Allow agent_query to Use an Existing Search Index Without Rebuilding (#847)

Co-authored-by: James Braza <[email protected]>

Author: psl-schaefer

paperqa/agents/main.py

@@ -120,7 +120,7 @@ async def run_agent(
     # Build the index once here, and then all tools won't need to rebuild it
     # only build if the a search tool is requested
     if PaperSearch.TOOL_FN_NAME in (settings.agent.tool_names or DEFAULT_TOOL_NAMES):
-        await get_directory_index(settings=settings)
+        await get_directory_index(settings=settings, build=settings.agent.rebuild_index)
 
     if isinstance(agent_type, str) and agent_type.lower() == FAKE_AGENT_TYPE:
         session, agent_status = await run_fake_agent(

paperqa/settings.py

@@ -523,6 +523,14 @@ class AgentSettings(BaseModel):
     )
     index: IndexSettings = Field(default_factory=IndexSettings)
 
+    rebuild_index: bool = Field(
+        default=True,
+        description=(
+            "Flag to rebuild the index at the start of agent runners, default is True"
+            " for CLI users to ensure all source PDFs are pulled in."
+        ),
+    )
+
     callbacks: Mapping[str, Sequence[Callable[[_EnvironmentState], Any]]] = Field(
         default_factory=dict,
         description="""

tests/test_agents.py

@@ -1080,6 +1080,20 @@ async def test_search_pagination(agent_test_settings: Settings) -> None:
     ), "Second page should match second slice of all results"
 
 
+@pytest.mark.asyncio
+async def test_empty_index_without_index_rebuild(agent_test_settings: Settings):
+    """Test that empty index and `rebuild_index=False` lead to a RuntimeError."""
+    agent_test_settings.agent = AgentSettings(index=IndexSettings())  # empty index
+    agent_test_settings.agent.rebuild_index = False
+    with pytest.raises(RuntimeError, match=r"Index .* was empty, please rebuild it."):
+        await agent_query(
+            query="Are COVID-19 vaccines effective?",
+            settings=agent_test_settings,
+            agent_type=FAKE_AGENT_TYPE,
+            force_index_rebuild=False,
+        )
+
+
 class TestClinicalTrialSearchTool:
     @pytest.mark.asyncio
     async def test_continuation(self) -> None: