -H "Authorization: Bearer ${OPENAI_ADMIN_KEY}" \
       -H "Content-Type: application/json" |
  jq -r '["NAME", "GROUP ID"], (.data[].projects.data[] | [.title, .id]) | @tsv' |
  column -t -s$'\t'
NAME             GROUP ID
Default project  proj_f**********************4
marslo           proj_E**********************m

$ curl -s https://api.openai.com/v1/organization/projects \
       -H "Authorization: Bearer ${OPENAI_ADMIN_KEY}" \
       -H "Content-Type: application/json" |
  jq -r '.data[] | select(.name == "marslo") | .id'

# or show name and id
$ curl -s https://api.openai.com/v1/organization/projects \
       -H "Authorization: Bearer ${OPENAI_ADMIN_KEY}" \
       -H "Content-Type: application/json" |
  jq -r '.data[] | "\(.name)|\(.id)"' | column -t -s'|'

using chatgpt to generate git commits

#!/usr/bin/env bash

set -euo pipefail

model="${COMMITX_MODEL:-gpt-4-1106-preview}"
temperature="${COMMITX_TEMPERATURE:-0.3}"
OPENAI_API_KEY="${OPENAI_API_KEY:? OPENAI_API_KEY not set}"
diff=$(git --no-pager diff --color=never)
prompt="Generate a Git commit message in the Conventional Commits format, with the following structure:

<type>(<scope1,scope2,...>): short summary

- detail 1
- detail 2
- ...

Use multiple scopes if the diff includes changes in multiple areas.
Use markdown-style bullet points for details.
Do not include explanations or code.
Base it on the following diff:

$diff"

declare payload
payload=$(jq -n \
  --arg model "$model" \
  --arg temp "$temperature" \
  --arg prompt "$prompt" \
  '{
    model: $model,
    temperature: ($temp | tonumber),
    messages: [
      { role: "system", content: "You are an assistant that writes Git commit messages." },
      { role: "user", content: $prompt }
    ]
  }')


curl -s https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d "$payload" |
jq -r '.choices[0].message.content'

# vim:tabstop=2:softtabstop=2:shiftwidth=2:expandtab:filetype=sh

using chatgpt to review git diff

#!/usr/bin/env bash

set -euo pipefail

MODEL="${COMMITX_MODEL:-gpt-4-turbo}"
TEMPERATURE="${COMMITX_TEMPERATURE:-0.3}"
OPENAI_API_KEY="${OPENAI_API_KEY:? OPENAI_API_KEY not set}"
diff=$(git --no-pager diff --color=never)
prompt="You are a Principal Software Architect reviewing a Git diff.

You may not have the full code context, but please still provide a professional code review.

Focus on:
- What the changes are doing
- Code quality, style, naming, structure
- Potential bugs or anti-patterns
- Suggestions for improvement

Use markdown. Be helpful, concise, and insightful.

--- START DIFF ---
${diff}
--- END DIFF ---"

curl -s https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d "$(jq -n \
    --arg model "$MODEL" \
    --arg temp "$TEMPERATURE" \
    --arg prompt "$prompt" \
    ' {
      model: $model,
      temperature: ($temp | tonumber),
      messages: [
        { role: "system", content: "You are a senior engineer performing code review." },
        { role: "user", content: $prompt }
      ]
    }')" |
jq -r '.choices[0].message.content'

# vim:tabstop=2:softtabstop=2:shiftwidth=2:expandtab:filetype=sh

RAG

[!TIP]

• RAG - Retrieval-Augmented Generation

tips

chatgpt

see details

import os
from pathlib import Path

# Define script contents
table_script = """#!/usr/bin/env bash

# print_table.sh — CLI 表格打印工具，支持列对齐和彩色输出

_strip_ansi() {
  sed 's/\\x1B\\[[0-9;]*[mK]//g'
}

_str_length() {
  echo -e "$1" | _strip_ansi | awk '{ print length }'
}

repeat_char() {
  printf "%*s" "$2" '' | tr ' ' "$1"
}

print_table() {
  local -n _headers=$1
  local -n _rows=$2
  local -n _aligns=$3

  local cols=${#_headers[@]}
  local -a widths

  # 计算每列最大宽度
  for ((i = 0; i < cols; i++)); do
    widths[i]=$(_str_length "${_headers[i]}")
  done

  for row in "${_rows[@]}"; do
    IFS='|' read -r -a fields <<<"$row"
    for ((i = 0; i < cols; i++)); do
      len=$(_str_length "${fields[i]}")
      ((len > widths[i])) && widths[i]=$len
    done
  done

  # 打印分割线
  _print_line() {
    for ((i = 0; i < cols; i++)); do
      printf "+-%s" "$(repeat_char '-' "${widths[i]}")"
    done
    echo "+"
  }

  # 打印一行
  _print_row() {
    local -a fields=("$@")
    for ((i = 0; i < cols; i++)); do
      local raw="${fields[i]}"
      local clean=$(_str_length "$raw")
      local pad=${widths[i]}
      case "${_aligns[i]}" in
        right)
          printf "| %*s " "$pad" "$raw"
          ;;
        center)
          local left=$(( (pad - clean) / 2 ))
          local right=$(( pad - clean - left ))
          printf "| %*s%s%*s " "$left" "" "$raw" "$right" ""
          ;;
        *)
          printf "| %-*s " "$pad" "$raw"
          ;;
      esac
    done
    echo "|"
  }

  # 打印整张表格
  _print_line
  _print_row "${_headers[@]}"
  _print_line
  for row in "${_rows[@]}"; do
    IFS='|' read -r -a fields <<<"$row"
    _print_row "${fields[@]}"
  done
  _print_line
}
"""

# Define output path
output_path = Path("/mnt/data/cli-ui/print_table.sh")
output_path.parent.mkdir(parents=True, exist_ok=True)

# Save script to file
with open(output_path, "w") as f:
    f.write(table_script)

output_path.name