Implement Customize Build Commands and Test Commands

docs/GitHub-Issue-Debug-Guide.md

@@ -161,7 +161,6 @@ After execution, the script outputs results in JSON format, including the follow
   "prometheus_result": {
     "patch": "Generated code patch",
     "passed_reproducing_test": true,
-    "passed_build": true,
     "passed_existing_test": false,
     "passed_regression_test": true, 
     "passed_reproduction_test": true,

prometheus/app/api/routes/issue.py

@@ -60,6 +60,18 @@ async def answer_issue(issue: IssueRequest, request: Request) -> Response[IssueR
                 code=400,
                 message="workdir must be provided for user defined environment",
             )
+
+    # Validate build and test commands if required
+    if issue.run_build and not issue.build_commands:
+        raise ServerException(
+            code=400, message="No build commands available, please provide build commands"
+        )
+
+    if issue.run_existing_test and not issue.test_commands:
+        raise ServerException(
+            code=400, message="No test commands available, please provide test commands"
+        )
+
     # Ensure the repository is not currently being used
     if repository.is_working:
         raise ServerException(
@@ -83,7 +95,6 @@ async def answer_issue(issue: IssueRequest, request: Request) -> Response[IssueR
     (
         patch,
         passed_reproducing_test,
-        passed_build,
         passed_regression_test,
         passed_existing_test,
         issue_response,
@@ -115,12 +126,11 @@ async def answer_issue(issue: IssueRequest, request: Request) -> Response[IssueR
     if (
         patch,
         passed_reproducing_test,
-        passed_build,
         passed_regression_test,
         passed_existing_test,
         issue_response,
         issue_type,
-    ) == (None, False, False, False, False, None, None):
+    ) == (None, False, False, False, None, None):
         raise ServerException(
             code=500,
             message="Failed to process the issue. Please try again later.",
@@ -135,7 +145,6 @@ async def answer_issue(issue: IssueRequest, request: Request) -> Response[IssueR
         data=IssueResponse(
             patch=patch,
             passed_reproducing_test=passed_reproducing_test,
-            passed_build=passed_build,
             passed_regression_test=passed_regression_test,
             passed_existing_test=passed_existing_test,
             issue_response=issue_response,

prometheus/app/models/response/issue.py

@@ -6,7 +6,6 @@
 class IssueResponse(BaseModel):
     patch: str | None = None
     passed_reproducing_test: bool
-    passed_build: bool
     passed_regression_test: bool
     passed_existing_test: bool
     issue_response: str | None = None

prometheus/app/services/issue_service.py

@@ -48,10 +48,7 @@ def answer_issue(
         dockerfile_content: Optional[str] = None,
         image_name: Optional[str] = None,
         workdir: Optional[str] = None,
-    ) -> (
-        tuple[None, bool, bool, bool, bool, None, None]
-        | tuple[str, bool, bool, bool, bool, str, IssueType]
-    ):
+    ) -> tuple[None, bool, bool, bool, None, None] | tuple[str, bool, bool, bool, str, IssueType]:
         """
         Processes an issue, generates patches if needed, runs optional builds and tests, and returning the results.
 
@@ -76,9 +73,10 @@ def answer_issue(
             Tuple containing:
                 - edit_patch (str): The generated patch for the issue.
                 - passed_reproducing_test (bool): Whether the reproducing test passed.
-                - passed_build (bool): Whether the build passed.
+                - passed_regression_test (bool): Whether the regression tests passed.
                 - passed_existing_test (bool): Whether the existing tests passed.
                 - issue_response (str): Response generated for the issue.
+                - issue_type (IssueType): The type of the issue (BUG or QUESTION).
         """
 
         # Set up a dedicated logger for this thread
@@ -111,7 +109,6 @@ def answer_issue(
             kg=knowledge_graph,
             git_repo=repository,
             container=container,
-            build_commands=build_commands,
             test_commands=test_commands,
         )
 
@@ -131,15 +128,14 @@ def answer_issue(
             return (
                 output_state["edit_patch"],
                 output_state["passed_reproducing_test"],
-                output_state["passed_build"],
                 output_state["passed_regression_test"],
                 output_state["passed_existing_test"],
                 output_state["issue_response"],
                 output_state["issue_type"],
             )
         except Exception as e:
             logger.error(f"Error in answer_issue: {str(e)}\n{traceback.format_exc()}")
-            return None, False, False, False, False, None, None
+            return None, False, False, False, None, None
         finally:
             logger.removeHandler(file_handler)
             file_handler.close()

prometheus/docker/base_container.py

@@ -99,7 +99,9 @@ def update_files(
         Creates a tar archive of the new files and copies them into the workdir of the container.
 
