feat: add profile editing and richer request ids

README.md

@@ -61,7 +61,7 @@ uv run codex-retry-gateway-tui
 Views:
 
 - overview: current gateway status and summary counts
-- requests: recent requests, request ID, timing, usage, attempts, and `usage_last_updated_at`
+- requests: recent requests, response/request/thread IDs, timing, usage, cached ratio, attempts, and `usage_last_updated_at`
 - logs: recent gateway logs
 - profiles: saved profiles, active profile default selection, profile actions
 
@@ -71,6 +71,7 @@ Controls:
 - `2` requests
 - `3` logs
 - `4` profiles
+- `e` edit selected profile in `$VISUAL` / `$EDITOR`
 - `/` filter
 - `r` refresh
 - `p` probe selected profile
@@ -80,3 +81,5 @@ Controls:
 - `o` open selected profile upstream URL
 
 This client uses only the public gateway admin API and does not need SSH or secrets.
+
+Profile editing uses a temp TOML draft opened in `$VISUAL` or `$EDITOR` and then saves it back through the gateway profiles API.

codex_retry_gateway_tui.py

@@ -9,8 +9,12 @@ import importlib.metadata
 import json
 import os
 import re
+import shlex
+import shutil
 import subprocess
 import sys
+import tempfile
+import tomllib
 import urllib.parse
 import urllib.request
 from pathlib import Path
@@ -30,6 +34,21 @@ DEFAULT_VERSION_CHECK_URL = "https://gitea.shujk.top/shujakuin/codex-retry-gatew
 DEFAULT_REFRESH_SECONDS = 10
 DEFAULT_TIMEOUT_SECONDS = 5
 DEFAULT_VERSION_CHECK_TIMEOUT_SECONDS = 2
+DEFAULT_PROFILE_REASONING_EQUALS = [516, 1034, 1552]
+DEFAULT_PROFILE_RETRYABLE_STATUS_CODES = [429, 503]
+DEFAULT_PROFILE_RETRYABLE_ERROR_MESSAGES = [
+    "Selected model is at capacity. Please try a different model.",
+    "stream disconnected before completion: Concurrency limit exceeded for account, please retry later",
+]
+DEFAULT_PROFILE_UPSTREAM_FETCH_RETRY_ATTEMPTS = 5
+DEFAULT_PROFILE_UPSTREAM_FETCH_RETRY_BACKOFF_MS = 350
+DEFAULT_PROFILE_REQUEST_HISTORY_LIMIT = 0
+DEFAULT_PROFILE_ENDPOINTS = [
+    "/responses",
+    "/chat/completions",
+    "/v1/responses",
+    "/v1/chat/completions",
+]
 INSTALL_COMMAND = "uv tool install --force git+https://gitea.shujk.top/shujakuin/codex-retry-gateway-tui.git"
 INSTALL_COMMAND_ARGS = [
     "uv",
@@ -312,6 +331,39 @@ def format_percent(value: Any) -> str:
     return f"{rounded}%"
 
 
+def cached_ratio(input_tokens: Any, cached_tokens: Any) -> float | None:
+    total = as_float(input_tokens)
+    if total <= 0:
+        return None
+    cached = max(0.0, as_float(cached_tokens))
+    return max(0.0, min(1.0, cached / total))
+
+
+def effective_input_tokens(input_tokens: Any, cached_tokens: Any) -> int | None:
+    total = parse_int_value(input_tokens)
+    if total is None:
+        return None
+    cached = max(0, as_int(cached_tokens))
+    return max(0, total - cached)
+
+
+def primary_request_id(row: dict[str, Any]) -> str:
+    return str(row.get("response_id") or row.get("request_id") or "").strip() or "-"
+
+
+def request_usage_summary(row: dict[str, Any]) -> str:
+    effective_in = effective_input_tokens(row.get("input_tokens"), row.get("cached_tokens"))
+    output_tokens = row.get("output_tokens")
+    cached_tokens = row.get("cached_tokens")
+    ratio = cached_ratio(row.get("input_tokens"), cached_tokens)
+    in_text = format_count(effective_in) if effective_in is not None else "-"
+    out_text = format_count(output_tokens) if output_tokens is not None else "-"
+    cached_text = format_count(cached_tokens) if cached_tokens is not None else "-"
+    if ratio is not None:
+        cached_text = f"{cached_text} ({format_percent(ratio * 100)})"
+    return f"in {in_text} | out {out_text} | cached {cached_text}"
+
+
 def format_count(value: Any) -> str:
     number = as_int(value)
     if abs(number) >= 1_000_000:
@@ -416,6 +468,93 @@ def format_bool(value: Any) -> str:
     return "yes" if bool(value) else "no"
 
