diff --git a/docgen.py b/docgen.py
index 4f8922b..6ebdadb 100644
--- a/docgen.py
+++ b/docgen.py
@@ -121,7 +121,7 @@ Key source files (with their generated documentation):
 
 Project Overview:"""
 
-    return _call_llm(prompt)
+    return _call_llm(prompt, max_tokens=1024)
 
 
 def generate_docs_batch(items: list[tuple[str, str]], doc_fn) -> list[str]:
@@ -149,7 +149,7 @@ def generate_docs_batch(items: list[tuple[str, str]], doc_fn) -> list[str]:
     return results
 
 
-def _call_ollama(prompt: str, retries: int = 3) -> str:
+def _call_ollama(prompt: str, retries: int = 3, max_tokens: int = 256) -> str:
     """Call Ollama API with retries."""
     for attempt in range(retries):
         try:
@@ -161,7 +161,7 @@ def _call_ollama(prompt: str, retries: int = 3) -> str:
                     "stream": False,
                     "options": {
                         "temperature": 0.3,
-                        "num_predict": 256,
+                        "num_predict": max_tokens,
                     },
                 },
                 timeout=120,
@@ -175,7 +175,7 @@ def _call_ollama(prompt: str, retries: int = 3) -> str:
             return f"[doc generation failed after {retries} attempts: {e}]"
 
 
-def _call_openai(prompt: str, retries: int = 3) -> str:
+def _call_openai(prompt: str, retries: int = 3, max_tokens: int = 256) -> str:
     """Call OpenAI-compatible API (Kiro gateway, OpenRouter, etc.)."""
     for attempt in range(retries):
         try:
@@ -186,7 +186,7 @@ def _call_openai(prompt: str, retries: int = 3) -> str:
                     "model": OPENAI_MODEL,
                     "messages": [{"role": "user", "content": prompt}],
                     "temperature": 0.3,
-                    "max_tokens": 256,
+                    "max_tokens": max_tokens,
                 },
                 timeout=120,
             )
@@ -199,8 +199,8 @@ def _call_openai(prompt: str, retries: int = 3) -> str:
             return f"[doc generation failed after {retries} attempts: {e}]"
 
 
-def _call_llm(prompt: str) -> str:
+def _call_llm(prompt: str, max_tokens: int = 256) -> str:
     """Route to the configured backend."""
     if LLM_BACKEND == "openai":
-        return _call_openai(prompt)
-    return _call_ollama(prompt)
+        return _call_openai(prompt, max_tokens=max_tokens)
+    return _call_ollama(prompt, max_tokens=max_tokens)