Pablan, as it stands

Self-hosted knowledge management for SMEs: a split-screen Markdown editor
whose sections an LLM refines while you write, and RAG question answering
over the documents that result. FastAPI + Postgres/pgvector on the back,
SvelteKit on the front, everything OpenAI-compatible and self-hostable.

Squashed into a single commit; the development history stays local.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CA43ZJda8Rbp2hKXNy8f6b

backend/app/log.py

@@ -0,0 +1,103 @@
+"""Structured JSON logging (stdlib only).
+
+Logging policy: log lines carry metadata only — never
+prompts, LLM responses, user messages or document text. Exceptions are
+reduced to their type plus a sanitized message; SQLAlchemy statement/param
+dumps are stripped because parameters can contain user content.
+"""
+
+import json
+import logging
+import sys
+from contextvars import ContextVar
+from datetime import UTC, datetime
+from typing import Any
+
+from app.config import get_settings
+
+correlation_id: ContextVar[str | None] = ContextVar("correlation_id", default=None)
+conversation_id: ContextVar[str | None] = ContextVar("conversation_id", default=None)
+
+# LogRecord attributes that are not user-supplied extras.
+_STANDARD_ATTRS = frozenset(
+    {
+        "args",
+        "asctime",
+        "created",
+        "exc_info",
+        "exc_text",
+        "filename",
+        "funcName",
+        "levelname",
+        "levelno",
+        "lineno",
+        "message",
+        "module",
+        "msecs",
+        "msg",
+        "name",
+        "pathname",
+        "process",
+        "processName",
+        "relativeCreated",
+        "stack_info",
+        "taskName",
+        "thread",
+        "threadName",
+    }
+)
+
+
+def safe_error(exc: BaseException, limit: int = 300) -> str:
+    """Exception text that is safe to log or persist (no content leaks).
+
+    SQLAlchemy appends "[SQL: ...] [parameters: (...)]" to its messages;
+    parameters can contain user content, so everything from "[SQL" on is cut.
+    """
+    text = str(exc).split("[SQL", 1)[0].strip()
+    return f"{type(exc).__name__}: {text[:limit]}"
+
+
+class JsonFormatter(logging.Formatter):
+    def format(self, record: logging.LogRecord) -> str:
+        payload: dict[str, Any] = {
+            "ts": datetime.fromtimestamp(record.created, tz=UTC).isoformat(
+                timespec="milliseconds"
+            ),
+            "level": record.levelname,
+            "logger": record.name,
+            "message": record.getMessage(),
+        }
+        cid = correlation_id.get()
+        if cid:
+            payload["correlation_id"] = cid
+        conv = conversation_id.get()
+        if conv:
+            payload["conversation_id"] = conv
+        for key, value in record.__dict__.items():
+            if key not in _STANDARD_ATTRS and not key.startswith("_"):
+                payload[key] = value
+        if record.exc_info and record.exc_info[1] is not None:
+            payload["error"] = safe_error(record.exc_info[1])
+        return json.dumps(payload, default=str)
+
+
+# These third-party loggers dump request/response bodies at DEBUG — with
+# prompts and user content in them, so they stay capped at INFO unless
+# content debug logging is explicitly enabled.
+_CONTENT_DEBUG_LOGGERS = ("openai", "httpx", "httpcore")
+
+
+def apply_content_log_guard() -> None:
+    level = logging.DEBUG if get_settings().debug_log_prompts else logging.INFO
+    for name in _CONTENT_DEBUG_LOGGERS:
+        logging.getLogger(name).setLevel(level)
+
+
+def setup_logging() -> None:
+    handler = logging.StreamHandler(sys.stdout)
+    handler.setFormatter(JsonFormatter())
+    root = logging.getLogger()
+    root.handlers = [handler]
+    root.setLevel(get_settings().log_level.upper())
+    apply_content_log_guard()