 
+def parse_bool_value(value: Any, default: bool = False) -> bool:
+    if isinstance(value, bool):
+        return value
+    if value is None:
+        return default
+    text = str(value).strip().lower()
+    if text in {"1", "true", "yes", "on"}:
+        return True
+    if text in {"0", "false", "no", "off", ""}:
+        return False
+    return bool(value)
+
+
+def parse_int_value(value: Any) -> int | None:
+    if value is None or str(value).strip() == "":
+        return None
+    try:
+        return int(str(value).strip())
+    except Exception:
+        try:
+            return int(float(str(value).strip()))
+        except Exception:
+            return None
+
+
+def normalize_editor_integer_list(value: Any) -> list[int]:
+    if isinstance(value, (list, tuple, set)):
+        raw_values = list(value)
+    else:
+        raw_values = re.split(r"[\s,]+", str(value or ""))
+    numbers: list[int] = []
+    for item in raw_values:
+        parsed = parse_int_value(item)
+        if parsed is not None:
+            numbers.append(parsed)
+    return numbers
+
+
+def normalize_editor_string_list(value: Any) -> list[str]:
+    if isinstance(value, (list, tuple, set)):
+        raw_values = list(value)
+    else:
+        raw_values = re.split(r"[\r\n,]+", str(value or ""))
+    values = []
+    for item in raw_values:
+        text = str(item or "").strip()
+        if text:
+            values.append(text)
+    return values
+
+
+def normalize_editor_phrase_list(value: Any) -> list[str]:
+    if isinstance(value, (list, tuple, set)):
+        raw_values = list(value)
+    else:
+        raw_values = str(value or "").splitlines()
+    values = []
+    for item in raw_values:
+        text = str(item or "").strip()
+        if text:
+            values.append(text)
+    return values
+
+
+def toml_string(value: Any) -> str:
+    return json.dumps(str(value or ""), ensure_ascii=False)
+
+
+def toml_bool(value: Any) -> str:
+    return "true" if parse_bool_value(value) else "false"
+
+
+def toml_int_list(values: Any) -> str:
+    numbers = normalize_editor_integer_list(values)
+    return "[" + ", ".join(str(value) for value in numbers) + "]"
+
+
+def toml_string_list(values: Any) -> str:
+    items = [toml_string(value) for value in normalize_editor_string_list(values)]
+    return "[" + ", ".join(items) + "]"
+
+
+def toml_phrase_list(values: Any) -> str:
+    items = [toml_string(value) for value in normalize_editor_phrase_list(values)]
+    return "[" + ", ".join(items) + "]"
+
+
 def summary_line(payload: dict[str, Any]) -> str:
     config = payload.get("config") if isinstance(payload.get("config"), dict) else {}
     state = payload.get("state") if isinstance(payload.get("state"), dict) else {}
@@ -466,6 +605,8 @@ def request_match_score(row: dict[str, Any], needle: str) -> bool:
         str(row.get(key) or "")
         for key in (
             "request_id",
+            "response_id",
+            "thread_id",
             "profile_name",
             "method",
             "path",
@@ -520,6 +661,8 @@ def normalize_request_rows(payload: dict[str, Any], filter_text: str = "") -> li
             {
                 "seq": as_int(entry.get("seq")),
                 "request_id": str(entry.get("request_id") or ""),
+                "response_id": str(entry.get("response_id") or ""),
+                "thread_id": str(entry.get("thread_id") or ""),
                 "profile_name": str(entry.get("profile_name") or ""),
                 "method": str(entry.get("method") or ""),
                 "path": str(entry.get("path") or ""),
@@ -562,7 +705,8 @@ def normalize_request_rows(payload: dict[str, Any], filter_text: str = "") -> li
 def render_request_detail(row: dict[str, Any]) -> str:
     bits = [
         f"seq {row['seq']}",
-        f"id {row['request_id'] or '-'}",
+        f"id {primary_request_id(row)}",
+        f"thread {row.get('thread_id') or '-'}",
         f"{row['method']} {row['path']}",
         f"profile {row.get('profile_name') or '-'}",
         f"status {row['status_code'] or '-'}",
@@ -586,12 +730,10 @@ def render_request_detail(row: dict[str, Any]) -> str:
         bits.append(f"forwarded {row['forwarded_model']}")
     if row.get("reasoning_tokens") is not None:
         bits.append(f"reasoning {row['reasoning_tokens']}")
-    if row.get("input_tokens") is not None or row.get("output_tokens") is not None or row.get("total_tokens") is not None:
-        bits.append(
-            f"usage in/out/total {row.get('input_tokens') or 0}/{row.get('output_tokens') or 0}/{row.get('total_tokens') or 0}"
-        )
+    if row.get("input_tokens") is not None or row.get("output_tokens") is not None or row.get("total_tokens") is not None or row.get("cached_tokens") is not None:
+        bits.append(f"usage {request_usage_summary(row)} | total {format_count(row.get('total_tokens'))}")
     if row.get("cached_tokens") is not None:
-        bits.append(f"cached {row['cached_tokens']}")
+        bits.append(f"cached raw {format_count(row['cached_tokens'])}")
     if row.get("error"):
         bits.append(f"error {row['error']}")
     if row.get("upstream_origin"):
@@ -661,6 +803,171 @@ def render_profile_detail(row: dict[str, Any]) -> str:
     return " | ".join(bits)
 
