logdetective 2.6.0__tar.gz → 2.7.0__tar.gz

{logdetective-2.6.0 → logdetective-2.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: logdetective
-Version: 2.6.0
+Version: 2.7.0
 Summary: Log using LLM AI to search for build/test failures and provide ideas for fixing these.
 License: Apache-2.0
 License-File: LICENSE

{logdetective-2.6.0 → logdetective-2.7.0}/logdetective/logdetective.py

@@ -174,11 +174,6 @@ async def run():  # pylint: disable=too-many-statements,too-many-locals,too-many
     log_summary = format_snippets(log_summary)
     LOG.info("Log summary: \n %s", log_summary)
 
-    prompt = (
-        f"{prompts_configuration.default_system_prompt}\n"
-        f"{prompts_configuration.prompt_template}"
-    )
-
     stream = True
     if args.no_stream:
         stream = False
@@ -186,30 +181,38 @@ async def run():  # pylint: disable=too-many-statements,too-many-locals,too-many
         log_summary,
         model,
         stream,
-        prompt_template=prompt,
+        prompt_templates=prompts_configuration,
         temperature=args.temperature,
     )
     probs = []
     print("Explanation:")
     # We need to extract top token probability from the response
-    # CreateCompletionResponse structure of llama-cpp-python.
+    # CreateChatCompletionResponse structure of llama-cpp-python.
     # `compute_certainty` function expects list of dictionaries with form
     # { 'logprob': <float> } as expected from the OpenAI API.
 
     if args.no_stream:
-        print(response["choices"][0]["text"])
+        print(response["choices"][0]["message"]["content"])
         probs = [
-            {"logprob": e} for e in response["choices"][0]["logprobs"]["token_logprobs"]
+            {"logprob": e["logprob"]} for e in response["choices"][0]["logprobs"]["content"]
         ]
 
     else:
         # Stream the output
         for chunk in response:
+            # What might happen, is that first (or possibly any other) chunk may not contain
+            # fields choices[0].delta.content or choices[0].logprobs -> if so, we just skip it
+            if any([
+                'content' not in chunk["choices"][0]["delta"],
+                'logprobs' not in chunk["choices"][0]
+            ]):
+                continue
+
             if isinstance(chunk["choices"][0]["logprobs"], dict):
                 probs.append(
-                    {"logprob": chunk["choices"][0]["logprobs"]["token_logprobs"][0]}
+                    {"logprob": chunk["choices"][0]["logprobs"]["content"][0]["logprob"]}
                 )
-            delta = chunk["choices"][0]["text"]
+            delta = chunk["choices"][0]["delta"]["content"]
             print(delta, end="", flush=True)
     certainty = compute_certainty(probs)
 

{logdetective-2.6.0 → logdetective-2.7.0}/logdetective/utils.py

@@ -8,7 +8,11 @@ import aiohttp
 import numpy as np
 import yaml
 
-from llama_cpp import Llama, CreateCompletionResponse, CreateCompletionStreamResponse
+from llama_cpp import (
+    Llama,
+    CreateChatCompletionResponse,
+    CreateChatCompletionStreamResponse,
+)
 from logdetective.constants import SNIPPET_DELIMITER
 from logdetective.models import PromptConfig, SkipSnippets
 from logdetective.remote_log import RemoteLog
@@ -123,8 +127,8 @@ def compute_certainty(probs: List[Dict]) -> float:
 
 
 def process_log(
-    log: str, model: Llama, stream: bool, prompt_template: str, temperature: float
-) -> CreateCompletionResponse | Iterator[CreateCompletionStreamResponse]:
+    log: str, model: Llama, stream: bool, prompt_templates: PromptConfig, temperature: float
+) -> CreateChatCompletionResponse | Iterator[CreateChatCompletionStreamResponse]:
     """Processes a given log using the provided language model and returns its summary.
 
     Args:
@@ -136,11 +140,23 @@ def process_log(
     Returns:
         str: The summary of the given log generated by the language model.
     """
-    response = model(
-        prompt=prompt_template.format(log),
+    messages = [
+        {
+            "role": "system",
+            "content": prompt_templates.default_system_prompt
+        },
+        {
+            "role": "user",
+            "content": prompt_templates.prompt_template.format(log)
+        },
+    ]
+
+    response = model.create_chat_completion(
+        messages=messages,
         stream=stream,
         max_tokens=0,
-        logprobs=1,
+        logprobs=True,
+        top_logprobs=1,
         temperature=temperature,
     )
 

{logdetective-2.6.0 → logdetective-2.7.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "logdetective"
-version = "2.6.0"
+version = "2.7.0"
 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"