✨ Add Line Match algorithm and Fix Various Bugs (#300)

* Created line_match algorithm Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Added tests and fixed bugs Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Fixed security vulnerability with langchain-experimental=0.0.63 Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Touched up some stuff Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Fixed versioning issue Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Added some more comments Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Fixed stateful modification of ctx Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Updated requirements.in Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Various bugfixes when putting together the demo Signed-off-by: JonahSussman <sussmanjonah@gmail.com> * Removed wrapper from model_provider.py Signed-off-by: JonahSussman <sussmanjonah@gmail.com> --------- Signed-off-by: JonahSussman <sussmanjonah@gmail.com>
konveyor · Aug 22, 2024 · 46460dd · 46460dd
1 parent 9bc5781
commit 46460dd
Show file tree

Hide file tree

Showing 27 changed files with 603 additions and 55 deletions.
diff --git a/kai/data/templates/solution_handling/before_and_after.jinja b/kai/data/templates/solution_handling/before_and_after.jinja
@@ -1,9 +1,9 @@
-Solution before changes:
+Solved example before changes:
 ```
 {{ solution.original_code }}
 ```
 
-Solution after changes:
+Solved example after changes:
 ```
 {{ solution.updated_code }}
 ```
diff --git a/kai/data/templates/solution_handling/diff_only.jinja b/kai/data/templates/solution_handling/diff_only.jinja
@@ -1,3 +1,3 @@
-Solution diff:
+Solved example diff:
 ```diff
 {{ solution.file_diff }}
diff --git a/kai/data/templates/solution_handling/llm_summary.jinja b/kai/data/templates/solution_handling/llm_summary.jinja
@@ -1,3 +1,3 @@
-Summary of changes for solution:
+Summary of changes for solved example:
 
 {{ solution.llm_summary }}
diff --git a/kai/models/util.py b/kai/models/util.py
@@ -12,7 +12,11 @@
 # These are known unique variables that can be included by incidents
 # They would prevent matches that we actually want, so we filter them
 # before adding to the database or searching
-FILTERED_INCIDENT_VARS = ("file", "package")
+FILTERED_INCIDENT_VARS = [
+    "file",  # Java, URI of the offending file
+    "package",  # Java, shows the package
+    "name",  # Java, shows the name of the method that caused the incident
+]
 
 
 def remove_known_prefixes(path: str) -> str:

diff --git a/kai/routes/load_analysis_report.py b/kai/routes/load_analysis_report.py
@@ -21,16 +21,17 @@ class PostLoadAnalysisReportApplication(BaseModel):
 
 
 class PostLoadAnalysisReportParams(BaseModel):
-    path_to_report: str
     application: PostLoadAnalysisReportApplication
+    report_data: dict | list[dict]
+    report_id: str
 
 
 @to_route("post", "/load_analysis_report")
 async def post_load_analysis_report(request: Request):
     params = PostLoadAnalysisReportParams.model_validate(await request.json())
 
     application = Application(**params.application.model_dump())
-    report = Report.load_report_from_file(params.path_to_report)
+    report = Report(params.report_data, params.report_id)
 
     count = request.app["kai_application"].incident_store.load_report(
         application, report

diff --git a/kai/service/incident_store/incident_store.py b/kai/service/incident_store/incident_store.py
@@ -181,6 +181,8 @@ def __init__(
         self.solution_detector = solution_detector
         self.solution_producer = solution_producer
 
+        self.create_tables()  # This is a no-op if the tables already exist
+
     def load_report(self, app: Application, report: Report) -> tuple[int, int, int]:
         """
         Load incidents from a report and given application object. Returns a
@@ -273,6 +275,7 @@ def load_report(self, app: Application, report: Report) -> tuple[int, int, int]:
                         session.commit()
 
                     for incident in violation_obj.incidents:
+                        filtered_vars = filter_incident_vars(incident.variables)
                         report_incidents.append(
                             SQLIncident(
                                 violation_name=violation.violation_name,
@@ -281,7 +284,7 @@ def load_report(self, app: Application, report: Report) -> tuple[int, int, int]:
                                 incident_uri=incident.uri,