 
+def profile_form_state(row: dict[str, Any]) -> dict[str, Any]:
+    raw = row.get("raw") if isinstance(row.get("raw"), dict) else {}
+    form = raw.get("form") if isinstance(raw.get("form"), dict) else {}
+    return {
+        "name": row.get("name") or "",
+        "listen_host": form.get("listen_host") or row.get("listen_host") or "",
+        "listen_port": parse_int_value(form.get("listen_port") or row.get("listen_port")),
+        "upstream_base_url": form.get("upstream_base_url") or row.get("upstream_base_url") or "",
+        "auth_mode": form.get("auth_mode") or row.get("auth_mode") or "passthrough",
+        "auth_env": form.get("auth_env") or "",
+        "auth_file": form.get("auth_file") or "",
+        "manual_secret": "",
+        "manual_secret_file": form.get("manual_secret_file") or "",
+        "manual_secret_configured": parse_bool_value(form.get("manual_secret_configured"), False),
+        "auth_json_path": form.get("auth_json_path") or "",
+        "auth_json_key": form.get("auth_json_key") or row.get("auth_json_key") or "",
+        "request_history_limit": parse_int_value(form.get("request_history_limit") or row.get("request_history_limit"))
+        if parse_int_value(form.get("request_history_limit") or row.get("request_history_limit")) is not None
+        else DEFAULT_PROFILE_REQUEST_HISTORY_LIMIT,
+        "model_remap": form.get("model_remap") or row.get("model_remap") or "",
+        "reasoning_equals": normalize_editor_integer_list(form.get("reasoning_equals") or row.get("reasoning_equals"))
+        or list(DEFAULT_PROFILE_REASONING_EQUALS),
+        "retryable_status_codes": normalize_editor_integer_list(form.get("retryable_status_codes"))
+        or list(DEFAULT_PROFILE_RETRYABLE_STATUS_CODES),
+        "retryable_error_messages": normalize_editor_phrase_list(form.get("retryable_error_messages"))
+        or list(DEFAULT_PROFILE_RETRYABLE_ERROR_MESSAGES),
+        "upstream_fetch_retry_attempts": parse_int_value(form.get("upstream_fetch_retry_attempts"))
+        if parse_int_value(form.get("upstream_fetch_retry_attempts")) is not None
+        else DEFAULT_PROFILE_UPSTREAM_FETCH_RETRY_ATTEMPTS,
+        "upstream_fetch_retry_backoff_ms": parse_int_value(form.get("upstream_fetch_retry_backoff_ms"))
+        if parse_int_value(form.get("upstream_fetch_retry_backoff_ms")) is not None
+        else DEFAULT_PROFILE_UPSTREAM_FETCH_RETRY_BACKOFF_MS,
+        "endpoints": normalize_editor_string_list(form.get("endpoints")) or list(DEFAULT_PROFILE_ENDPOINTS),
+    }
+
+
+def profile_payload_from_row(row: dict[str, Any]) -> dict[str, Any]:
+    state = profile_form_state(row)
+    return {
+        "name": str(state.get("name") or "").strip(),
+        "listen_host": str(state.get("listen_host") or "").strip(),
+        "listen_port": state.get("listen_port"),
+        "upstream_base_url": str(state.get("upstream_base_url") or "").strip(),
+        "auth_mode": str(state.get("auth_mode") or "passthrough").strip(),
+        "auth_env": str(state.get("auth_env") or "").strip(),
+        "auth_file": str(state.get("auth_file") or "").strip(),
+        "manual_secret": "",
+        "manual_secret_file": str(state.get("manual_secret_file") or "").strip(),
+        "manual_secret_configured": bool(state.get("manual_secret_configured")),
+        "auth_json_path": str(state.get("auth_json_path") or "").strip(),
+        "auth_json_key": str(state.get("auth_json_key") or "").strip(),
+        "request_history_limit": state.get("request_history_limit"),
