diff --git a/graphify/__main__.py b/graphify/__main__.py
index d23059d..fb28cf4 100644
--- a/graphify/__main__.py
+++ b/graphify/__main__.py
@@ -1546,7 +1546,15 @@ def main() -> None:
         cohesion = score_all(G, communities)
         gods = god_nodes(G)
         surprises = surprising_connections(G, communities)
-        labels = {cid: f"Community {cid}" for cid in communities}
+        out = watch_path / "graphify-out"
+        labels_path = out / ".graphify_labels.json"
+        if labels_path.exists():
+            try:
+                labels = {int(k): v for k, v in json.loads(labels_path.read_text(encoding="utf-8")).items()}
+            except Exception:
+                labels = {cid: f"Community {cid}" for cid in communities}
+        else:
+            labels = {cid: f"Community {cid}" for cid in communities}
         questions = suggest_questions(G, communities, labels)
         tokens = {"input": 0, "output": 0}
         from graphify.export import _git_head as _gh
@@ -1555,9 +1563,9 @@ def main() -> None:
                           {"warning": "cluster-only mode — file stats not available"},
                           tokens, str(watch_path), suggested_questions=questions,
                           min_community_size=min_community_size, built_at_commit=_commit)
-        out = watch_path / "graphify-out"
         (out / "GRAPH_REPORT.md").write_text(report, encoding="utf-8")
         to_json(G, communities, str(out / "graph.json"))
+        labels_path.write_text(json.dumps({str(k): v for k, v in labels.items()}, ensure_ascii=False), encoding="utf-8")
 
         # Mirror watch.py pattern: gate to_html so core outputs (graph.json +
         # GRAPH_REPORT.md) always land. Honor --no-viz explicitly; otherwise
@@ -1872,6 +1880,13 @@ def main() -> None:
         elif subcmd == "wiki":
             from graphify.wiki import to_wiki as _to_wiki
             from graphify.analyze import god_nodes as _god_nodes
+            if not communities:
+                print(
+                    "error: .graphify_analysis.json is missing or empty — refusing to export wiki to prevent data loss.\n"
+                    "Run `graphify extract .` (or `graphify cluster-only .`) to regenerate community data first.",
+                    file=sys.stderr,
+                )
+                sys.exit(1)
             if not gods_data:
                 gods_data = _god_nodes(G)
             n = _to_wiki(G, communities, str(out_dir / "wiki"),
diff --git a/graphify/llm.py b/graphify/llm.py
index 07c8455..77fcbb7 100644
--- a/graphify/llm.py
+++ b/graphify/llm.py
@@ -144,9 +144,10 @@ def _call_openai_compat(
     try:
         from openai import OpenAI
     except ImportError as exc:
+        pkg_hint = "graphifyy[kimi]" if backend == "kimi" else "openai"
         raise ImportError(
-            "Kimi/OpenAI-compatible extraction requires the openai package. "
-            "Run: pip install openai"
+            f"{'Ollama' if backend == 'ollama' else 'Kimi'}/OpenAI-compatible extraction requires the openai package. "
+            f"Run: pip install {pkg_hint}"
         ) from exc
 
     client = OpenAI(api_key=api_key, base_url=base_url)
diff --git a/graphify/wiki.py b/graphify/wiki.py
index 8cc845f..973ed89 100644
--- a/graphify/wiki.py
+++ b/graphify/wiki.py
@@ -196,6 +196,12 @@ def to_wiki(
     out = Path(output_dir)
     out.mkdir(parents=True, exist_ok=True)
 
+    if not communities:
+        raise ValueError(
+            "communities dict is empty — refusing to clear wiki/. "
+            "Run `graphify extract .` or `graphify cluster-only .` first."
+        )
+
     # Clear stale .md files from previous runs to prevent orphan accumulation.
     # Community labels are LLM-generated (per skill.md Step 5) and non-deterministic
     # across runs — the same conceptual community may be named differently each time
diff --git a/pyproject.toml b/pyproject.toml
index 378a3c9..837c9b1 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -54,6 +54,7 @@ leiden = ["graspologic; python_version < '3.13'"]
 office = ["python-docx", "openpyxl"]
 video = ["faster-whisper", "yt-dlp"]
 kimi = ["openai", "tiktoken"]
+ollama = ["openai"]
 sql = ["tree-sitter-sql"]
 all = ["mcp", "neo4j", "pypdf", "markdownify", "watchdog", "graspologic; python_version < '3.13'", "python-docx", "openpyxl", "faster-whisper", "yt-dlp", "matplotlib", "openai", "tiktoken", "tree-sitter-sql"]