logdetective 2.4.1__tar.gz → 2.6.0__tar.gz

{logdetective-2.4.1 → logdetective-2.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: logdetective
-Version: 2.4.1
+Version: 2.6.0
 Summary: Log using LLM AI to search for build/test failures and provide ideas for fixing these.
 License: Apache-2.0
 License-File: LICENSE
@@ -27,6 +27,8 @@ Requires-Dist: aiohttp (>=3.7.4,<4.0.0)
 Requires-Dist: aiolimiter (>=1.0.0,<2.0.0) ; extra == "server"
 Requires-Dist: aioresponses (>=0.7.8,<0.8.0) ; extra == "testing"
 Requires-Dist: alembic (>=1.13.3,<2.0.0) ; extra == "server" or extra == "server-testing"
+Requires-Dist: asciidoc[testing] (>=10.2.1,<11.0.0) ; extra == "testing"
+Requires-Dist: asyncpg (>=0.30.0,<0.31.0) ; extra == "server" or extra == "server-testing"
 Requires-Dist: backoff (==2.2.1) ; extra == "server" or extra == "server-testing"
 Requires-Dist: drain3 (>=0.9.11,<0.10.0)
 Requires-Dist: fastapi (>=0.111.1,<1.0.0) ; extra == "server" or extra == "server-testing"
@@ -37,11 +39,10 @@ Requires-Dist: llama-cpp-python (>0.2.56,!=0.2.86,<1.0.0)
 Requires-Dist: matplotlib (>=3.8.4,<4.0.0) ; extra == "server" or extra == "server-testing"
 Requires-Dist: numpy (>=1.26.0)
 Requires-Dist: openai (>=1.82.1,<2.0.0) ; extra == "server" or extra == "server-testing"
-Requires-Dist: psycopg2 (>=2.9.9,<3.0.0) ; extra == "server"
-Requires-Dist: psycopg2-binary (>=2.9.9,<3.0.0) ; extra == "server-testing"
 Requires-Dist: pydantic (>=2.8.2,<3.0.0)
 Requires-Dist: pytest (>=8.4.1,<9.0.0) ; extra == "testing"
 Requires-Dist: pytest-asyncio (>=1.1.0,<2.0.0) ; extra == "testing"
+Requires-Dist: pytest-cov[testing] (>=7.0.0,<8.0.0) ; extra == "testing"
 Requires-Dist: pytest-mock (>=3.14.1,<4.0.0) ; extra == "server-testing"
 Requires-Dist: python-gitlab (>=4.4.0)
 Requires-Dist: pyyaml (>=6.0.1,<7.0.0)
@@ -127,6 +128,20 @@ Note that streaming with some models (notably Meta-Llama-3) is broken and can be
 
     logdetective https://example.com/logs.txt --model QuantFactory/Meta-Llama-3-8B-Instruct-GGUF --filename_suffix Q5_K_M.gguf --no-stream
 
+Choice of LLM
+-------------
+
+While Log Detective is compatible with a wide range of LLMs, it does require an instruction tuned model to function properly.
+
+Whether or not the model has been trained to work with instructions can be determined by examining the model card, or simply by checking if it has `instruct` in its name.
+
+When deployed as a server, Log Detective uses `/chat/completions` API as defined by OpenAI. The API must support both `system` and `user` roles, in order to properly work with a system prompt.
+
+Configuration fields `system_role` and `user_role` can be used to set role names for APIs with non-standard roles.
+
+> **Note:**
+> In cases when no system role is available, it is possible to set both fields to the same value. This will concatenate system and standard prompt.
+> This may have negative impact coherence of response.
 
 Real Example
 ------------

{logdetective-2.4.1 → logdetective-2.6.0}/README.md

@@ -73,6 +73,20 @@ Note that streaming with some models (notably Meta-Llama-3) is broken and can be
 
     logdetective https://example.com/logs.txt --model QuantFactory/Meta-Llama-3-8B-Instruct-GGUF --filename_suffix Q5_K_M.gguf --no-stream
 
+Choice of LLM
+-------------
+
+While Log Detective is compatible with a wide range of LLMs, it does require an instruction tuned model to function properly.
+
+Whether or not the model has been trained to work with instructions can be determined by examining the model card, or simply by checking if it has `instruct` in its name.
+
+When deployed as a server, Log Detective uses `/chat/completions` API as defined by OpenAI. The API must support both `system` and `user` roles, in order to properly work with a system prompt.
+
+Configuration fields `system_role` and `user_role` can be used to set role names for APIs with non-standard roles.
+
+> **Note:**
+> In cases when no system role is available, it is possible to set both fields to the same value. This will concatenate system and standard prompt.
+> This may have negative impact coherence of response.
 
