Feature/fix broken regr (#282)

* Change eval to use task workflow

* Change eval to use task workflow

* fix eval harness test

* Add sql tests

* 'Refactored by Sourcery' (#280)

Co-authored-by: Sourcery AI <>

* tweak regr

* add type check

* remove prints

* remove bare except

* Change node path

* Remove unused fstring, fix security issues

* fixing small bugs

* increment embedding data

* finish

* fix module regr

---------

Co-authored-by: sourcery-ai[bot] <58596630+sourcery-ai[bot]@users.noreply.github.com>
Co-authored-by: Nolan Tremelling <34580718+NolanTrem@users.noreply.github.com>

tests/regression/test_agents.py

@@ -1,6 +1,7 @@
 # TODO - Agent tests should depend on actions for verification, not specific output
 import ast
 import logging
+import random
 
 import pytest
 
@@ -94,18 +95,21 @@ def test_agent_py_reader_and_context(
         )
 
 
+random_suffix = random.randint(0, 1000000)
+
+
 @pytest.mark.regression
 @pytest.mark.parametrize(
     "instructions, toolkit_list, model, agent_config_name, max_iterations, expected_output_module, expected_result, min_similarity",
     [
         # Extracting source code directly from a module
         (
-            "Create a new module with a hello world function at automata.test_module",
+            f"Create a new module with a hello world function at automata.test_module_{random_suffix}",
             ["py-writer"],
             "gpt-4",
             "automata-main",
             2,
-            "automata.test_output.test_module",
+            f"automata.test_module_{random_suffix}",
             "def hello_world():\n    print('Hello, world!')",
             0.9,
         ),