                                 incident_snip=incident.code_snip,
                                 incident_line=incident.line_number,
-                                incident_variables=deep_sort(incident.variables),
+                                incident_variables=deep_sort(filtered_vars),
                                 incident_message=incident.message,
                             )
                         )
@@ -401,9 +404,7 @@ def find_solutions(
             )
 
             result: list[Solution] = []
-            for incident in session.execute(
-                select_incidents_with_solutions_stmt
-            ).scalars():
+            for incident in session.scalars(select_incidents_with_solutions_stmt).all():
                 select_accepted_solution_stmt = select(SQLAcceptedSolution).where(
                     SQLAcceptedSolution.solution_id == incident.solution_id
                 )
@@ -415,9 +416,18 @@ def find_solutions(
                 processed_solution = self.solution_producer.post_process_one(
                     incident, accepted_solution.solution
                 )
+
+                # TODO: This first line doesn't work for some reason. The second
+                # line is a hack to get around it.
                 accepted_solution.solution = processed_solution
+                session.query(SQLAcceptedSolution).filter(
+                    SQLAcceptedSolution.solution_id == incident.solution_id
+                ).update({"solution": processed_solution})
+
                 result.append(processed_solution)
 
+                session.commit()
+
             session.commit()
 
             return result

diff --git a/kai/service/incident_store/sql_types.py b/kai/service/incident_store/sql_types.py
@@ -23,7 +23,7 @@
 
 class SQLSolutionType(TypeDecorator):
     impl = VARCHAR
-    cache_ok = True
+    cache_ok = False
 
     def process_bind_param(self, value: Optional[Solution], dialect: Dialect):
         # Into the db
@@ -148,7 +148,7 @@ class SQLIncident(SQLBase):
     incident_uri: Mapped[str]
     incident_message: Mapped[str]
     incident_snip: Mapped[str]
-    incident_line: Mapped[int]
+    incident_line: Mapped[int]  # 0-indexed!
     incident_variables: Mapped[dict[str, Any]]
     solution_id: Mapped[Optional[str]] = mapped_column(
         ForeignKey("accepted_solutions.solution_id")

diff --git a/kai/service/kai_application/kai_application.py b/kai/service/kai_application/kai_application.py
@@ -148,9 +148,10 @@ def get_incident_solutions_for_file(
                     )
 
                     if len(solutions) != 0:
-                        pb_incident["solution_str"] = self.solution_consumer(
-                            solutions[0]
-                        )
+                        solution_str = self.solution_consumer(solutions[0])
+
+                        if len(solution_str) != 0:
+                            pb_incident["solution_str"] = solution_str
 
             pb_vars = {
                 "src_file_name": file_name,

diff --git a/kai/service/llm_interfacing/model_provider.py b/kai/service/llm_interfacing/model_provider.py
@@ -100,7 +100,22 @@ def __init__(self, config: KaiConfigModels):
                 model_class = FakeListChatModel
 
                 defaults = {
-                    "responses": ["Default LLM response."],
+                    "responses": [
+                        "## Reasoning\n"
+                        "\n"
+                        "Default reasoning.\n"
+                        "\n"
+                        "## Updated File\n"
+                        "\n"
+                        "```\n"
+                        "Default updated file.\n"
+                        "```\n"
+                        "\n"
+                        "## Additional Information\n"
+                        "\n"
+                        "Default additional information.\n"
+                        "\n"
+                    ],
                     "sleep": None,
                 }
 

diff --git a/kai/service/solution_handling/consumption.py b/kai/service/solution_handling/consumption.py
@@ -42,6 +42,9 @@ def solution_consumer_before_and_after(solution: Solution) -> str:
 
 
 def solution_consumer_llm_summary(solution: Solution) -> str:
+    if solution.llm_summary is None:
+        return ""
+
     return (
         __create_jinja_env().get_template("llm_summary.jinja").render(solution=solution)
     )