@oneciel-ai/claude-any 0.1.39 → 0.1.43

package/README.md

@@ -48,7 +48,7 @@ arguments through unchanged.
 
 Credits: One Ciel LLC
 
-Current version: `0.1.39`
+Current version: `0.1.43`
 
 ## Why This Exists
 
@@ -381,6 +381,27 @@ steps under that larger model's supervision.
 
 ## Changelog
 
+### 0.1.43
+
+- **429 backoff retry**: upstream `429 Too Many Requests` responses are now
+  handled as retry/backoff events across all retry attempts instead of leaking
+  the raw upstream error after the first backoff.
+
+### 0.1.42
+
+- **Live stream progress**: the statusline now updates streamed upstream output
+  progress with formatted input/output token estimates and chunk counts.
+
+### 0.1.41
+
+- **Statusline formatting**: upstream token counts now use thousands separators
+  and a space before `tok`, for example `27,501 tok`.
+
+### 0.1.40
+
+- **RPM 0 is preserved**: setting `rate_limit_rpm=0` now stores an explicit
+  unlimited mode instead of falling back to the provider default.
+
 ### 0.1.39
 
 - **Menu input fixes**: restores terminal line/echo mode before text or number

package/claude_any.py

@@ -85,7 +85,7 @@ PROVIDER_LABELS = {
     "self-hosted-nim": "Self Hosted NIM",
 }
 APP_NAME = "Claude Any"
-VERSION = "0.1.39"
+VERSION = "0.1.43"
 CREDITS = "Credits: One Ciel LLC"
 
 LOG_LEVELS = {"SILENT": 0, "ERROR": 1, "WARN": 2, "INFO": 3, "DEBUG": 4, "TRACE": 5}
@@ -1339,7 +1339,22 @@ def main():
                 tokens = activity.get("tokens")
                 rpm_text += f" | upstream {age:.0f}s"
                 if tokens:
-                    rpm_text += f" {tokens}tok"
+                    try:
+                        rpm_text += f" {int(tokens):,} tok"
+                    except Exception:
+                        rpm_text += f" {tokens} tok"
+                output_tokens = activity.get("output_tokens")
+                if output_tokens:
+                    try:
+                        rpm_text += f" -> {int(output_tokens):,} tok"
+                    except Exception:
+                        rpm_text += f" -> {output_tokens} tok"
+                chunks = activity.get("chunks")
+                if chunks:
+                    try:
+                        rpm_text += f" ({int(chunks):,} chunks)"
+                    except Exception:
+                        rpm_text += f" ({chunks} chunks)"
             elif event in ("success", "error"):
                 rpm_text += f" | {event} {age:.0f}s"
     print(f"{left} | {color(rpm_text)}")
@@ -2207,8 +2222,10 @@ def router_rate_limit_recent(timestamps: Any, now: float, window: float, *, incl
 
 def router_rate_limit_usage(provider: str, pcfg: dict[str, Any], model: str | None = None) -> tuple[int, int | None]:
     rpm = router_rate_limit_effective_rpm(provider, pcfg, model)
-    if rpm is None:
-        return 0, None
+    if rpm is None:
+        return 0, None
+    if rpm == 0:
+        return 0, 0
     key = router_rate_limit_key(provider, pcfg, model)
     now = time.time()
     try:
@@ -4697,6 +4714,8 @@ def stream_openai_chat_to_anthropic_sse(
     source_body: dict[str, Any] | None = None,
     start_index: int = 0,
     word_chunking: bool = False,
+    input_tokens: int | None = None,
+    input_bytes: int | None = None,
 ) -> None:
     next_content_index = start_index
     text_started = False
@@ -4709,6 +4728,8 @@ def stream_openai_chat_to_anthropic_sse(
     tool_fragments: dict[int, dict[str, Any]] = {}
     output_tokens = 0
     finish_reason = "stop"
+    chunks_seen = 0
+    last_activity_update = 0.0
 
     def emit(event_name: str, payload: dict[str, Any]) -> None:
         handler.wfile.write(f"event: {event_name}\ndata: {json.dumps(payload, ensure_ascii=False)}\n\n".encode())
@@ -4736,8 +4757,27 @@ def stream_openai_chat_to_anthropic_sse(
             {"type": "content_block_delta", "index": idx, "delta": {"type": "text_delta", "text": text}},
         )
 