+        "model_remap": str(state.get("model_remap") or "").strip(),
+        "reasoning_equals": state.get("reasoning_equals") or [],
+        "retryable_status_codes": state.get("retryable_status_codes") or [],
+        "retryable_error_messages": state.get("retryable_error_messages") or [],
+        "upstream_fetch_retry_attempts": state.get("upstream_fetch_retry_attempts"),
+        "upstream_fetch_retry_backoff_ms": state.get("upstream_fetch_retry_backoff_ms"),
+        "endpoints": state.get("endpoints") or [],
+    }
+
+
+def profile_editor_document(row: dict[str, Any]) -> str:
+    state = profile_form_state(row)
+    model_remap = str(state.get("model_remap") or "")
+    if "\n" in model_remap:
+        model_remap_value = '"""\n' + model_remap.rstrip("\n").replace('"""', '\\"""') + '\n"""'
+    else:
+        model_remap_value = toml_string(model_remap)
+    return "\n".join(
+        [
+            "# Edit the selected codex-retry-gateway profile and save.",
+            "# Leave manual_secret empty to keep the current secret file.",
+            "# Changing name creates a new profile file; it does not delete the old one.",
+            "",
+            f"name = {toml_string(state.get('name'))}",
+            f"listen_host = {toml_string(state.get('listen_host'))}",
+            f"listen_port = {state.get('listen_port') if state.get('listen_port') is not None else 4610}",
+            f"upstream_base_url = {toml_string(state.get('upstream_base_url'))}",
+            f"auth_mode = {toml_string(state.get('auth_mode'))}",
+            f"auth_env = {toml_string(state.get('auth_env'))}",
+            f"auth_file = {toml_string(state.get('auth_file'))}",
+            'manual_secret = ""',
+            f"manual_secret_file = {toml_string(state.get('manual_secret_file'))}",
+            f"manual_secret_configured = {toml_bool(state.get('manual_secret_configured'))}",
+            f"auth_json_path = {toml_string(state.get('auth_json_path'))}",
+            f"auth_json_key = {toml_string(state.get('auth_json_key'))}",
+            f"request_history_limit = {state.get('request_history_limit') if state.get('request_history_limit') is not None else 0}",
+            f"model_remap = {model_remap_value}",
+            f"reasoning_equals = {toml_int_list(state.get('reasoning_equals'))}",
+            f"retryable_status_codes = {toml_int_list(state.get('retryable_status_codes'))}",
+            f"retryable_error_messages = {toml_phrase_list(state.get('retryable_error_messages'))}",
+            f"upstream_fetch_retry_attempts = {state.get('upstream_fetch_retry_attempts') if state.get('upstream_fetch_retry_attempts') is not None else 5}",
+            f"upstream_fetch_retry_backoff_ms = {state.get('upstream_fetch_retry_backoff_ms') if state.get('upstream_fetch_retry_backoff_ms') is not None else 350}",
+            f"endpoints = {toml_string_list(state.get('endpoints'))}",
+            "",
+        ]
+    )
+
+
+def profile_payload_from_editor_text(text: str) -> dict[str, Any]:
+    data = tomllib.loads(text)
+    if not isinstance(data, dict):
+        raise ValueError("editor payload must be a TOML object")
+    return {
+        "name": str(data.get("name") or "").strip(),
+        "listen_host": str(data.get("listen_host") or "").strip(),
+        "listen_port": parse_int_value(data.get("listen_port")),
+        "upstream_base_url": str(data.get("upstream_base_url") or "").strip(),
+        "auth_mode": str(data.get("auth_mode") or "passthrough").strip(),