 Real Example
 ------------

{logdetective-2.4.1 → logdetective-2.6.0}/logdetective/prompts-summary-first.yml

@@ -18,5 +18,3 @@ prompt_template: |
   Snippets:
 
   {}
-
-  Analysis:

{logdetective-2.4.1 → logdetective-2.6.0}/logdetective/prompts.yml

@@ -19,7 +19,6 @@ prompt_template: |
 
   {}
 
-  Analysis:
 
 snippet_prompt_template: |
   Analyse following RPM build log snippet. Describe contents accurately, without speculation or suggestions for resolution
@@ -30,7 +29,6 @@ snippet_prompt_template: |
 
   {}
 
-  Analysis:
 
 prompt_template_staged: |
   Given following log snippets, their explanation, and nothing else, explain what failure, if any, occurred during build of this package.
@@ -47,7 +45,6 @@ prompt_template_staged: |
 
   {}
 
-  Analysis:
 
 # System prompts
 # System prompts are meant to serve as general guide for model behavior,

logdetective-2.6.0/logdetective/server/database/base.py

@@ -0,0 +1,71 @@
+from os import getenv
+from contextlib import asynccontextmanager
+from sqlalchemy.orm import DeclarativeBase
+from sqlalchemy.ext.asyncio import create_async_engine, async_sessionmaker
+from logdetective import logger
+
+
+def get_pg_url() -> str:
+    """create postgresql connection string"""
+    return (
+        f"postgresql+asyncpg://{getenv('POSTGRESQL_USER')}"
+        f":{getenv('POSTGRESQL_PASSWORD')}@{getenv('POSTGRESQL_HOST', 'postgres')}"
+        f":{getenv('POSTGRESQL_PORT', '5432')}/{getenv('POSTGRESQL_DATABASE')}"
+    )
+
+
+# To log SQL statements, set SQLALCHEMY_ECHO env. var. to True|T|Yes|Y|1
+sqlalchemy_echo = getenv("SQLALCHEMY_ECHO", "False").lower() in (
+    "true",
+    "t",
+    "yes",
+    "y",
+    "1",
+)
+engine = create_async_engine(get_pg_url(), echo=sqlalchemy_echo)
+SessionFactory = async_sessionmaker(autoflush=True, bind=engine)  # pylint: disable=invalid-name
+
+
+class Base(DeclarativeBase):
+    """Declarative base class for all ORM models."""
+
+
+@asynccontextmanager
+async def transaction(commit: bool = False):
+    """
+    Context manager for 'framing' a db transaction.
+
+    Args:
+        commit: Whether to call `Session.commit()` upon exiting the context. Should be set to True
+            if any changes are made within the context. Defaults to False.
+    """
+
+    session = SessionFactory()
+    async with session:
+        try:
+            yield session
+            if commit:
+                await session.commit()
+        except Exception as ex:
+            logger.warning("Exception while working with database: %s", str(ex))
+            await session.rollback()
+            raise
+        finally:
+            await session.close()
+
+
+async def init():
+    """Init db"""
+    async with engine.begin() as conn:
+        await conn.run_sync(Base.metadata.create_all)
+    logger.debug("Database initialized")
+
+
+async def destroy():
+    """Destroy db"""
+    async with engine.begin() as conn:
+        await conn.run_sync(Base.metadata.drop_all)
+    logger.warning("Database cleaned")
+
+
+DB_MAX_RETRIES = 3  # How many times retry a db operation

{logdetective-2.4.1 → logdetective-2.6.0}/logdetective/server/database/models/__init__.py

@@ -1,4 +1,3 @@
-from logdetective.server.database.base import Base
 from logdetective.server.database.models.merge_request_jobs import (
     Forge,
     GitlabMergeRequestJobs,
@@ -18,8 +17,9 @@ from logdetective.server.database.models.exceptions import (
     KojiTaskAnalysisTimeoutError,
 )
 
+# pylint: disable=undefined-all-variable
+
 __all__ = [
-    Base.__name__,
     GitlabMergeRequestJobs.__name__,
     Comments.__name__,
     Reactions.__name__,

{logdetective-2.4.1 → logdetective-2.6.0}/logdetective/server/database/models/koji.py

@@ -1,6 +1,9 @@
+from __future__ import annotations
+from typing import Optional
 from datetime import datetime, timedelta, timezone
-from sqlalchemy import Column, BigInteger, DateTime, ForeignKey, Integer, String
-from sqlalchemy.orm import relationship
+from sqlalchemy import BigInteger, DateTime, ForeignKey, Integer, String, select
+from sqlalchemy.orm import Mapped, mapped_column, relationship
+
 from sqlalchemy.exc import OperationalError
 import backoff
 
@@ -21,42 +24,47 @@ class KojiTaskAnalysis(Base):
 
