Fix feedback log path: use absolute path so agent always finds it

The feedback log (.docetl_feedback.log) used a relative path, so if
the server and the agent had different cwds, the agent's Monitor would
watch the wrong file and never see feedback.

Now FeedbackStore resolves the log path to an absolute path at init.
The pipeline prints [FEEDBACK_LOG] <path> to stdout so the agent knows
where to monitor. Added /feedback/log_path endpoint for discovery.

https://claude.ai/code/session_013WUwghjHVh6r5mRHEMuhGP

Author: claude

.claude/skills/docetl/SKILL.md

@@ -585,6 +585,8 @@ You can also start it manually with `docetl serve` if you want it running before
 Monitor(command="tail -F .docetl_feedback.log", description="Watch for human feedback")
 ```
 
+If the pipeline prints a `[FEEDBACK_LOG] /absolute/path` line, use that path instead — the log may be in a different directory than your cwd. You can also query the server: `curl -s http://localhost:<PORT>/feedback/log_path`.
+
 This is MANDATORY. Without the Monitor, you will not see feedback until the pipeline finishes. The Monitor sends you a notification the instant the human submits feedback.
 
 Each feedback event appears as a line:

docetl/tui/web_reporter.py

@@ -34,6 +34,19 @@
 _FEEDBACK_LOG = ".docetl_feedback.log"
 
 
+def _resolve_feedback_log() -> str:
+    """Return the absolute path to the feedback log.
+
+    When a port file exists, write the log next to it so the agent always
+    knows where to find it regardless of cwd.  Falls back to cwd.
+    """
+    try:
+        port_path = os.path.abspath(_PORT_FILE)
+        return os.path.join(os.path.dirname(port_path), _FEEDBACK_LOG)
+    except Exception:
+        return os.path.abspath(_FEEDBACK_LOG)
+
+
 # ---------------------------------------------------------------------------
 # Feedback store
 # ---------------------------------------------------------------------------
@@ -47,13 +60,13 @@ def __init__(self):
         self.kill_reason: str | None = None
         self._agent_messages: list[dict] = []
         self._agent_msg_counter = 0
-        # Truncate log at start of each session
-        with open(_FEEDBACK_LOG, "w") as f:
+        self._log_path = _resolve_feedback_log()
+        with open(self._log_path, "w") as f:
             pass
 
     def _log(self, line: str):
         try:
-            with open(_FEEDBACK_LOG, "a") as f:
+            with open(self._log_path, "a") as f:
                 f.write(line + "\n")
                 f.flush()
         except OSError:
@@ -312,6 +325,8 @@ def do_GET(self):
                 self._json_response(broadcaster._last_event or {})
             elif self.path == "/health":
                 self._json_response({"ok": True})
+            elif self.path == "/feedback/log_path":
+                self._json_response({"path": feedback._log_path})
             else:
                 self.send_error(404)
 
@@ -2241,6 +2256,8 @@ def _run_with_remote_server(runner: "DSLRunner", tracker: ProgressTracker, port:
     runner.console.log(
         f"[bold blue]Using persistent server:[/bold blue] http://localhost:{port}"
     )
+    log_path = _resolve_feedback_log()
+    print(f"[FEEDBACK_LOG] {log_path}", flush=True)
 
     # Reset the server UI for this new run
     try: