mempalace.auto_query.runner
Source: mempalace/auto_query/runner.py
Auto-query runner — chains signal extraction, routing, and formatting.
This module is the single entry point for the auto-query pipeline. It is called from the UserPromptSubmit hook (via __main__.py) or directly from tests.
Pipeline::
load config → check enabled/mode → extract signals → pick tool
→ execute MCP call → format injection → log decision
The MCP call is made via _call_mcp(), which delegates to the palace-daemon HTTP proxy (/mcp endpoint). When the daemon is not reachable, the call is skipped and a dry-run decision is logged.
Classes
class AutoQueryResult
Result of a single auto-query pipeline run.
__init__
python
def __init__(self, injection = None, decision = None, tool_call = None, mcp_result = None)Functions
run_auto_query
python
def run_auto_query(prompt, session_id, turn, project_wing = '', known_wings = None, known_entities = None, has_recent_drawers = False, config = None, queried_entities = None, log_dir = None)Run the auto-query pipeline for a single user turn.
Returns an AutoQueryResult with the injection text (or None), the decision record, and the raw MCP result.