fix(agents): cite the pages the graph tool read

The graph tool records every page read_entity_pages returns in
retrieved_docs, but agents only ever collected internal_search's. A turn
answered from graph pages therefore emitted no sources: on the multi-hop e2e
run, a question answered from two read_entity_pages calls came back with none.

_search_tool_docs gathers both tools' documents the way the executor caches
them, and both the classic/agentic collector and the research agent's
per-step citations use it. Checked against the real executor and a real
graph: a graph-only turn now cites the pages it read.

docsgpt/agents/research_agent.py

@@ -7,10 +7,7 @@ from typing import Dict, Generator, List, Optional
 from docsgpt.agents.base import BaseAgent
 from docsgpt.agents.tool_executor import ToolExecutor
 from docsgpt.agents.tools.graph_search import add_graph_search_tool
-from docsgpt.agents.tools.internal_search import (
-    INTERNAL_TOOL_ID,
-    add_internal_search_tool,
-)
+from docsgpt.agents.tools.internal_search import add_internal_search_tool
 from docsgpt.agents.tools.wiki import add_wiki_tool
 from docsgpt.agents.tools.think import THINK_TOOL_ENTRY, THINK_TOOL_ID
 from docsgpt.logging import LogContext
@@ -622,12 +619,9 @@ class ResearchAgent(BaseAgent):
         return messages, search_returned_empty
 
     def _collect_step_sources(self):
-        """Collect sources from InternalSearchTool and register with CitationManager."""
-        cache_key = f"internal_search:{INTERNAL_TOOL_ID}:{self.user or ''}"
-        tool = self.tool_executor._loaded_tools.get(cache_key)
-        if tool and hasattr(tool, "retrieved_docs"):
-            for doc in tool.retrieved_docs:
-                self.citations.add(doc)
+        """Register the search tools' docs (internal search and graph pages) with CitationManager."""
+        for doc in self._search_tool_docs():
+            self.citations.add(doc)
 
     # ------------------------------------------------------------------
     # Phase 3: Synthesis