+        "auth_env": str(data.get("auth_env") or "").strip(),
+        "auth_file": str(data.get("auth_file") or "").strip(),
+        "manual_secret": str(data.get("manual_secret") or "").strip(),
+        "manual_secret_file": str(data.get("manual_secret_file") or "").strip(),
+        "manual_secret_configured": parse_bool_value(data.get("manual_secret_configured"), False),
+        "auth_json_path": str(data.get("auth_json_path") or "").strip(),
+        "auth_json_key": str(data.get("auth_json_key") or "").strip(),
+        "request_history_limit": parse_int_value(data.get("request_history_limit")),
+        "model_remap": str(data.get("model_remap") or "").strip(),
+        "reasoning_equals": normalize_editor_integer_list(data.get("reasoning_equals")),
+        "retryable_status_codes": normalize_editor_integer_list(data.get("retryable_status_codes")),
+        "retryable_error_messages": normalize_editor_phrase_list(data.get("retryable_error_messages")),
+        "upstream_fetch_retry_attempts": parse_int_value(data.get("upstream_fetch_retry_attempts")),
+        "upstream_fetch_retry_backoff_ms": parse_int_value(data.get("upstream_fetch_retry_backoff_ms")),
+        "endpoints": normalize_editor_string_list(data.get("endpoints")),
+    }
+
+
+def resolve_editor_command() -> list[str]:
+    configured = (os.environ.get("VISUAL") or os.environ.get("EDITOR") or "").strip()
+    if configured:
+        return shlex.split(configured)
+    for candidate in ("nano", "vim", "vi"):
+        if shutil.which(candidate):
+            return [candidate]
+    raise RuntimeError("no editor found; set $VISUAL or $EDITOR")
+
+
+def edit_profile_payload_with_editor(row: dict[str, Any]) -> dict[str, Any] | None:
+    original_text = profile_editor_document(row)
+    editor_command = resolve_editor_command()
+    fd, raw_path = tempfile.mkstemp(prefix=f"codex-retry-profile-{row.get('name') or 'profile'}-", suffix=".toml")
+    os.close(fd)
+    temp_path = Path(raw_path)
+    temp_path.write_text(original_text, encoding="utf-8")
+    try:
+        result = subprocess.run(editor_command + [str(temp_path)], check=False)
+        edited_text = temp_path.read_text(encoding="utf-8")
+        if edited_text == original_text:
+            return None
+        payload = profile_payload_from_editor_text(edited_text)
+        if result.returncode != 0:
+            raise RuntimeError(f"editor exited with status {result.returncode}")
+        return payload
+    except Exception as exc:
+        raise RuntimeError(f"{exc}; kept draft at {temp_path}") from exc
+    finally:
+        if temp_path.exists():
+            try:
+                if temp_path.read_text(encoding="utf-8") == original_text:
+                    temp_path.unlink()
+            except Exception:
+                pass
+
+
 def action_url(api_url: str, suffix: str) -> str:
     base = api_url.rstrip("/")
     return f"{base}{suffix}"
@@ -785,6 +1092,12 @@ def run_textual(api_url: str, status_url: str, refresh_seconds: int, timeout: in
             ("2", "show_requests", "Requests"),
             ("3", "show_logs", "Logs"),
             ("4", "show_profiles", "Profiles"),
+            ("e", "edit_profile", "Edit Profile"),
+            ("p", "probe_profile", "Probe Profile"),
+            ("s", "switch_profile", "Switch Profile"),
+            ("w", "save_active_profile", "Save Profile"),
+            ("d", "delete_profile", "Delete Profile"),
+            ("u", "open_upstream", "Open Upstream"),
             ("/", "focus_filter", "Filter"),
         ]
 
