logdetective 1.2.0__tar.gz → 1.3.0__tar.gz

{logdetective-1.2.0 → logdetective-1.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: logdetective
-Version: 1.2.0
+Version: 1.3.0
 Summary: Log using LLM AI to search for build/test failures and provide ideas for fixing these.
 License: Apache-2.0
 Author: Jiri Podivin
@@ -30,6 +30,7 @@ Requires-Dist: huggingface-hub (>0.23.2)
 Requires-Dist: llama-cpp-python (>0.2.56,!=0.2.86)
 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)

{logdetective-1.2.0 → logdetective-1.3.0}/logdetective/server/config.py

@@ -1,8 +1,10 @@
 import os
 import logging
 import yaml
+from openai import AsyncOpenAI
+
 from logdetective.utils import load_prompts
-from logdetective.server.models import Config
+from logdetective.server.models import Config, InferenceConfig
 
 
 def load_server_config(path: str | None) -> Config:
@@ -49,6 +51,14 @@ def get_log(config: Config):
     return log
 
 
+def get_openai_api_client(ineference_config: InferenceConfig):
+    """Set up AsyncOpenAI client with default configuration.
+    """
+    return AsyncOpenAI(
+        api_key=ineference_config.api_token,
+        base_url=ineference_config.url)
+
+
 SERVER_CONFIG_PATH = os.environ.get("LOGDETECTIVE_SERVER_CONF", None)
 SERVER_PROMPT_PATH = os.environ.get("LOGDETECTIVE_PROMPTS", None)
 
@@ -56,3 +66,5 @@ SERVER_CONFIG = load_server_config(SERVER_CONFIG_PATH)
 PROMPT_CONFIG = load_prompts(SERVER_PROMPT_PATH)
 
 LOG = get_log(SERVER_CONFIG)
+
+CLIENT = get_openai_api_client(SERVER_CONFIG.inference)

{logdetective-1.2.0 → logdetective-1.3.0}/logdetective/server/emoji.py

@@ -44,7 +44,7 @@ async def _handle_gitlab_operation(func: Callable, *args):
     """
     try:
         return await asyncio.to_thread(func, *args)
-    except gitlab.GitlabError as e:
+    except (gitlab.GitlabError, gitlab.GitlabGetError) as e:
         log_msg = f"Error during GitLab operation {func}{args}: {e}"
         if "Not Found" in str(e):
             LOG.error(log_msg)
@@ -64,6 +64,8 @@ async def collect_emojis_in_comments(  # pylint: disable=too-many-locals
     mrs = {}
     for comment in comments:
         mr_job_db = GitlabMergeRequestJobs.get_by_id(comment.merge_request_job_id)
+        if not mr_job_db:
+            continue
         if mr_job_db.id not in projects:
             projects[mr_job_db.id] = project = await _handle_gitlab_operation(
                 gitlab_conn.projects.get, mr_job_db.project_id

{logdetective-1.2.0 → logdetective-1.3.0}/logdetective/server/llm.py

@@ -1,21 +1,21 @@
 import os
 import asyncio
-import json
 import random
-from typing import List, Tuple, Dict, Any, Union
+from typing import List, Tuple, Union
 
 import backoff
-from aiohttp import StreamReader
 from fastapi import HTTPException
 
 import aiohttp
+from openai import AsyncStream
+from openai.types.chat import ChatCompletionChunk
 
 from logdetective.constants import SNIPPET_DELIMITER
 from logdetective.extractors import DrainExtractor
 from logdetective.utils import (
     compute_certainty,
 )
-from logdetective.server.config import LOG, SERVER_CONFIG, PROMPT_CONFIG
+from logdetective.server.config import LOG, SERVER_CONFIG, PROMPT_CONFIG, CLIENT
 from logdetective.server.models import (
     AnalyzedSnippet,
     InferenceConfig,
@@ -54,59 +54,6 @@ def mine_logs(log: str) -> List[Tuple[int, str]]:
     return log_summary
 
 
-async def submit_to_llm_endpoint(
-    url_path: str,
-    data: Dict[str, Any],
-    headers: Dict[str, str],
-    stream: bool,
-    inference_cfg: InferenceConfig = SERVER_CONFIG.inference,
-) -> Any:
-    """Send request to an API endpoint. Verifying successful request unless
-    the using the stream response.
-
-    url_path: The endpoint path to query. (e.g. "/v1/chat/completions"). It should
-    not include the scheme and netloc of the URL, which is stored in the
-    InferenceConfig.
-    data:
-    headers:
-    stream:
-    inference_cfg: An InferenceConfig object containing the URL, max_tokens
-    and other relevant configuration for talking to an inference server.
-    """
-    async with inference_cfg.get_limiter():
-        LOG.debug("async request %s headers=%s data=%s", url_path, headers, data)
-        session = inference_cfg.get_http_session()
-
-        if inference_cfg.api_token:
-            headers["Authorization"] = f"Bearer {inference_cfg.api_token}"
-
-        response = await session.post(
-            url_path,
-            headers=headers,
-            # we need to use the `json=` parameter here and let aiohttp
-            # handle the json-encoding
-            json=data,
-            timeout=int(LLM_CPP_SERVER_TIMEOUT),
-            # Docs says chunked takes int, but:
-            #   DeprecationWarning: Chunk size is deprecated #1615
-            # So let's make sure we either put True or None here
-            chunked=True if stream else None,
-            raise_for_status=True,
-        )
-        if stream:
-            return response
-        try:
-            return json.loads(await response.text())
-        except UnicodeDecodeError as ex:
-            LOG.error(
-                "Error encountered while parsing llama server response: %s", ex
-            )
-            raise HTTPException(
-                status_code=400,
-                detail=f"Couldn't parse the response.\nError: {ex}\nData: {response.text}",
-            ) from ex
-
-
 def should_we_giveup(exc: aiohttp.ClientResponseError) -> bool:
     """
     From backoff's docs:
@@ -141,7 +88,7 @@ async def submit_text(
     text: str,
     inference_cfg: InferenceConfig,
     stream: bool = False,
-) -> Union[Explanation, StreamReader]:
+) -> Union[Explanation, AsyncStream[ChatCompletionChunk]]:
     """Submit prompt to LLM.
     inference_cfg: The configuration section from the config.json representing
     the relevant inference server for this request.
@@ -149,40 +96,36 @@ async def submit_text(
     """
     LOG.info("Analyzing the text")
 
