secureagentrag-api / utils /conversation_store.py
LeomordKaly's picture
deploy: phase 3 BYOK backend (Dockerfile.hf, FastAPI on 7860)
09ed8ca verified
"""Persistent conversation history storage using JSONL files.
Provides durable, file-backed conversation storage that survives server restarts.
Conversations are keyed by thread_id and stored in daily-organized JSONL files
for easy rotation and backup.
"""
from __future__ import annotations
import json
import re
import time
from dataclasses import dataclass, field
from datetime import UTC, datetime
from pathlib import Path
from typing import Any
from config.settings import settings
from utils.logging import get_logger
logger = get_logger(__name__)
# Legacy citation markers used by earlier synthesizer prompts. Rendered as
# literal "[[N]]" in the UI, which looked like a bug. The current prompt
# emits `[N]`. We rewrite legacy strings at load time so historic threads
# display the new format without re-running the LLM.
_LEGACY_CITATION_RE = re.compile(r"\[\[(\d+)\]\]")
def _normalize_citations(text: str) -> str:
"""Convert legacy ``[[N]]`` markers to ``[N]`` in stored content."""
if not text:
return text
return _LEGACY_CITATION_RE.sub(r"[\1]", text)
# Default retention: 90 days for conversation threads
DEFAULT_RETENTION_DAYS: int = 90
@dataclass
class ConversationMessage:
"""A single message in a conversation thread.
Attributes:
role: Message role — "user", "assistant", "blocked", or "system".
content: Message text content.
timestamp: UTC timestamp of the message.
metadata: Optional metadata (citations, confidence, routing info, etc.).
"""
role: str
content: str
timestamp: str = field(default_factory=lambda: datetime.now(UTC).isoformat())
metadata: dict[str, Any] = field(default_factory=dict)
@dataclass
class ConversationThread:
"""A complete conversation thread with metadata.
Attributes:
thread_id: Unique thread identifier.
user_id: Owner of the conversation.
org_id: Organization identifier.
messages: List of messages in chronological order.
created_at: Thread creation timestamp.
updated_at: Last activity timestamp.
metadata: Thread-level metadata (model, mode, etc.).
"""
thread_id: str
user_id: str
org_id: str
messages: list[ConversationMessage] = field(default_factory=list)
created_at: str = field(default_factory=lambda: datetime.now(UTC).isoformat())
updated_at: str = field(default_factory=lambda: datetime.now(UTC).isoformat())
metadata: dict[str, Any] = field(default_factory=dict)
class ConversationStore:
"""Persistent conversation storage backed by JSONL files.
Stores conversations in a directory structure organized by date:
conversations/YYYY-MM-DD/<thread_id>.jsonl
Each file contains the full thread as a single JSON line for atomic writes.
Args:
base_dir: Base directory for conversation storage.
Defaults to "conversations" in the project root.
"""
def __init__(self, base_dir: str | None = None) -> None:
"""Initialize the conversation store.
Args:
base_dir: Base directory for conversation files.
"""
self._base_dir = Path(base_dir or settings.conversation_dir)
self._base_dir.mkdir(parents=True, exist_ok=True)
logger.info("conversation_store_initialized", path=str(self._base_dir))
def _thread_path(self, thread_id: str) -> Path:
"""Compute the file path for a given thread_id.
Uses today's date for new files; existing files are found via search.
Args:
thread_id: The thread identifier.
Returns:
Path to the thread's JSONL file.
"""
today = datetime.now(UTC).strftime("%Y-%m-%d")
date_dir = self._base_dir / today
date_dir.mkdir(parents=True, exist_ok=True)
return date_dir / f"{thread_id}.jsonl"
def _find_thread_path(self, thread_id: str) -> Path | None:
"""Search for an existing thread file across all date directories.
Args:
thread_id: The thread identifier to search for.
Returns:
Path to the existing file, or None if not found.
"""
for date_dir in self._base_dir.iterdir():
if date_dir.is_dir():
candidate = date_dir / f"{thread_id}.jsonl"
if candidate.exists():
return candidate
return None
def save_thread(self, thread: ConversationThread) -> bool:
"""Persist a conversation thread to disk.
Args:
thread: The conversation thread to save.
Returns:
True if saved successfully, False otherwise.
"""
try:
# Update timestamp
thread.updated_at = datetime.now(UTC).isoformat()
# Convert to dict
data = {
"thread_id": thread.thread_id,
"user_id": thread.user_id,
"org_id": thread.org_id,
"messages": [
{
"role": msg.role,
"content": msg.content,
"timestamp": msg.timestamp,
"metadata": msg.metadata,
}
for msg in thread.messages
],
"created_at": thread.created_at,
"updated_at": thread.updated_at,
"metadata": thread.metadata,
}
# Check if thread already exists in a different date directory
existing_path = self._find_thread_path(thread.thread_id)
file_path = existing_path or self._thread_path(thread.thread_id)
# Atomic write: write to temp then rename
temp_path = file_path.with_suffix(".tmp")
with open(temp_path, "w", encoding="utf-8") as f:
json.dump(data, f, ensure_ascii=False)
f.write("\n")
temp_path.replace(file_path)
logger.info(
"thread_saved",
thread_id=thread.thread_id,
message_count=len(thread.messages),
path=str(file_path),
)
return True
except Exception as exc:
logger.error("thread_save_failed", thread_id=thread.thread_id, error=str(exc))
return False
def load_thread(self, thread_id: str) -> ConversationThread | None:
"""Load a conversation thread from disk.