         Args:
-          new_project_path: Path to the directory containing new files.
+            project_root_path: Path to the project root directory.
+            updated_files: List of file paths (relative to project_root_path) to update in the container.
+            removed_files: List of file paths (relative to project_root_path) to remove from the container.
         """
         if not project_root_path.is_absolute():
             raise ValueError("project_root_path {project_root_path} must be a absolute path")
@@ -157,10 +159,10 @@ def execute_command(self, command: str) -> str:
 {command} timeout after {self.timeout} seconds
 *******************************************************************************
 """
-        timeout_command = f"timeout -k 5 {self.timeout}s {command}"
-        command = f'/bin/bash -l -c "{timeout_command}"'
+        bash_cmd = ["/bin/bash", "-lc", command]
+        full_cmd = ["timeout", "-k", "5", f"{self.timeout}s", *bash_cmd]
         self._logger.debug(f"Running command in container: {command}")
-        exec_result = self.container.exec_run(command, workdir=self.workdir)
+        exec_result = self.container.exec_run(full_cmd, workdir=self.workdir)
         exec_result_str = exec_result.output.decode("utf-8")
 
         if exec_result.exit_code in (124, 137):
@@ -169,13 +171,11 @@ def execute_command(self, command: str) -> str:
         self._logger.debug(f"Command output:\n{exec_result_str}")
         return exec_result_str
 
-    def restart_container(self):
-        self._logger.info("Restarting the container")
-        if self.container:
-            self.container.stop(timeout=10)
-            self.container.remove(force=True)
-
-        self.start_container()
+    def reset_repository(self):
+        """Reset the git repository in the container to a clean state."""
+        self._logger.info("Resetting git repository in the container")
+        self.execute_command("git reset --hard")
+        self.execute_command("git clean -fd")
 
     def cleanup(self):
         """Clean up container resources and temporary files.

prometheus/lang_graph/graphs/issue_graph.py

@@ -30,7 +30,6 @@ def __init__(
         kg: KnowledgeGraph,
         git_repo: GitRepository,
         container: BaseContainer,
-        build_commands: Optional[Sequence[str]] = None,
         test_commands: Optional[Sequence[str]] = None,
     ):
         self.git_repo = git_repo
@@ -52,7 +51,6 @@ def __init__(
             container=container,
             kg=kg,
             git_repo=git_repo,
-            build_commands=build_commands,
             test_commands=test_commands,
         )
 

prometheus/lang_graph/graphs/issue_state.py

@@ -26,7 +26,6 @@ class IssueState(TypedDict):
 
     passed_regression_test: bool
     passed_reproducing_test: bool
-    passed_build: bool
     passed_existing_test: bool
 
     issue_response: str

prometheus/lang_graph/nodes/context_extraction_node.py

@@ -67,10 +67,16 @@
 
 HUMAN_MESSAGE = """\
 This is the original user query:
+
+--- BEGIN ORIGINAL QUERY ---
 {original_query}
+--- END ORIGINAL QUERY ---
 
 The context or file content that you have seen so far (Some of the context may be IRRELEVANT to the query!!!):
+
+--- BEGIN CONTEXT ---
 {context}
+--- END CONTEXT ---
 
 REMEMBER: Your task is to summarize the relevant contexts to a given query and return it in the specified format!
 """
@@ -112,16 +118,6 @@ def __init__(self, model: BaseChatModel, root_path: str):
             f"thread-{threading.get_ident()}.prometheus.lang_graph.nodes.context_extraction_node"
         )
 
-    def get_human_message(self, state: ContextRetrievalState) -> str:
-        full_context_str = transform_tool_messages_to_str(
-            extract_last_tool_messages(state["context_provider_messages"])
-        )
-        original_query = state["query"]
-        return HUMAN_MESSAGE.format(
-            original_query=original_query,
-            context=full_context_str,
-        )
-
     def __call__(self, state: ContextRetrievalState):
         """
         Extract relevant code contexts from the codebase based on the user query and existing context.
@@ -130,9 +126,26 @@ def __call__(self, state: ContextRetrievalState):
         self._logger.info("Starting context extraction process")
         # Get Context List with existing context
         final_context = state.get("context", [])
-        # Get a human message
-        human_message = self.get_human_message(state)
+
+        # Transform the tool messages to a single string
+        full_context_str = transform_tool_messages_to_str(
+            extract_last_tool_messages(state["context_provider_messages"])
+        )
+
+        # return existing context if no new context is available
+        if not full_context_str:
+            self._logger.debug(
+                "No context available from tool messages, returning existing context"
+            )
+            return {"context": final_context}
+
+        # Format the human message
+        human_message = HUMAN_MESSAGE.format(
+            original_query=state["query"],
+            context=full_context_str,
+        )
         self._logger.debug(human_message)
+
         # Summarize the context based on the last messages and system prompt
         response = self.model.invoke({"human_prompt": human_message})
         self._logger.debug(f"Model response: {response}")

