logdetective 0.4.0__py3-none-any.whl → 2.11.0__py3-none-any.whl

logdetective/server/templates/base_response.html.j2

@@ -0,0 +1,59 @@
+<b>Log Detective report
+    {% if project %}for {{ project }}{% endif %}
+</b>
+<p>
+    Analysis:
+    <br>
+    {{ explanation }}
+    <br>
+    <details>
+        <summary>Analysis details</summary>
+        <ul>
+            {% for snippet in snippets %}
+                <li>
+                    <b>Line {{ snippet.line_number }}:</b> <code>{{ snippet.text }}</code>
+                    {{ snippet.explanation.text }}
+                </li>
+            {% endfor %}
+        </ul>
+    </details>
+</p>
+<br>
+<details>
+    <summary>Help</summary>
+    <p>
+        Log Detective analyses logs using template mining algorithm and LLMs.
+        Providing overview of the primary issue, while highlighting potentially important log lines.
+    </p>
+    <p>
+        Log Detective may offer potential solution to problem.
+        However, it is designed to do so only when sufficiently confident in the response.
+        <br>
+        Even so, as with any LLM based AI system, recommendations of Log Detective must be taken with caution.
+        Before acting on any recommendations of Log Detective, consider following:
+    </p>
+    <ul>
+        <li>
+            <b>Do you understand how does the proposed solution work? And why should it solve your problem?</b>
+        </li>
+        <li>Did Log Detective work with sufficient information to make this suggestion?</li>
+        <li>Are the suggested actions safe in the context of your work?</li>
+        <li>Do results of snippet analysis align with results of final analysis?</li>
+        <li>Are there any potentially adverse outcomes from implementing proposed solution?</li>
+    </ul>
+</details>
+<br>
+{% render_contacts %}
+<details>
+    <summary>Contacts</summary>
+    <p>Don't hesitate to reach out with your questions or suggestions.</p>
+    <ul>
+        <li>
+            <a href="https://github.com/fedora-copr/logdetective">Source</a>
+        </li>
+        <li>
+            <a href="https://github.com/fedora-copr/logdetective/issues">Issue tracker</a>
+        </li>
+    </ul>
+</details>
+{% endif %}

logdetective/server/templates/gitlab_full_comment.md.j2

@@ -0,0 +1,73 @@
+The package {{ package }} build has experienced an issue.
+Please know that the explanation was provided by AI and may be incorrect.
+{% if certainty > 0 %}
+    In this case, we are {{ "%.2f" | format(certainty) }}% certain of the response {{ emoji_face }}.
+{% endif %}
+
+<details open>
+    <summary>Description</summary>
+    {{ explanation }}
+</details>
+<details>
+{#
+    Formatted so that we don't trigger GitLab markdown
+#}
+<summary>Snippets</summary>
+<ul>
+{% for snippet in snippets -%}
+<li><div><b>Line {{ snippet.line_number }}:</b> <code>{{ snippet.text | e }}</code><br>{{ snippet.explanation.text | e }}</div></li>
+{%- endfor %}
+</ul>
+</details>
+<details>
+    <summary>Logs</summary>
+    <p>Log Detective analyzed the following logs files to provide an explanation:</p>
+    <ul>
+        <li>
+            <a href="{{ log_url }}">{{ log_url }}</a>
+        </li>
+    </ul>
+    <p>
+        Additional logs are available from:
+        <ul>
+            <li>
+                <a href="{{ artifacts_url }}">artifacts.zip</a>
+            </li>
+        </ul>
+    </p>
+    <p>
+        Please know that these log files are automatically removed after some
+        time, so you might need a backup.
+    </p>
+</details>
+<details>
+    <summary>Help</summary>
+    <p>Don't hesitate to reach out.</p>
+    <ul>
+        <li>
+            <a href="https://github.com/fedora-copr/logdetective">Upstream</a>
+        </li>
+        <li>
+            <a href="https://github.com/fedora-copr/logdetective/issues">Issue tracker</a>
+        </li>
+        <li>
+            <a href="https://redhat.enterprise.slack.com/archives/C06DWNVKKDE">Slack</a>
+        </li>
+        <li>
+            <a href="https://log-detective.com/documentation">Documentation</a>
+        </li>
+    </ul>
+</details>
+
+<hr>
+
+This explanation was provided by AI (<a href="https://logdetective.com">Log Detective</a>).
+Always review AI generated content prior to use.
+Was the provided feedback accurate and helpful?
+<br>
+Please vote with :thumbsup:
+or :thumbsdown: to help us improve.
+<br>
+<i>If this Log Detective report contains harmful content,
+please use the <a href="https://docs.gitlab.com/user/report_abuse/">Gitlab reporting feature for harmful content</a>
+and contact the <a href="https://github.com/fedora-copr/logdetective/issues">Log Detective developers</a>.</i>

