logdetective 1.5.0__tar.gz → 1.6.0__tar.gz

{logdetective-1.5.0 → logdetective-1.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: logdetective
-Version: 1.5.0
+Version: 1.6.0
 Summary: Log using LLM AI to search for build/test failures and provide ideas for fixing these.
 License: Apache-2.0
 Author: Jiri Podivin
@@ -90,6 +90,7 @@ To analyze a log file, run the script with the following command line arguments:
 - `--summarizer` DISABLED: LLM summarization option was removed. Argument is kept for backward compatibility only.(optional, default: "drain"): Choose between LLM and Drain template miner as the log summarizer. You can also provide the path to an existing language model file instead of using a URL.
 - `--n_lines` DISABLED: LLM summarization option was removed. Argument is kept for backward compatibility only. (optional, default: 8): The number of lines per chunk for LLM analysis. This only makes sense when you are summarizing with LLM.
 - `--n_clusters` (optional, default 8): Number of clusters for Drain to organize log chunks into. This only makes sense when you are summarizing with Drain
+- `--skip_snippets` Path to patterns for skipping snippets.
 
 Example usage:
 
@@ -438,6 +439,28 @@ with spaces, or replacement fields marked with curly braces, `{}` left for inser
 Number of replacement fields in new prompts, must be the same as in originals.
 Although their position may be different.
 
+
+Skip Snippets
+-------------
+
+Certain log chunks may not contribute to the analysis of the problem under any circumstances.
+User can specify regular expressions, matching such log chunks, along with simple description,
+using Skip Snippets feature.
+
+Patterns to be skipped must be defined yaml file as a dictionary, where key is a description
+and value is a regular expression. For example:
+
+```
+child_exit_code_zero: "Child return code was: 0"
+```
+
+Special care must be taken not to write a regular expression which may match
+too many chunks, or which may be evaluated as data structure by the yaml parser.
+
+Example of a valid pattern definition file: `logdetective/skip_patterns.yml`,
+can be used as a starting point and is used as a default if no other definition is provided.
+
+
 License
 -------
 

{logdetective-1.5.0 → logdetective-1.6.0}/README.md

@@ -46,6 +46,7 @@ To analyze a log file, run the script with the following command line arguments:
 - `--summarizer` DISABLED: LLM summarization option was removed. Argument is kept for backward compatibility only.(optional, default: "drain"): Choose between LLM and Drain template miner as the log summarizer. You can also provide the path to an existing language model file instead of using a URL.
 - `--n_lines` DISABLED: LLM summarization option was removed. Argument is kept for backward compatibility only. (optional, default: 8): The number of lines per chunk for LLM analysis. This only makes sense when you are summarizing with LLM.
 - `--n_clusters` (optional, default 8): Number of clusters for Drain to organize log chunks into. This only makes sense when you are summarizing with Drain
+- `--skip_snippets` Path to patterns for skipping snippets.
 
 Example usage:
 
@@ -394,6 +395,28 @@ with spaces, or replacement fields marked with curly braces, `{}` left for inser
 Number of replacement fields in new prompts, must be the same as in originals.
 Although their position may be different.
 
+
+Skip Snippets
+-------------
+
+Certain log chunks may not contribute to the analysis of the problem under any circumstances.
+User can specify regular expressions, matching such log chunks, along with simple description,
+using Skip Snippets feature.
+
+Patterns to be skipped must be defined yaml file as a dictionary, where key is a description
+and value is a regular expression. For example:
+
+```
+child_exit_code_zero: "Child return code was: 0"
+```
+
+Special care must be taken not to write a regular expression which may match
+too many chunks, or which may be evaluated as data structure by the yaml parser.
+
+Example of a valid pattern definition file: `logdetective/skip_patterns.yml`,
+can be used as a starting point and is used as a default if no other definition is provided.
+
+
 License
 -------
 

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/extractors.py

@@ -5,7 +5,8 @@ from typing import Tuple
 import drain3
 from drain3.template_miner_config import TemplateMinerConfig
 
-from logdetective.utils import get_chunks
+from logdetective.utils import get_chunks, filter_snippet_patterns
+from logdetective.models import SkipSnippets
 
 LOG = logging.getLogger("logdetective")
 
@@ -13,7 +14,13 @@ LOG = logging.getLogger("logdetective")
 class DrainExtractor:
     """A class that extracts information from logs using a template miner algorithm."""
 
-    def __init__(self, verbose: bool = False, context: bool = False, max_clusters=8):
+    def __init__(
+        self,
+        verbose: bool = False,
+        context: bool = False,
+        max_clusters=8,
+        skip_snippets: SkipSnippets = SkipSnippets({}),
+    ):
         config = TemplateMinerConfig()
         config.load(f"{os.path.dirname(__file__)}/drain3.ini")
         config.profiling_enabled = verbose
@@ -21,22 +28,28 @@ class DrainExtractor:
         self.miner = drain3.TemplateMiner(config=config)
         self.verbose = verbose
         self.context = context
+        self.skip_snippets = skip_snippets
 
     def __call__(self, log: str) -> list[Tuple[int, str]]:
         out = []
+        # Create chunks
+        chunks = list(get_chunks(log))
+        # Keep only chunks that don't match any of the excluded patterns
+        chunks = [
+            (_, chunk)
+            for _, chunk in chunks
+            if not filter_snippet_patterns(chunk, self.skip_snippets)
+        ]
         # First pass create clusters
-        for _, chunk in get_chunks(log):
+        for _, chunk in chunks:
             processed_chunk = self.miner.add_log_message(chunk)
             LOG.debug(processed_chunk)
-        # Sort found clusters by size, descending order
-        sorted_clusters = sorted(
-            self.miner.drain.clusters, key=lambda it: it.size, reverse=True
-        )
+        clusters = list(self.miner.drain.clusters)
         # Second pass, only matching lines with clusters,
         # to recover original text
-        for chunk_start, chunk in get_chunks(log):
+        for chunk_start, chunk in chunks:
             cluster = self.miner.match(chunk, "always")
-            if cluster in sorted_clusters:
+            if cluster in clusters:
                 out.append((chunk_start, chunk))
-                sorted_clusters.remove(cluster)
+                clusters.remove(cluster)
         return out

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/logdetective.py

@@ -14,6 +14,7 @@ from logdetective.utils import (
     format_snippets,
     compute_certainty,
     load_prompts,
+    load_skip_snippet_patterns,
 )
 from logdetective.extractors import DrainExtractor
 
@@ -82,6 +83,12 @@ def setup_args():
         default=DEFAULT_TEMPERATURE,
         help="Temperature for inference.",
     )
+    parser.add_argument(
+        "--skip_snippets",
+        type=str,
+        default=f"{os.path.dirname(__file__)}/skip_snippets.yml",
+        help="Path to patterns for skipping snippets.",
+    )
     return parser.parse_args()
 
 
@@ -120,9 +127,18 @@ async def run():  # pylint: disable=too-many-statements,too-many-locals
         LOG.error("You likely do not have enough memory to load the AI model")
         sys.exit(3)
 
+    try:
+        skip_snippets = load_skip_snippet_patterns(args.skip_snippets)
+    except OSError as e:
+        LOG.error(e)
+        sys.exit(5)
+
     # Log file summarizer initialization
     extractor = DrainExtractor(
-        args.verbose > 1, context=True, max_clusters=args.n_clusters
+        args.verbose > 1,
+        context=True,
+        max_clusters=args.n_clusters,
+        skip_snippets=skip_snippets,
     )
 
     LOG.info("Getting summary")

