jarvis/dev-intel-poc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: increase max_tokens for repo-level docs to 1024

Browse Source 
The new structured repo doc prompt (5 sections with headers) needs more
room than the old 4-6 sentence prompt. File and relationship docs stay
at 256 tokens. Plumbed max_tokens through _call_llm → _call_ollama/_call_openai.
This commit is contained in:
Jarvis Prime
2026-03-05 04:25:33 +00:00
parent c9e4337e15
commit cbf3c8141c
1 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
docgen.py
View File

@@ -121,7 +121,7 @@ Key source files (with their generated documentation):
Project Overview:""" 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]: 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 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.""" """Call Ollama API with retries."""
for attempt in range(retries): for attempt in range(retries):
try: try:
@@ -161,7 +161,7 @@ def _call_ollama(prompt: str, retries: int = 3) -> str:
"stream": False, "stream": False,
"options": { "options": {
"temperature": 0.3, "temperature": 0.3,
"num_predict": 256, "num_predict": max_tokens,
}, },
}, },
timeout=120, timeout=120,
@@ -175,7 +175,7 @@ def _call_ollama(prompt: str, retries: int = 3) -> str:
return f"[doc generation failed after {retries} attempts: {e}]" 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.).""" """Call OpenAI-compatible API (Kiro gateway, OpenRouter, etc.)."""
for attempt in range(retries): for attempt in range(retries):
try: try:
@@ -186,7 +186,7 @@ def _call_openai(prompt: str, retries: int = 3) -> str:
"model": OPENAI_MODEL, "model": OPENAI_MODEL,
"messages": [{"role": "user", "content": prompt}], "messages": [{"role": "user", "content": prompt}],
"temperature": 0.3, "temperature": 0.3,
"max_tokens": 256, "max_tokens": max_tokens,
}, },
timeout=120, timeout=120,
) )
@@ -199,8 +199,8 @@ def _call_openai(prompt: str, retries: int = 3) -> str:
return f"[doc generation failed after {retries} attempts: {e}]" 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.""" """Route to the configured backend."""
if LLM_BACKEND == "openai": if LLM_BACKEND == "openai":
return _call_openai(prompt) return _call_openai(prompt, max_tokens=max_tokens)
return _call_ollama(prompt) return _call_ollama(prompt, max_tokens=max_tokens)