@@ -831,7 +1144,7 @@ def run_textual(api_url: str, status_url: str, refresh_seconds: int, timeout: in
             requests = self.query_one("#requests_table", DataTable)
             requests.cursor_type = "row"
             requests.zebra_stripes = True
-            requests.add_columns("Seq", "ID", "Status", "Path", "Model", "Reasoning", "Req", "Resp", "Chunks", "First", "Duration", "Updated", "Note")
+            requests.add_columns("Seq", "ID", "Thread", "Status", "Path", "Model", "Reasoning", "Usage", "Resp", "Chunks", "First", "Duration", "Updated", "Note")
 
             logs = self.query_one("#logs_table", DataTable)
             logs.cursor_type = "row"
@@ -857,7 +1170,7 @@ def run_textual(api_url: str, status_url: str, refresh_seconds: int, timeout: in
             self.query_one("#logs_table").display = active_table == "logs"
             self.query_one("#profiles_table").display = active_table == "profiles"
             self.query_one("#controls", Static).update(
-                "views: 1 overview | 2 requests | 3 logs | 4 profiles | / filter | r refresh | enter action | n newest request | p probe | s switch | w save | d delete | u open upstream"
+                "views: 1 overview | 2 requests | 3 logs | 4 profiles | / filter | r refresh | e edit | p probe | s switch | w save | d delete | u open upstream"
             )
             self._focus_table()
             self._update_table_headers()
@@ -1005,12 +1318,13 @@ def run_textual(api_url: str, status_url: str, refresh_seconds: int, timeout: in
                 self.request_by_key[key] = row
                 table.add_row(
                     str(row["seq"]),
-                    short_text(row["request_id"], 18) or "-",
+                    short_text(primary_request_id(row), 18) or "-",
+                    short_text(row["thread_id"], 18) or "-",
                     f"{status_symbol(row)} {row.get('status_code') or '-'}",
                     short_text(row["path"], 22),
                     short_text(row["model"] or row["requested_model"] or row["forwarded_model"], 16),
                     row["reasoning_tokens"] if row["reasoning_tokens"] is not None else "-",
-                    format_bytes(row["request_body_bytes"]),
+                    short_text(request_usage_summary(row), 24),
                     format_bytes(row["response_bytes_received"]),
                     request_chunk_progress(row),
                     format_ms(row["first_response_delay_ms"]),
@@ -1101,6 +1415,26 @@ def run_textual(api_url: str, status_url: str, refresh_seconds: int, timeout: in
                 return
             open_url(upstream)
 