+    def update_stream_activity(force: bool = False) -> None:
+        nonlocal last_activity_update
+        now = time.time()
+        if not force and now - last_activity_update < 0.5:
+            return
+        last_activity_update = now
+        estimated_output = output_tokens or max(0, len(text_so_far) // 4)
+        write_router_activity(
+            "request",
+            provider,
+            model,
+            tokens=input_tokens,
+            bytes=input_bytes,
+            output_tokens=estimated_output,
+            chunks=chunks_seen,
+            stream=True,
+        )
+
     try:
         for raw_line in resp:
+            chunks_seen += 1
             line = raw_line.decode("utf-8", errors="ignore").strip()
             if not line or line.startswith(":"):
                 continue
@@ -4789,6 +4829,7 @@ def stream_openai_chat_to_anthropic_sse(
                     emit_text_delta(to_flush)
                 else:
                     emit_text_delta(text_chunk)
+                update_stream_activity()
             for call in delta.get("tool_calls") or []:
                 if not isinstance(call, dict):
                     continue
@@ -4804,6 +4845,8 @@ def stream_openai_chat_to_anthropic_sse(
                     slot["name"] += str(fn.get("name"))
                 if fn.get("arguments"):
                     slot["arguments"] += str(fn.get("arguments"))
+                update_stream_activity()
+        update_stream_activity(force=True)
         if word_chunking and text_buffer:
             to_flush, text_buffer = _split_word_buffer(text_buffer, force=True)
             emit_text_delta(to_flush)
@@ -4928,19 +4971,30 @@ def upstream_http_error_message(exc: urllib.error.HTTPError, raw: str | None = N
 UPSTREAM_RETRY_HTTP_CODES: frozenset[int] = frozenset({502, 503, 504})
 
 
-def upstream_retry_message(attempt: int, total: int) -> str:
-    lang = str(load_config().get("language") or "en")
-    if lang == "ko":
-        return f"서버가 응답하지 않아 재시도합니다 ({attempt}/{total})."
+def upstream_retry_message(attempt: int, total: int) -> str:
+    lang = str(load_config().get("language") or "en")
+    if lang == "ko":
+        return f"서버가 응답하지 않아 재시도합니다 ({attempt}/{total})."
     if lang == "ja":
         return f"サーバーが応答しないため再試行します ({attempt}/{total})。"
     if lang == "zh":
         return f"服务器未响应，正在重试 ({attempt}/{total})。"
-    return f"Upstream server did not respond; retrying ({attempt}/{total})."
-
-
-def upstream_retry_wait_seconds(attempt: int) -> float:
-    return min(20.0, 2.0 * max(1, attempt))
+    return f"Upstream server did not respond; retrying ({attempt}/{total})."
+
+
+def upstream_rate_limit_retry_message(attempt: int, total: int) -> str:
+    lang = str(load_config().get("language") or "en")
+    if lang == "ko":
+        return f"Upstream rate limit에 도달해 대기 후 재시도합니다 ({attempt}/{total})."
+    if lang == "ja":
+        return f"Upstream rate limit に達したため、待機して再試行します ({attempt}/{total})。"
+    if lang == "zh":
+        return f"已达到 upstream rate limit，等待后重试 ({attempt}/{total})。"
+    return f"Upstream rate limit reached; waiting before retry ({attempt}/{total})."
+
+
+def upstream_retry_wait_seconds(attempt: int) -> float:
+    return min(20.0, 2.0 * max(1, attempt))
 