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/models.py

@@ -1,5 +1,6 @@
+import re
 from typing import Optional
-from pydantic import BaseModel
+from pydantic import BaseModel, model_validator
 
 from logdetective.constants import (
     PROMPT_TEMPLATE,
@@ -40,3 +41,33 @@ class PromptConfig(BaseModel):
         self.staged_system_prompt = data.get(
             "staged_system_prompt", DEFAULT_SYSTEM_PROMPT
         )
+
+
+class SkipSnippets(BaseModel):
+    """Regular expressions defining snippets we should not analyze"""
+
+    snippet_patterns: dict[str, re.Pattern] = {}
+
+    def __init__(self, data: Optional[dict] = None):
+        super().__init__(data=data)
+        if data is None:
+            return
+        self.snippet_patterns = {
+            key: re.compile(pattern) for key, pattern in data.items()
+        }
+
+    @model_validator(mode="before")
+    @classmethod
+    def check_patterns(cls, data: dict):
+        """Check if all supplied patterns are valid regular expressions.
+        Techically replicating what is done in __init__ but with nicer error message."""
+        patterns = data["data"]
+        for key, pattern in patterns.items():
+            try:
+                re.compile(pattern=pattern)
+            except (TypeError, re.error) as ex:
+                raise ValueError(
+                    f"Invalid pattern `{pattern}` with name `{key}` supplied for skipping in logs."
+                ) from ex
+
+        return data

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/remote_log.py

@@ -53,7 +53,7 @@ class RemoteLog:
             LOG.debug("process url %s", self.url)
             try:
                 response = await self._http_session.get(self.url, raise_for_status=True)
-            except aiohttp.ClientResponseError as ex:
+            except (aiohttp.ClientResponseError, aiohttp.ClientConnectorError) as ex:
                 raise RuntimeError(f"We couldn't obtain the logs: {ex}") from ex
             return await response.text()
         LOG.error("Invalid URL received ")

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/server/config.py

@@ -3,8 +3,9 @@ import logging
 import yaml
 from openai import AsyncOpenAI
 
-from logdetective.utils import load_prompts
+from logdetective.utils import load_prompts, load_skip_snippet_patterns
 from logdetective.server.models import Config, InferenceConfig
+import logdetective
 
 
 def load_server_config(path: str | None) -> Config:
@@ -60,9 +61,16 @@ def get_openai_api_client(ineference_config: InferenceConfig):
 
 SERVER_CONFIG_PATH = os.environ.get("LOGDETECTIVE_SERVER_CONF", None)
 SERVER_PROMPT_PATH = os.environ.get("LOGDETECTIVE_PROMPTS", None)
+# The default location for skip patterns is in the same directory
+# as logdetective __init__.py file.
+SERVER_SKIP_PATTERNS_PATH = os.environ.get(
+    "LOGDETECIVE_SKIP_PATTERNS",
+    f"{os.path.dirname(logdetective.__file__)}/skip_snippets.yml",
+)
 
 SERVER_CONFIG = load_server_config(SERVER_CONFIG_PATH)
 PROMPT_CONFIG = load_prompts(SERVER_PROMPT_PATH)
+SKIP_SNIPPETS_CONFIG = load_skip_snippet_patterns(SERVER_SKIP_PATTERNS_PATH)
 
 LOG = get_log(SERVER_CONFIG)
 

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/server/llm.py

@@ -16,7 +16,13 @@ from logdetective.utils import (
     compute_certainty,
     prompt_to_messages,
 )
-from logdetective.server.config import LOG, SERVER_CONFIG, PROMPT_CONFIG, CLIENT
+from logdetective.server.config import (
+    LOG,
+    SERVER_CONFIG,
+    PROMPT_CONFIG,
+    CLIENT,
+    SKIP_SNIPPETS_CONFIG,
+)
 from logdetective.server.models import (
     AnalyzedSnippet,
     InferenceConfig,
@@ -42,7 +48,10 @@ def format_analyzed_snippets(snippets: list[AnalyzedSnippet]) -> str:
 def mine_logs(log: str) -> List[Tuple[int, str]]:
     """Extract snippets from log text"""
     extractor = DrainExtractor(
-        verbose=True, context=True, max_clusters=SERVER_CONFIG.extractor.max_clusters
+        verbose=True,
+        context=True,
+        max_clusters=SERVER_CONFIG.extractor.max_clusters,
+        skip_snippets=SKIP_SNIPPETS_CONFIG,
     )
 