+        def action_edit_profile(self) -> None:
+            if self.current_view != "profiles":
+                self._set_status("switch to profiles view first")
+                return
+            row = self._selected_profile_row()
+            if not row:
+                self._set_status("no profile selected")
+                return
+            try:
+                with self.suspend():
+                    payload = edit_profile_payload_with_editor(row)
+                if payload is None:
+                    self._set_status(f"edit {row['name']} cancelled (no changes)")
+                    return
+                status, result = post_json(build_api_url(gateway_base_url, "/api/profiles"), timeout, payload)
+                self._set_status(f"edit {payload.get('name') or row['name']} -> {status} | {short_text(result, 96)}")
+                self.refresh_data(refresh=True)
+            except Exception as exc:
+                self._set_status(f"edit error: {exc}")
+
         def action_probe_profile(self) -> None:
             row = self._selected_profile_row()
             if not row:
@@ -1134,22 +1468,7 @@ def run_textual(api_url: str, status_url: str, refresh_seconds: int, timeout: in
             row = self._selected_profile_row()
             if not row:
                 return
-            raw = row.get("raw") if isinstance(row.get("raw"), dict) else {}
-            form = raw.get("form") if isinstance(raw.get("form"), dict) else {}
-            payload = {
-                "name": row["name"],
-                "listen_host": form.get("listen_host") or row["listen_host"],
-                "listen_port": as_int(form.get("listen_port") or row["listen_port"]),
-                "upstream_base_url": form.get("upstream_base_url") or row["upstream_base_url"],
-                "upstream_auth_mode": form.get("auth_mode") or row["auth_mode"],
-                "upstream_auth_env": form.get("auth_env") or "",
-                "upstream_auth_file": form.get("auth_file") or "",
-                "upstream_auth_json_path": form.get("auth_json_path") or "",
-                "upstream_auth_json_key": form.get("auth_json_key") or "",
-                "request_history_limit": as_int(form.get("request_history_limit") or row["request_history_limit"]),
-                "model_remap": form.get("model_remap") or "",
-                "reasoning_equals": form.get("reasoning_equals") or "",
-            }
+            payload = profile_payload_from_row(row)
             try:
                 status, result = post_json(build_api_url(gateway_base_url, "/api/profiles"), timeout, payload)
                 self._set_status(f"save profile -> {status} | {short_text(result, 96)}")
@@ -1253,9 +1572,9 @@ def print_once(snapshot: dict[str, Any], filter_text: str = "") -> None:
     print("requests")
     for row in snapshot.get("requests") or []:
         print(
-            f"{row['seq']:>6} {short_text(row['request_id'], 16):<16} {status_symbol(row)} {row.get('status_code') or '-':<4} "
+            f"{row['seq']:>6} {short_text(primary_request_id(row), 16):<16} {short_text(row.get('thread_id'), 14):<14} {status_symbol(row)} {row.get('status_code') or '-':<4} "
             f"{short_text(row['path'], 20):<20} {short_text(row['model'] or row['requested_model'] or row['forwarded_model'], 16):<16} "
-            f"{request_chunk_progress(row):<14} {format_ms(row['duration_ms']):<8} {request_updated_elapsed(row):<8} "
+            f"{short_text(request_usage_summary(row), 26):<26} {request_chunk_progress(row):<14} {format_ms(row['duration_ms']):<8} {request_updated_elapsed(row):<8} "
             f"{format_count(row['upstream_attempt_count'] or 0):<4} {short_text(row['error'], 30)}"
         )
 