prometheus/lang_graph/nodes/context_retrieval_subgraph_node.py

@@ -3,6 +3,7 @@
 from typing import Dict, Sequence
 
 from langchain_core.language_models.chat_models import BaseChatModel
+from langgraph.errors import GraphRecursionError
 
 from prometheus.graph.knowledge_graph import KnowledgeGraph
 from prometheus.lang_graph.subgraphs.context_retrieval_subgraph import ContextRetrievalSubgraph
@@ -31,8 +32,12 @@ def __init__(
 
     def __call__(self, state: Dict) -> Dict[str, Sequence[Context]]:
         self._logger.info("Enter context retrieval subgraph")
-        output_state = self.context_retrieval_subgraph.invoke(
-            state[self.query_key_name], state["max_refined_query_loop"]
-        )
+        try:
+            output_state = self.context_retrieval_subgraph.invoke(
+                state[self.query_key_name], state["max_refined_query_loop"]
+            )
+        except GraphRecursionError as e:
+            self._logger.debug("Graph recursion limit reached, returning empty context")
+            raise e
         self._logger.info(f"Context retrieved: {output_state['context']}")
         return {self.context_key_name: output_state["context"]}

prometheus/lang_graph/nodes/issue_bug_analyzer_message_node.py

@@ -82,8 +82,6 @@ def format_human_message(self, state: Dict):
             )
         elif "reproducing_test_fail_log" in state and state["reproducing_test_fail_log"]:
             edit_error = f"The patch failed to pass the bug exposing test cases:\n{state['reproducing_test_fail_log']}"
-        elif "build_fail_log" in state and state["build_fail_log"]:
-            edit_error = f"The patch failed to pass the build:\n{state['build_fail_log']}"
         elif "existing_test_fail_log" in state and state["existing_test_fail_log"]:
             edit_error = (
                 f"The patch failed to existing test cases:\n{state['existing_test_fail_log']}"

prometheus/lang_graph/nodes/issue_bug_reproduction_context_message_node.py

@@ -9,7 +9,7 @@ class IssueBugReproductionContextMessageNode:
     BUG_REPRODUCING_QUERY = """\
 {issue_info}
 
-OBJECTIVE: Find three relevant existing test cases that demonstrates similar functionality to the reported bug,
+OBJECTIVE: Find 5 relevant existing test cases that demonstrates similar functionality to the reported bug,
 including ALL necessary imports, test setup, mocking, assertions, and any test method used in the test case.
 
 <reasoning>
@@ -36,7 +36,7 @@ class IssueBugReproductionContextMessageNode:
 </reasoning>
 
 REQUIREMENTS:
-- Return THREE complete, self-contained test cases most similar to bug scenario
+- Return 5 complete, self-contained test cases most similar to bug scenario
 - Must include ALL necessary imports at the start of each test file
 - Must include full test method implementation
 - Must include ALL mock/fixture setup
@@ -105,7 +105,7 @@ def test_file_permission_denied(self, mock_open, mock_access):
 3. Tests with comparable mocking patterns
 4. Tests demonstrating similar assertions
 
-Find the THREE most relevant test cases with complete context, ensuring ALL necessary imports are included at the start of each test file.
+Find the 5 most relevant test cases with complete context, ensuring ALL necessary imports are included at the start of each test file.
 """
 
     def __init__(self):

prometheus/lang_graph/nodes/issue_bug_responder_node.py

@@ -1,10 +1,10 @@
 import logging
 import threading
-from typing import Dict
 
 from langchain_core.language_models.chat_models import BaseChatModel
 from langchain_core.messages import HumanMessage, SystemMessage
 
+from prometheus.lang_graph.subgraphs.issue_bug_state import IssueBugState
 from prometheus.utils.issue_util import format_issue_info
 
 
@@ -53,15 +53,15 @@ def __init__(self, model: BaseChatModel):
             f"thread-{threading.get_ident()}.prometheus.lang_graph.nodes.issue_bug_responder_node"
         )
 
-    def format_human_message(self, state: Dict) -> HumanMessage:
+    def format_human_message(self, state: IssueBugState) -> HumanMessage:
         verification_messages = []
 
         # We only report successful verifications that were performed
         if state["passed_reproducing_test"]:
             verification_messages.append("✓ The bug reproducing test passed")
 
-        if state["passed_build"]:
-            verification_messages.append("✓ Build passes successfully")
+        if state["passed_regression_test"]:
+            verification_messages.append("✓ All selected regression tests passes successfully")
 
         if state["passed_existing_test"]:
             verification_messages.append("✓ All existing tests pass successfully")
@@ -78,7 +78,7 @@ def format_human_message(self, state: Dict) -> HumanMessage:
 
         return HumanMessage(content=formatted_message)
 
-    def __call__(self, state: Dict):
+    def __call__(self, state: IssueBugState):
         messages = [
             self.system_prompt,
             self.format_human_message(state),