     LOG.info("Getting summary")

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/server/plot.py

@@ -2,12 +2,10 @@ import datetime
 from typing import Optional, Union, Dict
 
 import numpy
-import matplotlib
-import matplotlib.figure
-import matplotlib.pyplot
+from numpy.typing import ArrayLike
+from matplotlib import dates, colormaps, axes, pyplot, figure
 
-from matplotlib.pyplot import cm
-from logdetective.server import models
+from logdetective.server.models import TimePeriod
 from logdetective.server.database.models import (
     AnalyzeRequestMetrics,
     EndpointType,
@@ -18,25 +16,25 @@ from logdetective.server.database.models import (
 class Definition:
     """Define plot details, given a time period."""
 
-    def __init__(self, time_period: models.TimePeriod):
+    def __init__(self, time_period: TimePeriod):
         self.time_period = time_period
         self.days_diff = time_period.get_time_period().days
         if self.time_period.hours:
             self._freq = "H"
             self._time_format = "%Y-%m-%d %H"
-            self._locator = matplotlib.dates.HourLocator(interval=2)
+            self._locator = dates.HourLocator(interval=2)
             self._time_unit = "hour"
             self._time_delta = datetime.timedelta(hours=1)
         elif self.time_period.days:
             self._freq = "D"
             self._time_format = "%Y-%m-%d"
-            self._locator = matplotlib.dates.DayLocator(interval=1)
+            self._locator = dates.DayLocator(interval=1)
             self._time_unit = "day"
             self._time_delta = datetime.timedelta(days=1)
         elif self.time_period.weeks:
             self._freq = "W"
             self._time_format = "%Y-%m-%d"
-            self._locator = matplotlib.dates.WeekdayLocator(interval=1)
+            self._locator = dates.WeekdayLocator(interval=1)
             self._time_unit = "week"
             self._time_delta = datetime.timedelta(weeks=1)
 
@@ -120,10 +118,10 @@ def create_time_series_arrays(
 
 
 def _add_bar_chart(
-    ax: matplotlib.figure.Axes,
+    ax: axes.Axes,
     plot_def: Definition,
-    timestamps: numpy.array,
-    values: numpy.array,
+    timestamps: ArrayLike,
+    values: ArrayLike,
     label: str,
 ) -> None:
     """Add a blue bar chart"""
@@ -142,18 +140,18 @@ def _add_bar_chart(
     ax.set_ylabel(label, color="blue")
     ax.tick_params(axis="y", labelcolor="blue")
 
-    ax.xaxis.set_major_formatter(matplotlib.dates.DateFormatter(plot_def.time_format))
+    ax.xaxis.set_major_formatter(dates.DateFormatter(plot_def.time_format))
     ax.xaxis.set_major_locator(plot_def.locator)
 
-    matplotlib.pyplot.xticks(rotation=45)
+    pyplot.xticks(rotation=45)
 
     ax.grid(True, alpha=0.3)
 
 
 def _add_line_chart(  # pylint: disable=too-many-arguments disable=too-many-positional-arguments
-    ax: matplotlib.figure.Axes,
-    timestamps: numpy.array,
-    values: numpy.array,
+    ax: axes.Axes,
+    timestamps: ArrayLike,
+    values: ArrayLike,
     label: str,
     color: str = "red",
     set_label: bool = True,
@@ -166,10 +164,10 @@ def _add_line_chart(  # pylint: disable=too-many-arguments disable=too-many-posi
 
 
 def requests_per_time(
-    period_of_time: models.TimePeriod,
+    period_of_time: TimePeriod,
     endpoint: EndpointType = EndpointType.ANALYZE,
     end_time: Optional[datetime.datetime] = None,
-) -> matplotlib.figure.Figure:
+) -> figure.Figure:
     """
     Generate a visualization of request counts over a specified time period.
 