Args:
thread_id: The thread identifier to load.
Returns:
The loaded ConversationThread, or None if not found.
"""
file_path = self._find_thread_path(thread_id)
if not file_path:
return None
try:
with open(file_path, encoding="utf-8") as f:
line = f.readline().strip()
if not line:
return None
data = json.loads(line)
messages = [
ConversationMessage(
role=msg["role"],
content=_normalize_citations(msg["content"]),
timestamp=msg.get("timestamp", ""),
metadata=msg.get("metadata", {}),
)
for msg in data.get("messages", [])
]
return ConversationThread(
thread_id=data["thread_id"],
user_id=data["user_id"],
org_id=data["org_id"],
messages=messages,
created_at=data.get("created_at", ""),
updated_at=data.get("updated_at", ""),
metadata=data.get("metadata", {}),
)
except Exception as exc:
logger.error("thread_load_failed", thread_id=thread_id, error=str(exc))
return None
def list_threads(
self,
user_id: str | None = None,
org_id: str | None = None,
limit: int = 50,
) -> list[dict[str, Any]]:
"""List conversation threads with optional filtering.
Args:
user_id: Filter by user ID.
org_id: Filter by organization ID.
limit: Maximum number of threads to return.
Returns:
List of thread summary dicts (thread_id, user_id, org_id, message_count, updated_at).
"""
threads: list[dict[str, Any]] = []
for date_dir in sorted(self._base_dir.iterdir(), reverse=True):
if not date_dir.is_dir():
continue
for file_path in sorted(date_dir.glob("*.jsonl"), reverse=True):
try:
with open(file_path, encoding="utf-8") as f:
line = f.readline().strip()
if not line:
continue
data = json.loads(line)
if user_id and data.get("user_id") != user_id:
continue
if org_id and data.get("org_id") != org_id:
continue
threads.append(
{
"thread_id": data["thread_id"],
"user_id": data.get("user_id", ""),
"org_id": data.get("org_id", ""),
"message_count": len(data.get("messages", [])),
"updated_at": data.get("updated_at", ""),
"metadata": data.get("metadata", {}),
}
)
if len(threads) >= limit:
break
except Exception as exc:
logger.warning("thread_list_parse_failed", file=str(file_path), error=str(exc))
if len(threads) >= limit:
break
return threads
def delete_thread(self, thread_id: str) -> bool:
"""Delete a conversation thread from disk.
Args:
thread_id: The thread identifier to delete.
Returns:
True if deleted successfully, False otherwise.
"""
file_path = self._find_thread_path(thread_id)
if not file_path:
return False
try:
file_path.unlink()
logger.info("thread_deleted", thread_id=thread_id, path=str(file_path))
return True
except Exception as exc:
logger.error("thread_delete_failed", thread_id=thread_id, error=str(exc))
return False
def append_message(
self,
thread_id: str,
message: ConversationMessage,
user_id: str = "",
org_id: str = "",
metadata: dict[str, Any] | None = None,
) -> bool:
"""Append a message to an existing thread, creating it if necessary.
Args:
thread_id: The thread identifier.
message: The message to append.
user_id: User ID for new threads.
org_id: Org ID for new threads.
metadata: Thread-level metadata for new threads.
Returns:
True if appended successfully, False otherwise.
"""
thread = self.load_thread(thread_id)
if thread is None:
thread = ConversationThread(
thread_id=thread_id,
user_id=user_id,
org_id=org_id,
metadata=metadata or {},
)
thread.messages.append(message)
return self.save_thread(thread)
def cleanup_old_threads(self, retention_days: int = DEFAULT_RETENTION_DAYS) -> int:
"""Delete conversation threads older than the retention period.
Args:
retention_days: Number of days to retain conversations. Default 90.
Returns:
Number of threads deleted.
"""
cutoff_timestamp = time.time() - (retention_days * 86400)
deleted_count = 0
for date_dir in self._base_dir.iterdir():
if not date_dir.is_dir():
continue
for file_path in date_dir.glob("*.jsonl"):
try:
mtime = file_path.stat().st_mtime
if mtime < cutoff_timestamp:
file_path.unlink()
deleted_count += 1
logger.debug("old_thread_deleted", path=str(file_path))
except Exception as exc:
logger.warning("thread_cleanup_failed", path=str(file_path), error=str(exc))
# Remove empty date directories
try:
if date_dir.exists() and not any(date_dir.iterdir()):
date_dir.rmdir()
except Exception:
pass
logger.info(
"conversation_cleanup_completed", deleted=deleted_count, retention_days=retention_days
)
return deleted_count
# Module-level singleton
conversation_store = ConversationStore()