From 2c394e617cb69bb6553e35c1782e80348cdb1ff5 Mon Sep 17 00:00:00 2001
From: AMATH <116212274+amathxbt@users.noreply.github.com>
Date: Sat, 18 Apr 2026 10:09:57 +0100
Subject: [PATCH] Improve error handling for invalid model identifiers

Add a shared parser to validate model identifiers and raise a descriptive ValueError for invalid formats.

Signed-off-by: AMATH <116212274+amathxbt@users.noreply.github.com>
---
 ...ise clear error for invalid model identifiers | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
 create mode 100644 fix(llm): raise clear error for invalid model identifiers

diff --git a/fix(llm): raise clear error for invalid model identifiers b/fix(llm): raise clear error for invalid model identifiers
new file mode 100644
index 0000000..a205e9a
--- /dev/null
+++ b/fix(llm): raise clear error for invalid model identifiers	
@@ -0,0 +1,16 @@
+## Summary
+
+Fixes invalid model string handling in the LLM client.
+
+`LLM.completion()` and `LLM.chat()` currently assume the model identifier always contains a `/` separator and directly access `model.split("/")[1]`. When a caller passes a plain string such as `"gpt-5"`, the SDK raises an unhelpful `IndexError` instead of a clear validation error.
+
+This PR adds a shared parser that validates model identifiers and raises a descriptive `ValueError` when the input is not in `provider/model-name` format.
+
+Fixes #249
+
+## Problem
+
+Both `LLM.completion()` and `LLM.chat()` strip the provider prefix with direct string splitting:
+
+```python
+model.split("/")[1]