community-detection

Community Detection Workflow

Run a complete community detection and visualization workflow on the current Gephi graph.

Tell the user what you're doing at each step — narrate briefly before each tool call.

Steps

1. Health check: Call gephi_health_check. If it fails, tell the user to start Gephi and stop.

2. Graph info: Call gephi_get_project_info. Tell the user the node/edge counts.

3. Ask which method (skip if $ARGUMENTS names one or the user already said). One question, with the trade-off stated plainly:

   • Louvain (Gephi's built-in Modularity; Blondel et al. 2008): maximizes modularity by greedy local moves. Fast and familiar; the default.
   • Leiden (Traag, Waltman, and van Eck 2019): the same objective with a refinement step that guarantees every community is internally connected and converges more reliably. Its partitions can be more uneven in size. Requires the CWTS Leiden plugin in Gephi; check gephi_list_statistics for "Leiden algorithm" before offering it as available.
   • Stochastic block model inference (Peixoto 2019): fits a generative model of the edges and selects the partition that best explains them, with model selection that returns a single block when the data support no structure. Modularity maximization has no such check and returns a partition for any graph, including a random one. SBM inference is not implemented in Gephi; if the user wants it, say so and point to graph-tool (minimize_blockmodel_dl) outside this workflow.

   Frame the choice as: modularity maximization gives a partition that describes how the observed edges cluster; SBM inference tests whether a block structure is supported at all. Cite the papers in the caption when the map is publication-bound.

4. Compute communities:

   • Louvain: call gephi_compute_modularity with resolution $ARGUMENTS resolution (default 1.0). Note Gephi's resolution runs opposite to the gamma convention in most papers: raising it merges communities.
   • Leiden: call gephi_run_statistic with name="Leiden algorithm" and params={"algorithm": "Leiden", "qualityFunction": "Modularity", "resolution": <resolution>}; the result column is what the plugin reports (check gephi_get_columns and use that name in step 6). Tell the user: "Running community detection..." then report the modularity score and number of communities.