@@ -200,13 +198,13 @@ def requests_per_time(
         requests_counts, plot_def, start_time, end_time
     )
 
-    fig, ax1 = matplotlib.pyplot.subplots(figsize=(12, 6))
+    fig, ax1 = pyplot.subplots(figsize=(12, 6))
     _add_bar_chart(ax1, plot_def, timestamps, counts, "Requests")
 
     ax2 = ax1.twinx()
     _add_line_chart(ax2, timestamps, numpy.cumsum(counts), "Cumulative Requests")
 
-    matplotlib.pyplot.title(
+    pyplot.title(
         f"Requests received for API {endpoint} ({start_time.strftime(plot_def.time_format)} "
         f"to {end_time.strftime(plot_def.time_format)})"
     )
@@ -215,16 +213,16 @@ def requests_per_time(
     lines2, labels2 = ax2.get_legend_handles_labels()
     ax1.legend(lines1 + lines2, labels1 + labels2, loc="center")
 
-    matplotlib.pyplot.tight_layout()
+    pyplot.tight_layout()
 
     return fig
 
 
 def average_time_per_responses(  # pylint: disable=too-many-locals
-    period_of_time: models.TimePeriod,
+    period_of_time: TimePeriod,
     endpoint: EndpointType = EndpointType.ANALYZE,
     end_time: Optional[datetime.datetime] = None,
-) -> matplotlib.figure.Figure:
+) -> figure.Figure:
     """
     Generate a visualization of average response time and length over a specified time period.
 
@@ -259,7 +257,7 @@ def average_time_per_responses(  # pylint: disable=too-many-locals
         float,
     )
 
-    fig, ax1 = matplotlib.pyplot.subplots(figsize=(12, 6))
+    fig, ax1 = pyplot.subplots(figsize=(12, 6))
     _add_bar_chart(
         ax1, plot_def, timestamps, average_time, "average response time (seconds)"
     )
@@ -280,7 +278,7 @@ def average_time_per_responses(  # pylint: disable=too-many-locals
     ax2 = ax1.twinx()
     _add_line_chart(ax2, timestamps, average_length, "average response length (chars)")
 
-    matplotlib.pyplot.title(
+    pyplot.title(
         f"average response time for API {endpoint} ({start_time.strftime(plot_def.time_format)} "
         f"to {end_time.strftime(plot_def.time_format)})"
     )
@@ -289,7 +287,7 @@ def average_time_per_responses(  # pylint: disable=too-many-locals
     lines2, labels2 = ax2.get_legend_handles_labels()
     ax1.legend(lines1 + lines2, labels1 + labels2, loc="center")
 
-    matplotlib.pyplot.tight_layout()
+    pyplot.tight_layout()
 
     return fig
 
@@ -322,7 +320,7 @@ def _collect_emoji_data(
 
 
 def _plot_emoji_data(  # pylint: disable=too-many-locals
-    ax: matplotlib.figure.Axes,
+    ax: axes.Axes,
     reactions_values_dict: Dict[str, Dict[datetime.datetime, int]],
     plot_def: Definition,
     start_time: datetime.datetime,
@@ -340,7 +338,10 @@ def _plot_emoji_data(  # pylint: disable=too-many-locals
         )
         all_counts.extend(counts)
 
-    colors = [cm.viridis(i) for i in numpy.linspace(0, 1, len(reactions_values_dict))]  # pylint: disable=no-member
+    colors = [
+        colormaps["viridis"](i)
+        for i in numpy.linspace(0, 1, len(reactions_values_dict))
+    ]
 
     first_emoji = True
     for i, (emoji, dict_counts) in enumerate(reactions_values_dict.items()):
@@ -369,9 +370,9 @@ def _plot_emoji_data(  # pylint: disable=too-many-locals
 
 
 def emojis_per_time(
-    period_of_time: models.TimePeriod,
+    period_of_time: TimePeriod,
     end_time: Optional[datetime.datetime] = None,
-) -> matplotlib.figure.Figure:
+) -> figure.Figure:
     """
     Generate a visualization of overall emoji feedback
     over a specified time period.
@@ -396,13 +397,13 @@ def emojis_per_time(
     start_time = period_of_time.get_period_start_time(end_time)
     reactions_values_dict = _collect_emoji_data(start_time, plot_def)
 