 
 def retryable_timeout_exception(exc: BaseException) -> bool:
@@ -4985,10 +5039,15 @@ def post_json_with_rate_retry(
         except urllib.error.HTTPError as exc:
             raw = exc.read().decode("utf-8", errors="ignore")
             learn_router_rate_limit_headers(provider, pcfg, model, exc.headers)
-            if exc.code == 429 and attempt == 0:
-                wait = register_router_rate_limit_backoff(provider, pcfg, model, exc.headers.get("Retry-After"))
-                time.sleep(wait)
-                continue
+            if exc.code == 429 and attempt + 1 < max_attempts:
+                retry_no = attempt + 1
+                wait = register_router_rate_limit_backoff(provider, pcfg, model, exc.headers.get("Retry-After"))
+                write_router_activity("retry", provider, model, attempt=retry_no, total=gateway_retries, code=exc.code, wait=wait, tokens=token_estimate, bytes=byte_estimate)
+                router_log("WARN", f"upstream_rate_limit_retry provider={provider} model={model} attempt={retry_no}/{gateway_retries} wait={wait:.2f}s tokens={token_estimate} bytes={byte_estimate}")
+                if retry_notice:
+                    retry_notice(upstream_rate_limit_retry_message(retry_no, gateway_retries))
+                time.sleep(wait)
+                continue
             if exc.code in UPSTREAM_RETRY_HTTP_CODES and attempt + 1 < max_attempts:
                 retry_no = attempt + 1
                 write_router_activity("retry", provider, model, attempt=retry_no, total=gateway_retries, code=exc.code, tokens=token_estimate, bytes=byte_estimate)
@@ -5049,8 +5108,13 @@ def open_openai_stream_with_rate_retry(
         except urllib.error.HTTPError as exc:
             raw = exc.read().decode("utf-8", errors="ignore")
             learn_router_rate_limit_headers(provider, pcfg, model, exc.headers)
-            if exc.code == 429 and attempt == 0:
+            if exc.code == 429 and attempt + 1 < max_attempts:
+                retry_no = attempt + 1
                 wait = register_router_rate_limit_backoff(provider, pcfg, model, exc.headers.get("Retry-After"))
+                write_router_activity("retry", provider, model, attempt=retry_no, total=gateway_retries, code=exc.code, wait=wait, tokens=token_estimate, bytes=byte_estimate, stream=True)
+                router_log("WARN", f"upstream_stream_rate_limit_retry provider={provider} model={model} attempt={retry_no}/{gateway_retries} wait={wait:.2f}s tokens={token_estimate} bytes={byte_estimate}")
+                if retry_notice:
+                    retry_notice(upstream_rate_limit_retry_message(retry_no, gateway_retries))
                 time.sleep(wait)
                 continue
             if exc.code in UPSTREAM_RETRY_HTTP_CODES and attempt + 1 < max_attempts:
@@ -5108,6 +5172,8 @@ def forward_openai_compatible_chat(handler: BaseHTTPRequestHandler, provider: st
                 model,
                 emit_retry_notice,
             )
+            req_tokens = estimate_tokens(req_body)
+            req_bytes = len(json.dumps(req_body, ensure_ascii=False).encode("utf-8"))
             stream_openai_chat_to_anthropic_sse(
                 handler,
                 resp,
@@ -5116,8 +5182,10 @@ def forward_openai_compatible_chat(handler: BaseHTTPRequestHandler, provider: st
                 source_body=body,
                 start_index=index,
                 word_chunking=bool(pcfg.get("stream_word_chunking", False)),
+                input_tokens=req_tokens,
+                input_bytes=req_bytes,
             )
-            write_router_activity("success", provider, model, tokens=estimate_tokens(req_body), bytes=len(json.dumps(req_body, ensure_ascii=False).encode("utf-8")), stream=True)
+            write_router_activity("success", provider, model, tokens=req_tokens, bytes=req_bytes, stream=True)
         except RuntimeError as exc:
             msg = str(exc)
             write_anthropic_stream_blocks(handler, [{"type": "text", "text": f"Upstream error: {msg}"}], index)
@@ -6663,11 +6731,11 @@ def apply_provider_option(provider: str, pcfg: dict[str, Any], token: str) -> No
             raise SystemExit("timeout must be a positive integer; values above 10000 are treated as milliseconds")
         pcfg["request_timeout_ms"] = fixed if key.endswith("_ms") or fixed > 10000 else fixed * 1000
         return
-    if key in ("rate_limit", "rate_limit_rpm", "rpm"):
-        fixed = positive_int(value)
-        if value in (0, "0", False, None):
-            pcfg.pop("rate_limit_rpm", None)
-            return
+    if key in ("rate_limit", "rate_limit_rpm", "rpm"):
+        fixed = positive_int(value)
+        if value in (0, "0", False, None):
+            pcfg["rate_limit_rpm"] = 0
+            return
         if not fixed:
             raise SystemExit("rate_limit_rpm must be a positive integer, or 0/unset to disable")
         pcfg["rate_limit_rpm"] = fixed

