run-llama/llama_index · error · ImportError

Please install networkx to visualize the graph: `pip install

Error message

Please install networkx to visualize the graph: `pip install networkx`

What it means

KnowledgeGraphIndex.get_networkx_graph() lazily imports networkx and, if missing, raises ImportError suggesting 'pip install networkx'. The method builds an nx.Graph from the graph store's rel_map (subjects, depth 1, up to `limit`) for visualization; networkx is an optional dependency only needed on this path.

Source

Thrown at llama-index-core/llama_index/core/indices/knowledge_graph/base.py:357

            all_ref_doc_info[ref_node.node_id] = ref_doc_info
        return all_ref_doc_info

    def get_networkx_graph(self, limit: int = 100) -> Any:
        """
        Get networkx representation of the graph structure.

        Args:
            limit (int): Number of starting nodes to be included in the graph.

        NOTE: This function requires networkx to be installed.
        NOTE: This is a beta feature.

        """
        try:
            import networkx as nx
        except ImportError:
            raise ImportError(
                "Please install networkx to visualize the graph: `pip install networkx`"
            )

        g = nx.Graph()
        subjs = list(self.index_struct.table.keys())

        # add edges
        rel_map = self._graph_store.get_rel_map(subjs=subjs, depth=1, limit=limit)

        added_nodes = set()
        for keyword in rel_map:
            for path in rel_map[keyword]:
                subj = keyword
                for i in range(0, len(path), 2):
                    if i + 2 >= len(path):
                        break

                    if subj not in added_nodes:

View on GitHub (pinned to afd0fef371)

Solutions

pip install networkx in the environment running the visualization.
Add networkx to an optional dev/visualization extras group so production images stay slim.
Guard the call: import networkx yourself first (or try/except ImportError) and show a graceful message instead of crashing the app.

Example fix

# before
g = kg_index.get_networkx_graph()  # ImportError without networkx

# after
pip install networkx
g = kg_index.get_networkx_graph()

Defensive patterns

Strategy: validation

Validate before calling

import importlib.util

if importlib.util.find_spec("networkx") is None:
    raise SystemExit("networkx required for graph visualization: pip install networkx")
g = kg_index.get_networkx_graph()

Try / catch

try:
    g = kg_index.get_networkx_graph()
except ImportError:
    g = None  # visualization unavailable; degrade gracefully

Prevention

Keep networkx in a 'viz' extras group and install it where graphs are inspected.
Guard debug/visualization endpoints with an import check.
Never let an optional visualization crash the main query path.

When it happens

Trigger: Calling kg_index.get_networkx_graph() or get_networkx_graph(limit=N) without networkx installed; visualization/inspection tooling over a KnowledgeGraphIndex on a base install of llama-index-core.

Common situations: Notebook exploration of a KG index on a slim environment; adding a 'view graph' debug endpoint to a service deployed from a minimal Docker image; local dev working (networkx present globally) but CI/prod failing.

Related errors

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/f6581fe793bd9d12. Report an issue: GitHub.