-    headers = {"Content-Type": "application/json"}
-
-    if SERVER_CONFIG.inference.api_token:
-        headers["Authorization"] = f"Bearer {SERVER_CONFIG.inference.api_token}"
-
     LOG.info("Submitting to /v1/chat/completions endpoint")
 
-    data = {
-        "messages": [
-            {
-                "role": "user",
-                "content": text,
-            }
-        ],
-        "max_tokens": inference_cfg.max_tokens,
-        "logprobs": inference_cfg.log_probs,
-        "stream": stream,
-        "model": inference_cfg.model,
-        "temperature": inference_cfg.temperature,
-    }
-
-    response = await submit_to_llm_endpoint(
-        "/v1/chat/completions",
-        data,
-        headers,
-        inference_cfg=inference_cfg,
-        stream=stream,
-    )
+    async with inference_cfg.get_limiter():
+        response = await CLIENT.chat.completions.create(
+            messages=[
+                {
+                    "role": "user",
+                    "content": text,
+                }
+            ],
+            max_tokens=inference_cfg.max_tokens,
+            logprobs=inference_cfg.log_probs,
+            stream=stream,
+            model=inference_cfg.model,
+            temperature=inference_cfg.temperature,
+        )
 
-    if stream:
+    if isinstance(response, AsyncStream):
         return response
+    if not response.choices[0].message.content:
+        LOG.error("No response content recieved from %s", inference_cfg.url)
+        raise RuntimeError()
+    if response.choices[0].logprobs and response.choices[0].logprobs.content:
+        logprobs = [e.to_dict() for e in response.choices[0].logprobs.content]
+    else:
+        logprobs = None
+
     return Explanation(
-        text=response["choices"][0]["message"]["content"],
-        logprobs=response["choices"][0]["logprobs"]["content"],
+        text=response.choices[0].message.content,
+        logprobs=logprobs,
     )
 
 

{logdetective-1.2.0 → logdetective-1.3.0}/logdetective/server/models.py

@@ -136,7 +136,9 @@ class InferenceConfig(BaseModel):  # pylint: disable=too-many-instance-attribute
     max_tokens: int = -1
     log_probs: bool = True
     url: str = ""
-    api_token: str = ""
+    # OpenAI client library requires a string to be specified for API token
+    # even if it is not checked on the server side
+    api_token: str = "None"
     model: str = ""
     temperature: NonNegativeFloat = DEFAULT_TEMPERATURE
     max_queue_size: int = LLM_DEFAULT_MAX_QUEUE_SIZE
@@ -153,7 +155,7 @@ class InferenceConfig(BaseModel):  # pylint: disable=too-many-instance-attribute
         self.log_probs = data.get("log_probs", True)
         self.url = data.get("url", "")
         self.http_timeout = data.get("http_timeout", 5.0)
-        self.api_token = data.get("api_token", "")
+        self.api_token = data.get("api_token", "None")
         self.model = data.get("model", "default-model")
         self.temperature = data.get("temperature", DEFAULT_TEMPERATURE)
         self.max_queue_size = data.get("max_queue_size", LLM_DEFAULT_MAX_QUEUE_SIZE)

{logdetective-1.2.0 → logdetective-1.3.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "logdetective"
-version = "1.2.0"
+version = "1.3.0"
 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"
@@ -52,10 +52,11 @@ alembic = {version = "^1.13.3", optional = true }
 matplotlib = {version = "^3.8.4", optional = true }
 backoff = {version = "2.2.1", optional = true }
 sentry-sdk = {version = "^2.17.0", optional = true, extras = ["fastapi"]}
+openai = {version = "^1.82.1", optional = true}
 
 [tool.poetry.extras]
-server = ["fastapi", "sqlalchemy", "psycopg2", "alembic", "matplotlib", "backoff", "aiolimiter", "sentry-sdk"]
-server-testing = ["fastapi", "sqlalchemy", "psycopg2-binary", "alembic", "matplotlib", "backoff", "pytest-asyncio", "sentry-sdk"]
+server = ["fastapi", "sqlalchemy", "psycopg2", "alembic", "matplotlib", "backoff", "aiolimiter", "sentry-sdk", "openai"]
+server-testing = ["fastapi", "sqlalchemy", "psycopg2-binary", "alembic", "matplotlib", "backoff", "pytest-asyncio", "sentry-sdk", "openai"]
 
 [build-system]
 requires = ["poetry-core"]