logdetective/server/templates/gitlab_short_comment.md.j2

@@ -0,0 +1,62 @@
+The package {{ package }} build has experienced an issue.
+Please know that the explanation was provided by AI and may be incorrect.
+{% if certainty > 0 %}
+    In this case, we are {{ "%.2f" | format(certainty) }}% certain of the response {{ emoji_face }}.
+{% endif %}
+
+<details open>
+    <summary>Description</summary>
+    {{ explanation }}
+</details>
+<details>
+    <summary>Logs</summary>
+    <p>Log Detective analyzed the following logs files to provide an explanation:</p>
+    <ul>
+        <li>
+            <a href="{{ log_url }}">{{ log_url }}</a>
+        </li>
+    </ul>
+    <p>
+        Additional logs are available from:
+        <ul>
+            <li>
+                <a href="{{ artifacts_url }}">artifacts.zip</a>
+            </li>
+        </ul>
+    </p>
+    <p>
+        Please know that these log files are automatically removed after some
+        time, so you might need a backup.
+    </p>
+</details>
+<details>
+    <summary>Help</summary>
+    <p>Don't hesitate to reach out.</p>
+    <ul>
+        <li>
+            <a href="https://github.com/fedora-copr/logdetective">Upstream</a>
+        </li>
+        <li>
+            <a href="https://github.com/fedora-copr/logdetective/issues">Issue tracker</a>
+        </li>
+        <li>
+            <a href="https://redhat.enterprise.slack.com/archives/C06DWNVKKDE">Slack</a>
+        </li>
+        <li>
+            <a href="https://log-detective.com/documentation">Documentation</a>
+        </li>
+    </ul>
+</details>
+
+<hr>
+
+This explanation was provided by AI (<a href="https://logdetective.com">Log Detective</a>).
+Always review AI generated content prior to use.
+Was the provided feedback accurate and helpful?
+<br>
+Please vote with :thumbsup:
+or :thumbsdown: to help us improve.
+<br>
+<i>If this Log Detective report contains harmful content,
+please use the <a href="https://docs.gitlab.com/user/report_abuse/">Gitlab reporting feature for harmful content</a>
+and contact the <a href="https://github.com/fedora-copr/logdetective/issues">Log Detective developers</a>.</i>

logdetective/server/utils.py

@@ -1,44 +1,110 @@
-import logging
-import yaml
-from logdetective.server.models import Config
+from typing import List
+from importlib.metadata import version
 
+import aiohttp
+from fastapi import HTTPException
 
-def load_server_config(path: str | None) -> Config:
-    """Load configuration file for logdetective server.
-    If no path was provided, or if the file doesn't exist, return defaults.
+from logdetective.constants import SNIPPET_DELIMITER
+from logdetective.server.config import LOG
+from logdetective.server.exceptions import LogDetectiveConnectionError
+from logdetective.server.models import AnalyzedSnippet, RatedSnippetAnalysis
+
+
+def format_analyzed_snippets(snippets: list[AnalyzedSnippet]) -> str:
+    """Format snippets for submission into staged prompt."""
+    summary = f"\n{SNIPPET_DELIMITER}\n".join(
+        [f"[{e.text}] at line [{e.line_number}]: [{e.explanation}]" for e in snippets]
+    )
+    return summary
+
+
+def connection_error_giveup(details: dict) -> None:
+    """Too many connection errors, give up.
     """
-    if path is not None:
-        try:
-            with open(path, "r") as config_file:
-                return Config(yaml.safe_load(config_file))
-        except FileNotFoundError:
-            pass
-    return Config()
+    LOG.error("Too many connection errors, giving up. %s", details["exception"])
+    raise LogDetectiveConnectionError() from details["exception"]
+
 
