[eric] agents: strip the temporary TTFT probes (pool drift-diag stays, gated OPENSWARM_POOL_DIAG)

backend/apps/agents/agent_manager.py

@@ -28,7 +28,6 @@ from backend.apps.agents.manager.RunSupport import RunSupport
 from backend.apps.agents.manager.run.handle_run_error import handle_run_error
 from backend.apps.agents.manager.run.TurnRunner import TurnRunner
 from backend.apps.agents.manager.run.RunOptions import RunOptions
-from backend.apps.agents.manager.ttft_probe import ttft_probe
 
 logger = logging.getLogger(__name__)
 
@@ -57,7 +56,6 @@ class AgentManager(SessionLifecycle, SessionPersistence, Messaging, SessionContr
         session = self.sessions.get(session_id)
         if not session:
             return
-        ttft_probe(session_id, "loop_start", fork=fork_session, model=session.model, msgs=len(session.messages))
 
         from backend.apps.agents.providers.registry import get_api_type as p_get_api_type
         p_api = p_get_api_type(session.model)
@@ -65,7 +63,6 @@ class AgentManager(SessionLifecycle, SessionPersistence, Messaging, SessionContr
             prompt, images, context_paths, forced_tools, attached_skills,
             api_type=p_api, model=session.model,
         )
-        ttft_probe(session_id, "prompt_built")
 
         try:
             # SDK presence check: fall to mock mode here, before the options build, so a missing SDK is a clean mock run, not an error card. The real use is in run_options / turn_runner (lazy-imported there).
@@ -98,7 +95,6 @@ class AgentManager(SessionLifecycle, SessionPersistence, Messaging, SessionContr
                 session, session_id, prompt, prompt_content, builtin_perms,
                 selected_browser_ids, selected_app_output_ids, selected_setting_ids,
                 fork_session, p_router_model_id, p_api_type_for_session)
-            ttft_probe(session_id, "options_built")
             resolved_model = p_router_model_id
             api_type = p_api_type_for_session
 

backend/apps/agents/manager/Messaging.py

@@ -17,7 +17,6 @@ from backend.apps.agents.manager.session.session_store import load_session_data
 from backend.apps.agents.manager.session.apply_context_window import apply_context_window
 from backend.apps.agents.manager.prompt.tool_catalog import get_all_tool_names
 from backend.apps.agents.manager.prompt.prompt_context import resolve_mode
-from backend.apps.agents.manager.ttft_probe import ttft_probe
 
 logger = logging.getLogger(__name__)
 
@@ -45,7 +44,6 @@ class Messaging(AgentManagerProtocol):
         client_message_id: Optional[str] = None,
     ):
         """Send a follow-up message to an existing session."""
-        ttft_probe(session_id, "msg_received")
         session = self.sessions.get(session_id)
         if not session:
             data = load_session_data(session_id)

backend/apps/agents/manager/run/TurnRunner.py

@@ -16,7 +16,6 @@ from backend.apps.agents.manager.streaming.state import ThinkingState, TurnState
 from backend.apps.agents.manager.streaming.handle_stream_event import handle_stream_event
 from backend.apps.agents.manager.streaming.handle_assistant_message import handle_assistant_message
 from backend.apps.agents.manager.streaming.handle_result_message import handle_result_message
-from backend.apps.agents.manager.ttft_probe import ttft_probe
 from backend.apps.agents.manager.run.client_pool import (
     acquire_client,
     boot_fingerprint,
@@ -42,7 +41,6 @@ class TurnRunner(AgentManagerProtocol):
                                     global_settings: AppSettings, force_respawn: bool = False) -> None:
         from claude_agent_sdk import query, ClaudeAgentOptions, AssistantMessage, ResultMessage
         from claude_agent_sdk.types import StreamEvent, SystemMessage
-        ttft_probe(session_id, "query_enter")
 
         async def prompt_stream():
             yield {
@@ -94,7 +92,6 @@ class TurnRunner(AgentManagerProtocol):
                             logger.exception("pre-emit thinking pill failed; continuing")
 
                 if turn.first_event:
-                    ttft_probe(session_id, "first_event", type=type(message).__name__)
                     logger.info(f"[MCP-DEBUG] First event received: {type(message).__name__}")
                     turn.first_event = False
 
