baidu-text-translate（百度文字翻译）

baidu_text_translate — Agent Reference

trans text wraps the Baidu AI Translation API. Use --json — it separates success (stdout) from errors (stderr) and gives you structured fields to act on.

trans text --json "你好世界"                             # auto-detect → English
trans text --json --from zh --to jp "你好"               # explicit languages
trans text --json --to fra --reference "使用正式语气" "你好"  # custom instruction
echo "早上好" | trans text --json                        # stdin pipe

Environment Diagnosis

Before translating, verify the environment with trans doctor:

trans doctor           # human-readable output
trans doctor --json    # JSON output for Agent parsing

JSON contract (stdout):

{
  "checks": [
    {"name":"api_key",     "ok":true,  "source":"env/config", "message":"configured"},
    {"name":"connectivity","ok":true,  "latency_ms":243,       "message":"reachable"},
    {"name":"account",     "ok":true,                          "message":"valid"}
  ],
  "all_ok": true
}

When a check fails, help_url in the failing check points to the fix:

• api_key.ok:false → set TRANS_API_KEY (see Configuration below)
• account.ok:false → key invalid/expired or quota exhausted

Run trans doctor --json as the first step when diagnosing any trans-cli failure.

JSON Contract

Success (stdout, exit 0):

{"from":"zh","to":"en","source":"你好","translated":"Hello"}

Error (stderr, exit ≠ 0):

{"code":"AUTH_FAILED","message":"…","help_url":"https://…","raw_code":"52001"}

help_url and raw_code are omitted when not applicable.

Error Decision Tree

Exit 2 requires human intervention — the agent cannot fix auth or config. Exit 1 means the agent passed bad arguments — fix the call, not the environment.

Language Codes

Baidu does not follow ISO 639. When unsure of a code, query locally:

trans languages                    # full list (200+ languages)
trans languages --filter jp        # search by code
trans languages --filter 日语      # search by name
trans languages --json             # JSON output for Agent parsing

The six most common ISO traps:

Other common codes that match ISO: zh cht en ru de pt it nl hi th Use auto for source-language auto-detection.

Configuration

Priority: --api-key flag > TRANS_API_KEY env > ~/.trans-cli/config.json

Config file field: api_key.

trans config init               # create empty config skeleton (~/.trans-cli/config.json)
trans config init --force       # overwrite existing config
trans config set api_key <KEY>  # write api_key (creates file if absent)

Getting an API Key (First-Time Setup)

Guide the user through these steps — the agent cannot do this on their behalf:

1. Open the Baidu Translation Open Platform and sign in: <https://fanyi-api.baidu.com/>
2. Go to Developer Center: <https://fanyi-api.baidu.com/manage/developer>
3. Click "Use Now" → select "General Translation" → complete identity verification
4. Get your API Key: <https://fanyi-api.baidu.com/manage/apiKey>
5. Write to config: trans config set api_key <KEY>

or set the environment variable TRANS_API_KEY

1. Verify: trans doctor --json — all_ok:true means success

Recharging (QUOTA_EXCEEDED)

Recharge portal: <https://fanyi-api.baidu.com/manage/account> Takes effect immediately — no restart needed. Auto-renewal can be enabled on the same page. Usage details: <https://fanyi-api.baidu.com/api/trans/user/usage>

Non-Obvious Behaviours

• Multiple positional args join with a space: trans text hello world → "hello world"
• When both stdin and positional args are given, positional args win silently
• The from field in JSON is the API's detected language, not the --from flag value
• trans doctor checks are short-circuit: if api_key fails, connectivity/account are skipped
• --reference is optional; omitting it sends no extra field to the API