package/docs/README.ja.md

@@ -47,7 +47,7 @@ vLLM、NVIDIA hosted、self-hosted NIM を選択し、通常の Claude Code 引
 
 Credits: One Ciel LLC
 
-現在のバージョン: `0.1.39`
+現在のバージョン: `0.1.43`
 
 ## 作られた理由
 
@@ -351,6 +351,26 @@ Windows/Linux 管理、クリーンアップスクリプト、定期的なセキ
 
 ## 変更履歴
 
+### 0.1.43
+
+- **429 backoff retry**: upstream `429 Too Many Requests` 応答を初回 backoff 後に
+  raw error として漏らさず、すべての retry attempt で backoff/retry event として処理します。
+
+### 0.1.42
+
+- **ライブストリーム進捗**: statusline が upstream streaming の出力進捗を
+  入力/出力 token 推定値と chunk 数で継続更新します。
+
+### 0.1.41
+
+- **Statusline 表示改善**: upstream token 数に桁区切りと `tok` 前の空白を入れ、
+  `27,501 tok` のように表示します。
+
+### 0.1.40
+
+- **RPM 0 を保持**: `rate_limit_rpm=0` の設定が provider 既定値に戻らず、
+  明示的な無制限モードとして保存されます。
+
 ### 0.1.39
 
 - **メニュー入力修正**: テキスト/数字プロンプトの前に terminal line/echo mode を

package/docs/README.ko.md

@@ -47,7 +47,7 @@ NVIDIA hosted, self-hosted NIM을 선택하고, Claude Code의 일반 인자는
 
 Credits: One Ciel LLC
 
-현재 버전: `0.1.39`
+현재 버전: `0.1.43`
 
 ## 왜 만들었나
 
@@ -351,6 +351,26 @@ Windows 이벤트 로그 리뷰, 바이러스/랜섬웨어 침입 시도 정리,
 
 ## 변경 이력
 
+### 0.1.43
+
+- **429 backoff retry**: upstream `429 Too Many Requests` 응답을 첫 backoff 이후
+  raw error로 흘리지 않고, 모든 retry attempt에서 backoff/retry 이벤트로 처리합니다.
+
+### 0.1.42
+
+- **실시간 스트림 진행 표시**: statusline이 upstream streaming 출력 진행을
+  입력/출력 token 추정치와 chunk 수로 계속 갱신합니다.
+
+### 0.1.41
+
+- **Statusline 표시 개선**: upstream token 수에 천 단위 구분자와 `tok` 앞 공백을
+  넣어 `27,501 tok`처럼 표시합니다.
+
+### 0.1.40
+
+- **RPM 0 유지**: `rate_limit_rpm=0` 설정이 provider 기본값으로 되돌아가지 않고
+  명시적인 무제한 모드로 저장됩니다.
+
 ### 0.1.39
 
 - **메뉴 입력 수정**: 텍스트/숫자 프롬프트 전에 터미널 line/echo 모드를 복구하여

package/docs/README.zh.md

@@ -47,7 +47,7 @@ NIM，并把普通 Claude Code 参数原样传递。
 
 Credits: One Ciel LLC
 
-当前版本: `0.1.39`
+当前版本: `0.1.43`
 
 ## 为什么存在
 
@@ -337,6 +337,26 @@ Hermes 格式模型或部分较旧的 Qwen tool template。
 
 ## 更新日志
 
+### 0.1.43
+
+- **429 backoff retry**：upstream `429 Too Many Requests` 响应现在会在所有 retry
+  attempt 中作为 backoff/retry event 处理，不再在首次 backoff 后泄漏 raw error。
+
+### 0.1.42
+
+- **实时流式进度**：statusline 会持续更新 upstream streaming 输出进度，
+  显示输入/输出 token 估算值和 chunk 数。
+
+### 0.1.41
+
+- **Statusline 格式优化**：upstream token 数现在带千位分隔符，并在 `tok` 前加入空格，
+  例如 `27,501 tok`。
+
+### 0.1.40
+
+- **保留 RPM 0**：`rate_limit_rpm=0` 现在会保存为明确的无限制模式，
+  不会回退到 provider 默认值。
+
 ### 0.1.39
 
 - **菜单输入修复**：在文本/数字提示前恢复 terminal line/echo mode，

package/docs/manual.md

@@ -10,7 +10,7 @@ Code starts, while passing normal Claude Code arguments through unchanged.
 
 Credits: One Ciel LLC
 
-Current version: `0.1.39`
+Current version: `0.1.43`
 
 ## Install
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@oneciel-ai/claude-any",
-  "version": "0.1.39",
+  "version": "0.1.43",
   "description": "Claude Code provider selector for Anthropic, Ollama, Ollama Cloud, vLLM, NVIDIA hosted, and self-hosted NIM.",
   "license": "MIT",
   "author": "One Ciel LLC",