From 2c5a13bba7da026af5b2ee891764e6eefbe317c9 Mon Sep 17 00:00:00 2001
From: Greg DeYoung <greg@deyoungnet.com>
Date: Thu, 9 Apr 2026 21:41:29 +0000
Subject: [PATCH] fix: normalize incomplete tool requests at extraction layer

When LLMs output valid JSON with tool_name but missing/malformed tool_args, json_parse_dirty() returned the incomplete dict as-is, causing ValueError crashes in validate_tool_request().

Root cause fix: normalize tool requests in json_parse_dirty() before returning them:

- Missing tool_args -> defaults to {}

- tool_args as string -> converts to dict

- tool_args as non-dict -> defaults to {}

- tool key -> normalized to tool_name

Non-tool requests are left untouched.

Related: #1458, #1466
---
 helpers/extract_tools.py | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/helpers/extract_tools.py b/helpers/extract_tools.py
index d939cc534..2b92c3aed 100644
--- a/helpers/extract_tools.py
+++ b/helpers/extract_tools.py
@@ -13,6 +13,16 @@ def json_parse_dirty(json: str) -> dict[str, Any] | None:
         try:
             data = DirtyJson.parse_string(ext_json)
             if isinstance(data, dict):
+                # Normalize: if it looks like a tool request, ensure required fields exist
+                if "tool_name" in data or "tool" in data:
+                    # Normalize tool_name (some models use "tool" instead of "tool_name")
+                    if "tool" in data and "tool_name" not in data:
+                        data["tool_name"] = data.pop("tool")
+                    # Normalize tool_args — default to empty dict if missing or not a dict
+                    if "tool_args" not in data:
+                        data["tool_args"] = data.get("args", {})
+                    if not isinstance(data.get("tool_args"), dict):
+                        data["tool_args"] = {}
                 return data
         except Exception:
             # If parsing fails, return None instead of crashing