     __tablename__ = "koji_task_analysis"
 
-    id = Column(Integer, primary_key=True)
-    koji_instance = Column(String(255), nullable=False, index=True)
-    task_id = Column(BigInteger, nullable=False, index=True, unique=True)
-    log_file_name = Column(String(255), nullable=False, index=True)
-    request_received_at = Column(
-        DateTime,
+    id: Mapped[int] = mapped_column(Integer, primary_key=True)
+    koji_instance: Mapped[str] = mapped_column(String(255), nullable=False, index=True)
+    task_id: Mapped[int] = mapped_column(BigInteger, nullable=False, index=True, unique=True)
+    log_file_name: Mapped[str] = mapped_column(String(255), nullable=False, index=True)
+    request_received_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True),
         nullable=False,
         index=True,
         default=datetime.now(timezone.utc),
         comment="Timestamp when the request was received",
     )
-    response_id = Column(
+    response_id: Mapped[Optional[int]] = mapped_column(
         Integer,
         ForeignKey("analyze_request_metrics.id"),
         nullable=True,
         index=False,
         comment="The id of the analyze request metrics for this task",
     )
-    response = relationship("AnalyzeRequestMetrics")
+    response: Mapped[Optional["AnalyzeRequestMetrics"]] = relationship(
+        "AnalyzeRequestMetrics",
+        back_populates="koji_tasks"
+    )
 
     @classmethod
     @backoff.on_exception(backoff.expo, OperationalError, max_tries=DB_MAX_RETRIES)
-    def create_or_restart(cls, koji_instance: str, task_id: int, log_file_name: str):
+    async def create_or_restart(
+        cls, koji_instance: str, task_id: int, log_file_name: str
+    ):
         """Create a new koji task analysis"""
-        with transaction(commit=True) as session:
+        query = select(cls).filter(
+            cls.koji_instance == koji_instance, cls.task_id == task_id
+        )
+        async with transaction(commit=True) as session:
             # Check if the task analysis already exists
-            koji_task_analysis = (
-                session.query(cls)
-                .filter_by(koji_instance=koji_instance, task_id=task_id)
-                .first()
-            )
+            query_result = await session.execute(query)
+            koji_task_analysis = query_result.first()
             if koji_task_analysis:
                 # If it does, update the request_received_at timestamp
                 koji_task_analysis.request_received_at = datetime.now(timezone.utc)
                 session.add(koji_task_analysis)
-                session.flush()
+                await session.flush()
                 return
 
             # If it doesn't, create a new one
@@ -65,14 +73,19 @@ class KojiTaskAnalysis(Base):
             koji_task_analysis.task_id = task_id
             koji_task_analysis.log_file_name = log_file_name
             session.add(koji_task_analysis)
-            session.flush()
+            await session.flush()
 
     @classmethod
     @backoff.on_exception(backoff.expo, OperationalError, max_tries=DB_MAX_RETRIES)
-    def add_response(cls, task_id: int, metric_id: int):
+    async def add_response(cls, task_id: int, metric_id: int):
         """Add a response to a koji task analysis"""
-        with transaction(commit=True) as session:
-            koji_task_analysis = session.query(cls).filter_by(task_id=task_id).first()
+        query = select(cls).filter(cls.task_id == task_id)
+        metrics_query = select(AnalyzeRequestMetrics).filter(
+            AnalyzeRequestMetrics.id == metric_id
+        )
+        async with transaction(commit=True) as session:
+            query_result = await session.execute(query)
+            koji_task_analysis = query_result.scalars().first()
             # Ensure that the task analysis doesn't already have a response
             if koji_task_analysis.response:
                 # This is probably due to an analysis that took so long that
@@ -81,20 +94,20 @@ class KojiTaskAnalysis(Base):
                 # returned to the consumer, so we'll just drop this extra one
                 # on the floor and keep the one saved in the database.
                 return
-
-            metric = (
-                session.query(AnalyzeRequestMetrics).filter_by(id=metric_id).first()
-            )
+            metrics_query_result = await session.execute(metrics_query)
+            metric = metrics_query_result.scalars().first()
             koji_task_analysis.response = metric
             session.add(koji_task_analysis)
-            session.flush()
+            await session.flush()
 
     @classmethod
     @backoff.on_exception(backoff.expo, OperationalError, max_tries=DB_MAX_RETRIES)
-    def get_response_by_task_id(cls, task_id: int) -> KojiStagedResponse:
+    async def get_response_by_task_id(cls, task_id: int) -> KojiStagedResponse:
         """Get a koji task analysis by task id"""
-        with transaction(commit=False) as session:
-            koji_task_analysis = session.query(cls).filter_by(task_id=task_id).first()
+        query = select(cls).filter(cls.task_id == task_id)
+        async with transaction(commit=False) as session:
+            query_result = await session.execute(query)
+            koji_task_analysis = query_result.scalars().first()
             if not koji_task_analysis:
                 raise KojiTaskNotFoundError(f"Task {task_id} not yet analyzed")