+def should_we_giveup(exc: aiohttp.ClientResponseError) -> bool:
+    """From backoff's docs:
 
-def get_log(config: Config):
+    > a function which accepts the exception and returns
+    > a truthy value if the exception should not be retried
     """
-    Initialize a logger for this server
+    LOG.info("Should we give up on retrying error %s", exc)
+    return exc.status < 400
+
+
+def we_give_up(details: dict):
+    """Retries didn't work (or we got a different exc)
+    we give up and raise proper 500 for our API endpoint
     """
-    log = logging.getLogger(config.log.name)
-    if getattr(log, "initialized", False):
-        return log
+    LOG.error("Last exception: %s", details["exception"])
+    LOG.error("Inference error: %s", details["args"])
+    raise HTTPException(500, "Request to the inference API failed")
+
+
+def select_relevance(snippet: AnalyzedSnippet) -> float:
+    """Retrieve relevance value from structure, if there is one."""
+    if not isinstance(snippet.explanation, RatedSnippetAnalysis):
+        LOG.exception("Only rated snippets can be ordered by relevance.")
+        raise ValueError
+    return snippet.explanation.relevance
+
+
+def select_line_number(explanation: AnalyzedSnippet) -> int:
+    """Returns line number of original snippet."""
+    return explanation.line_number
+
+
+def filter_snippets(
+    processed_snippets: List[AnalyzedSnippet], top_k: int
+) -> List[AnalyzedSnippet]:
+    """Filter snippets according to criteria in config while keeping them ordered by line number.
+    If all snippets recieved the same score, return them all.
+    AnalyzedSnippet objects must have `explanation` attribute set to `RatedSnippetAnalysis`,
+    otherwise raise `ValueError`."""
+
+    if top_k >= len(processed_snippets):
+        LOG.warning(
+            "The `top-k` parameter >= number of original snippets, skipping filtering."
+        )
+        return processed_snippets
+
+    # Sorting invokes `select_relevance` which also tests if objects actually
+    # have the score assigned. Otherwise it raises exception.
+    processed_snippets = sorted(processed_snippets, key=select_relevance, reverse=True)
+
+    # Check for failure mode when all snippets have
+    # the same relevance. In such cases there is no point in filtering
+    # and all snippets are returned.
+    max_relevance = processed_snippets[0].explanation.relevance
+    min_relevance = processed_snippets[-1].explanation.relevance
+
+    LOG.info(
+        "Analyzed snippets sorted. Max relevance: %d Min relevance: %e",
+        max_relevance,
+        min_relevance,
+    )
+    if max_relevance == min_relevance:
+        LOG.warning("All snippets recieved the same rating. Filtering disabled.")
+        return processed_snippets
+
+    processed_snippets = processed_snippets[:top_k]
+
+    # Re-sorting snippets by line number
+    processed_snippets = sorted(processed_snippets, key=select_line_number)
+
+    return processed_snippets
 
-    log.setLevel(config.log.level)
 
-    # Drop the default handler, we will create it ourselves
-    log.handlers = []
+def construct_final_prompt(formatted_snippets: str, prompt_template: str) -> str:
+    """Create final prompt from processed snippets and csgrep output, if it is available."""
 
-    # STDOUT
-    stream_handler = logging.StreamHandler()
-    stream_handler.setFormatter(logging.Formatter(config.log.format))
-    log.addHandler(stream_handler)
+    final_prompt = prompt_template.format(formatted_snippets)
+    return final_prompt
 
-    # Log to file
-    if config.log.path:
-        file_handler = logging.FileHandler(config.log.path)
-        file_handler.setFormatter(logging.Formatter(config.log.format))
-        log.addHandler(file_handler)
 
-    log.initialized = True
-    return log
+def get_version() -> str:
+    """Obtain the version number using importlib"""
+    return version('logdetective')

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"