-    fig, ax = matplotlib.pyplot.subplots(figsize=(12, 6))
+    fig, ax = pyplot.subplots(figsize=(12, 6))
 
     emoji_lines, emoji_labels = _plot_emoji_data(
         ax, reactions_values_dict, plot_def, start_time, end_time
     )
 
-    matplotlib.pyplot.title(
+    pyplot.title(
         f"Emoji feedback ({start_time.strftime(plot_def.time_format)} "
         f"to {end_time.strftime(plot_def.time_format)})"
     )
@@ -419,11 +420,11 @@ def emojis_per_time(
     ax.set_ylabel("Count")
 
     # Format x-axis
-    ax.xaxis.set_major_formatter(matplotlib.dates.DateFormatter(plot_def.time_format))
+    ax.xaxis.set_major_formatter(dates.DateFormatter(plot_def.time_format))
     ax.xaxis.set_major_locator(plot_def.locator)
     ax.tick_params(axis="x", labelrotation=45)
     ax.grid(True, alpha=0.3)
 
-    matplotlib.pyplot.tight_layout()
+    pyplot.tight_layout()
 
     return fig

logdetective-1.6.0/logdetective/skip_snippets.yml

@@ -0,0 +1,12 @@
+# This file holds patterns you want to skip during log parsing.
+# By default, no patterns are supplied.
+# Patterns are to be specified as values of dictionary,
+# with each key being a descriptive name of the pattern.
+# Patterns themselves are evaluated as a regular expression.
+# Make sure to avoid regular expressions that may be interpreted
+# as yaml syntax.
+# Example:
+
+# contains_capital_a: "^.*A.*"
+# starts_with_numeric: "^[0-9].*"
+child_exit_code_zero: "Child return code was: 0"

{logdetective-1.5.0 → logdetective-1.6.0}/logdetective/utils.py

@@ -8,7 +8,7 @@ import numpy as np
 import yaml
 
 from llama_cpp import Llama, CreateCompletionResponse, CreateCompletionStreamResponse
-from logdetective.models import PromptConfig
+from logdetective.models import PromptConfig, SkipSnippets
 from logdetective.remote_log import RemoteLog
 
 
@@ -223,3 +223,27 @@ def prompt_to_messages(
         ]
 
     return messages
+
+
+def filter_snippet_patterns(snippet: str, skip_snippets: SkipSnippets) -> bool:
+    """Try to match snippet agains provided patterns to determine if we should
+    filter it out or not."""
+    for key, pattern in skip_snippets.snippet_patterns.items():
+        if pattern.match(snippet):
+            LOG.debug("Snippet `%s` has matched agains skip pattern %s", snippet, key)
+            return True
+
+    return False
+
+
+def load_skip_snippet_patterns(path: str | None) -> SkipSnippets:
+    """Load dictionary of snippet patterns we want to skip."""
+    if path:
+        try:
+            with open(path, "r") as file:
+                return SkipSnippets(yaml.safe_load(file))
+        except OSError as e:
+            LOG.error("Couldn't open file with snippet skip patterns `%s`", path)
+            raise e
+
+    return SkipSnippets({})

{logdetective-1.5.0 → logdetective-1.6.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "logdetective"
-version = "1.5.0"
+version = "1.6.0"
 description = "Log using LLM AI to search for build/test failures and provide ideas for fixing these."
 authors = ["Jiri Podivin <jpodivin@gmail.com>"]
 license = "Apache-2.0"