Merge pull request #147 from dbpunk-labs/142-bug-fail-to-start-og-com…

…mand-in-gitpod

142 bug fail to start og command in gitpod

.github/workflows/cli_ci.yaml

@@ -13,7 +13,7 @@ jobs:
       matrix:
         #os: [ubuntu-latest, windows-latest, macos-latest]
         os: [ubuntu-latest, windows-latest]
-        python-version: ["3.10", "3.11"]
+        python-version: ["3.11"]
     runs-on: ${{ matrix.os }}
     steps:
       - name: Checkout repository

agent/src/og_agent/openai_agent.py

@@ -52,29 +52,6 @@
             "required": ["explanation", "code"],
         },
     },
-    {
-        "name": "python",
-        "description": "You must not call this function",
-        "parameters": {
-            "type": "object",
-            "properties": {
-                "explanation": {
-                    "type": "string",
-                    "description": "the explanation about the python code",
-                },
-                "code": {
-                    "type": "string",
-                    "description": "the python code to be executed",
-                },
-                "saved_filenames": {
-                    "type": "array",
-                    "items": {"type": "string"},
-                    "description": "A list of filenames that were created by the code",
-                },
-            },
-            "required": ["explanation", "code"],
-        },
-    },
 ]
 
 
@@ -108,6 +85,9 @@ def _merge_delta_for_content(self, message, delta):
             message["content"] = content + delta["content"]
 
     def _get_function_call_argument_new_typing(self, message):
+        if message['function_call']['name'] == 'python':
+            return TypingState.CODE, "", message["function_call"].get("arguments", "")
+
         arguments = message["function_call"].get("arguments", "")
         state = TypingState.START
         explanation_str = ""
@@ -240,11 +220,18 @@ async def handle_function(self, message, queue, context, task_context):
                 logging.debug("the client has cancelled the request")
                 return
             function_name = message["function_call"]["name"]
-            arguments = json.loads(message["function_call"]["arguments"])
-            logger.debug(f"call function {function_name} with args {arguments}")
-            code = arguments["code"]
-            explanation = arguments["explanation"]
-            saved_filenames = arguments.get("saved_filenames", [])
+            code = ""
+            explanation = ""
+            saved_filenames = []
+            if function_name == 'python':
+                code = message["function_call"]["arguments"]
+                logger.debug(f"call function {function_name} with args {code}")
+            else:
+                arguments = json.loads(message["function_call"]["arguments"])
+                logger.debug(f"call function {function_name} with args {arguments}")
+                code = arguments["code"]
+                explanation = arguments["explanation"]
+                saved_filenames = arguments.get("saved_filenames", [])
             tool_input = json.dumps({
                 "code": code,
                 "explanation": explanation,
@@ -299,7 +286,8 @@ async def arun(self, task, queue, context, max_iteration=5):
                 )
                 logger.debug(f"the response {chat_message}")
                 if "function_call" in chat_message:
-                    chat_message["content"] = None
+                    if 'content' not in chat_message:
+                        chat_message["content"] = None
                     messages.append(chat_message)
                     function_name = chat_message["function_call"]["name"]
                     if function_name not in ["execute_python_code", "python"]:

agent/src/og_agent/prompt.py

@@ -20,8 +20,8 @@
 Secondly, Being an expert in programming, you must follow the rules
 * To complete the goal, You must write a plan and execute it step by step, the followings are examples
     * The data visualization plan involves previewing, cleaning, and processing the data to generate the chart.
-* For each step, provide an **explanation** and the code block.
-    * Execute the python code using function `execute_python_code`
+* Every step must include the explanation and the code block
+    * Execute the python code using function `execute_python_code` 
     * If the code creates any files, add them to the saved_filenames of function `execute_python_code`.
     * If the code has any display data, save it as a file and add it to the saved_filenames of function `execute_python_code`
 * You must try to correct your code when you get errors from the output
@@ -66,3 +66,4 @@
 * language (string): The programming language used to execute the action.
 * is_final_answer (boolean): Whether this is the final answer to the question. If it is, the value of this field should be true. Otherwise, the value should be false.
 """
+

up/tests/up_tests.py

@@ -165,7 +165,7 @@ def test_start_azure_openai_smoketest():
     kernel_key = random_str(32)
     with Live(Group(*segments), console=console) as live:
         generate_kernel_env(live, segments, install_dir, kernel_key)
-        code = load_docker_image("v0.4.27", "dbpunk/octogen", live, segments)
+        code = load_docker_image("v0.4.41", "dbpunk/octogen", live, segments)
         assert code == 0, "bad result code of loading docker image"
         result = start_octogen_for_azure_openai(
             live,
@@ -175,7 +175,7 @@ def test_start_azure_openai_smoketest():
             admin_key,
             kernel_key,
             "dbpunk/octogen",
-            "v0.4.27",
+            "v0.4.41",
             "azure_open_api_key",
             "test_deployment",
             "https://azure_base",
@@ -193,7 +193,7 @@ def test_start_openai_smoketest():
     kernel_key = random_str(32)
     with Live(Group(*segments), console=console) as live:
         generate_kernel_env(live, segments, install_dir, kernel_key)
-        code = load_docker_image("v0.4.27", "dbpunk/octogen", live, segments)
+        code = load_docker_image("v0.4.41", "dbpunk/octogen", live, segments)
         assert code == 0, "bad result code of loading docker image"
         result = start_octogen_for_openai(
             live,
@@ -203,7 +203,7 @@ def test_start_openai_smoketest():
             admin_key,
             kernel_key,
             "dbpunk/octogen",
-            "v0.4.27",
+            "v0.4.41",
             "openai_api_key",
             "gpt-3.5-turbo",
         )
@@ -220,7 +220,7 @@ def test_start_codellama_smoketest():
     kernel_key = random_str(32)
     with Live(Group(*segments), console=console) as live:
         generate_kernel_env(live, segments, install_dir, kernel_key)
-        code = load_docker_image("v0.4.27", "dbpunk/octogen", live, segments)
+        code = load_docker_image("v0.4.41", "dbpunk/octogen", live, segments)
         assert code == 0, "bad result code of loading docker image"
         result = start_octogen_for_codellama(
             live,
@@ -232,7 +232,7 @@ def test_start_codellama_smoketest():
             admin_key,
             kernel_key,
             "dbpunk/octogen",
-            "v0.4.27",
+            "v0.4.41",
         )
         assert result
 
@@ -251,5 +251,5 @@ def test_load_valid_docker_image():
     console = Console()
     segments = []
     with Live(Group(*segments), console=console) as live:
-        code = load_docker_image("v0.4.26", "dbpunk/octogen", live, segments)
+        code = load_docker_image("v0.4.41", "dbpunk/octogen", live, segments)
         assert code == 0, "loading image should be ok"