@@ -113,7 +110,6 @@ class TurnRunner(AgentManagerProtocol):
                         message, session, session_id, turn, thinking, self.live_partial, self.sessions
                     )
                 elif isinstance(message, ResultMessage):
-                    ttft_probe(session_id, "result", chars=turn.assistant_text_chars)
                     await handle_result_message(
                         message, session, session_id, turn, thinking, self.sessions,
                         resolved_model, api_type, global_settings,

backend/apps/agents/manager/run/client_pool.py

@@ -45,7 +45,8 @@ def boot_fingerprint(options_kwargs: Dict, session: AgentSession) -> str:
     frozen = {k: v for k, v in options_kwargs.items() if k not in P_NON_BOOT_KEYS}
     frozen["p_branch"] = session.active_branch_id
     frozen["p_compacted_through"] = session.compacted_through_msg_id
-    if os.environ.get("OSW_TTFT_PROBE") == "1":
+    # Pool diagnostics (OPENSWARM_POOL_DIAG=1): on a respawn, names WHICH boot field drifted; the tool for debugging respawn churn (e.g. the thinking short/long-prompt flip) in the field.
+    if os.environ.get("OPENSWARM_POOL_DIAG") == "1":
         digests = {k: hashlib.sha256(json.dumps(v, sort_keys=True, default=str).encode()).hexdigest()[:10] for k, v in frozen.items()}
         prev = p_last_field_digests.get(session.id)
         if prev is not None:

backend/apps/agents/manager/streaming/handle_stream_event.py

@@ -14,7 +14,6 @@ from backend.apps.agents.core.models import AgentSession
 from backend.apps.agents.core.ws_manager import ws_manager
 from backend.apps.agents.manager.streaming.state import ThinkingState, TurnState
 from backend.apps.agents.manager.streaming.PartialReply import PartialReply
-from backend.apps.agents.manager.ttft_probe import ttft_probe
 
 try:
     from claude_agent_sdk.types import StreamEvent
@@ -45,7 +44,6 @@ async def handle_stream_event(
 
         if block_type == "text":
             if turn.stream_text_msg_id is None:
-                ttft_probe(session_id, "first_text_block")
                 turn.stream_text_msg_id = uuid4().hex
                 await ws_manager.send_to_session(session_id, "agent:stream_start", {
                     "session_id": session_id,
@@ -56,8 +54,6 @@ async def handle_stream_event(
 
         elif block_type == "thinking":
             # Reasoning trace from thinking-capable models (GPT-5.3 Codex, Gemini 3 Pro/Flash, Claude with extended thinking). Rendered as a collapsible "thinking" message in the UI via the existing stream infrastructure, the frontend already handles role="thinking" for the DynamicIsland/agent card rendering.
-            if not thinking.block_starts:
-                ttft_probe(session_id, "first_thinking_block")
             thinking_msg_id = uuid4().hex
             turn.stream_block_index_map[index] = thinking_msg_id
             # Server-stamp start so we can accumulate per-turn elapsed_ms across multiple thinking blocks (think → tool → think → answer turns sum correctly).
@@ -89,8 +85,6 @@ async def handle_stream_event(
 
         if msg_id and delta_type == "text_delta":
             text_chunk = delta.get("text", "")
-            if turn.assistant_text_chars == 0:
-                ttft_probe(session_id, "first_text_delta", chars=len(text_chunk))
             turn.assistant_text_chars += len(text_chunk)
             turn.stream_text_accum += text_chunk
             live_partial[session_id] = PartialReply(

backend/apps/agents/manager/ttft_probe.py

@@ -1,25 +0,0 @@
-"""Temporary time-to-first-token phase probe for the send->first-token A/B sweep. A no-op unless
-OSW_TTFT_PROBE=1, so it never spams a normal run. Strip once the persistent-client work lands."""
-
-import logging
-import os
-import time
-
-from typeguard import typechecked
-
-logger = logging.getLogger(__name__)
-
-P_TTFT_ENABLED = os.environ.get("OSW_TTFT_PROBE") == "1"
-
-
-@typechecked
-def ttft_probe(session_id: str, phase: str, **extra: object) -> None:
-    """One monotonic phase stamp for the TTFT breakdown; the A/B parser reads `phase=<name> mono=<t>`.
-    A no-op unless OSW_TTFT_PROBE=1, and it swallows any error so instrumentation can NEVER break a turn."""
-    if not P_TTFT_ENABLED:
-        return
-    try:
-        tail = " ".join(f"{k}={v}" for k, v in extra.items())
-        logger.warning(f"[TTFT] sid={session_id} phase={phase} mono={time.monotonic():.4f} {tail}".rstrip())
-    except Exception:
-        pass