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "codex-retry-gateway-tui"
-version = "0.1.2"
+version = "0.1.3"
 description = "Terminal UI for codex-retry-gateway monitoring and control"
 readme = "README.md"
 requires-python = ">=3.11"

tests/test_payload.py

@@ -35,12 +35,14 @@ class CodexRetryGatewayTUITests(unittest.TestCase):
         mod = load_module()
         payload = {
             "entries": [
-                {"seq": 1, "request_id": "req_a", "path": "/responses"},
-                {"seq": 2, "request_id": "req_b", "path": "/v1/responses"},
+                {"seq": 1, "request_id": "req_a", "response_id": "resp_a", "thread_id": "thread_a", "path": "/responses"},
+                {"seq": 2, "request_id": "req_b", "response_id": "resp_b", "thread_id": "thread_b", "path": "/v1/responses"},
             ]
         }
         rows = mod.normalize_request_rows(payload)
         self.assertEqual([row["request_id"] for row in rows], ["req_b", "req_a"])
+        self.assertEqual(rows[0]["response_id"], "resp_b")
+        self.assertEqual(rows[0]["thread_id"], "thread_b")
 
     def test_profile_rows_show_active_first(self) -> None:
         mod = load_module()
@@ -75,6 +77,8 @@ class CodexRetryGatewayTUITests(unittest.TestCase):
         row = {
             "seq": 1,
             "request_id": "r",
+            "response_id": "resp_1",
+            "thread_id": "thread_1",
             "method": "POST",
             "path": "/responses",
             "status_code": 200,
@@ -90,10 +94,50 @@ class CodexRetryGatewayTUITests(unittest.TestCase):
             "finished_at": "2026-06-30T00:00:01Z",
             "usage_last_updated_at": "2026-06-30T00:00:01Z",
         }
+        self.assertEqual(mod.primary_request_id(row), "resp_1")
+        self.assertIn("thread thread_1", mod.render_request_detail(row))
         self.assertIn("updated 1.0s", mod.render_request_detail(row))
         self.assertEqual(mod.request_updated_elapsed(row), "1.0s")
         self.assertEqual(mod.request_chunk_progress(row), "7 / 4.1KB")
 
+    def test_request_usage_summary_shows_cached_ratio(self) -> None:
+        mod = load_module()
+        row = {
+            "input_tokens": 100,
+            "output_tokens": 40,
+            "cached_tokens": 25,
+        }
+        self.assertEqual(mod.request_usage_summary(row), "in 75 | out 40 | cached 25 (25%)")
+
+    def test_profile_editor_payload_round_trip(self) -> None:
+        mod = load_module()
+        row = {
+            "name": "pc",
+            "listen_host": "100.115.235.115",
+            "listen_port": "4610",
+            "upstream_base_url": "https://example.com/v1",
+            "auth_mode": "passthrough",
+            "auth_json_key": "OPENAI_API_KEY",
+            "request_history_limit": 0,
+            "model_remap": "a=b",
+            "reasoning_equals": [516, 1034],
+            "raw": {
+                "form": {
+                    "retryable_error_messages": [
+                        "Selected model is at capacity. Please try a different model.",
+                        "stream disconnected before completion: Concurrency limit exceeded for account, please retry later",
+                    ],
+                }
+            },
+        }
+        text = mod.profile_editor_document(row)
+        payload = mod.profile_payload_from_editor_text(text)
+        self.assertEqual(payload["name"], "pc")
+        self.assertIn(
+            "stream disconnected before completion: Concurrency limit exceeded for account, please retry later",
+            payload["retryable_error_messages"],
+        )
+
     def test_default_api_url_discovers_gateway_state(self) -> None:
         mod = load_module()
         with tempfile.TemporaryDirectory() as tmpdir: