Refactor/unify prompt loader (#319)

* unify prompt loader

* unify prompt loader part 2

* fix bugs and tests

* resolve conflicts + osa version increment

Author: fl1pcoin

osa_tool/aboutgen/about_generator.py

@@ -2,12 +2,12 @@
 import re
 from typing import List
 
-from osa_tool.aboutgen.prompts_about_config import PromptAboutLoader
 from osa_tool.config.settings import ConfigLoader
 from osa_tool.git_agent.git_agent import GitAgent
-from osa_tool.logger import logger
 from osa_tool.models.models import ModelHandler, ModelHandlerFactory
-from osa_tool.utils import extract_readme_content, parse_folder_name
+from osa_tool.utils.logger import logger
+from osa_tool.utils.prompts_builder import PromptLoader, PromptBuilder
+from osa_tool.utils.utils import extract_readme_content, parse_folder_name
 
 HOMEPAGE_KEYS = [
     "documentation",
@@ -24,16 +24,15 @@
 class AboutGenerator:
     """Generates Git repository About section content."""
 
-    def __init__(self, config_loader: ConfigLoader, git_agent: GitAgent):
+    def __init__(self, config_loader: ConfigLoader, prompts: PromptLoader, git_agent: GitAgent):
         self.config = config_loader.config
+        self.prompts = prompts
         self.model_handler: ModelHandler = ModelHandlerFactory.build(self.config)
         self.repo_url = self.config.git.repository
         self.metadata = git_agent.metadata
         self.base_path = os.path.join(os.getcwd(), parse_folder_name(self.repo_url))
         self.readme_content = extract_readme_content(self.base_path)
-        self.prompts = PromptAboutLoader().prompts
         self.validate_topics = git_agent.validate_topics
-
         self._content: dict | None = None
 
     def generate_about_content(self) -> None:
@@ -93,10 +92,13 @@ def generate_description(self) -> str:
             logger.warning("No README content found. Cannot generate description.")
             return ""
 
-        formatted_prompt = self.prompts.description.format(readme_content=self.readme_content)
+        prompt = PromptBuilder.render(
+            self.prompts.get("about_section.description"),
+            readme_content=self.readme_content,
+        )
 
         try:
-            description = self.model_handler.send_request(formatted_prompt)
+            description = self.model_handler.send_request(prompt)
             logger.debug(f"Generated description: {description}")
             return description[:350]
         except Exception as e:
@@ -124,14 +126,15 @@ def generate_topics(self, amount: int = 7) -> List[str]:
                 logger.warning(f"{amount} topics already exist in the metadata. Skipping generation.")
                 return existing_topics
 
-        formatted_prompt = self.prompts.topics.format(
+        prompt = PromptBuilder.render(
+            self.prompts.get("about_section.topics"),
+            readme_content=self.readme_content,
             amount=amount,
             topics=existing_topics,
-            readme_content=self.readme_content,
         )
 
         try:
-            response = self.model_handler.send_request(formatted_prompt)
+            response = self.model_handler.send_request(prompt)
             topics = [topic.strip().lower().replace(" ", "-") for topic in response.split(",") if topic.strip()]
             logger.debug(f"Generated topics from LLM: {topics}")
             validated_topics = self.validate_topics(topics)
@@ -182,8 +185,11 @@ def _extract_readme_urls(readme_content: str) -> List[str]:
     def _analyze_urls(self, urls: List[str]) -> List[str]:
         """Generates LLM prompt for URL analysis"""
         logger.info(f"Analyzing {len(urls)} project URLs...")
-        formatted_prompt = self.prompts.analyze_urls.format(project_url=self.repo_url, urls=", ".join(urls))
-        response = self.model_handler.send_request(formatted_prompt)
+
+        prompt = PromptBuilder.render(
+            self.prompts.get("about_section.analyze_urls"), project_url=self.repo_url, urls=", ".join(urls)
+        )
+        response = self.model_handler.send_request(prompt)
         if not response:
             return []
 

osa_tool/aboutgen/prompts_about_config.py

@@ -8,8 +8,8 @@
 from dotenv import load_dotenv
 from requests import HTTPError
 
-from osa_tool.logger import logger
-from osa_tool.utils import get_base_repo_url
+from osa_tool.utils.logger import logger
+from osa_tool.utils.utils import get_base_repo_url
 
 load_dotenv()
 

osa_tool/analytics/metadata.py

@@ -1,10 +1,9 @@
 import os
 
-import tomli as tomllib
 from pydantic import ValidationError
 
 from osa_tool.analytics.metadata import RepositoryMetadata
-from osa_tool.analytics.prompt_builder import (
+from osa_tool.analytics.response_validation import (
     RepositoryReport,
     RepositoryStructure,
     ReadmeEvaluation,
@@ -13,21 +12,24 @@
 )
 from osa_tool.analytics.sourcerank import SourceRank
 from osa_tool.config.settings import ConfigLoader
-from osa_tool.logger import logger
 from osa_tool.models.models import ModelHandler, ModelHandlerFactory
 from osa_tool.readmegen.postprocessor.response_cleaner import JsonProcessor
-from osa_tool.utils import extract_readme_content, osa_project_root, parse_folder_name
+from osa_tool.utils.logger import logger
+from osa_tool.utils.prompts_builder import PromptLoader, PromptBuilder
+from osa_tool.utils.utils import extract_readme_content, parse_folder_name
 
 
 class TextGenerator:
-    def __init__(self, config_loader: ConfigLoader, sourcerank: SourceRank, metadata: RepositoryMetadata):
+    def __init__(
+        self, config_loader: ConfigLoader, sourcerank: SourceRank, prompts: PromptLoader, metadata: RepositoryMetadata
+    ):
         self.config = config_loader.config
         self.sourcerank = sourcerank
+        self.prompts = prompts
+        self.metadata = metadata
         self.model_handler: ModelHandler = ModelHandlerFactory.build(self.config)
         self.repo_url = self.config.git.repository
-        self.metadata = metadata
         self.base_path = os.path.join(os.getcwd(), parse_folder_name(self.repo_url))
-        self.prompt_path = os.path.join(osa_project_root(), "config", "settings", "prompt_for_analysis.toml")
 
     def make_request(self) -> RepositoryReport:
         """
@@ -36,7 +38,16 @@ def make_request(self) -> RepositoryReport:
         Returns:
             str: The generated repository analysis response from the model.
         """
-        response = self.model_handler.send_request(self._build_prompt())
+        response = self.model_handler.send_request(
+            PromptBuilder.render(
+                self.prompts.get("analysis.main_prompt"),
+                project_name=self.metadata.name,
+                metadata=self.metadata,
+                repository_tree=self.sourcerank.tree,
+                presence_files=self._extract_presence_files(),
+                readme_content=extract_readme_content(self.base_path),
+            )
+        )
         parsed_data = JsonProcessor.parse(response, expected_type=dict)
         try:
             parsed_report = RepositoryReport.model_validate(parsed_data)
@@ -56,30 +67,6 @@ def make_request(self) -> RepositoryReport:
             logger.error(f"Unexpected error while parsing RepositoryReport: {e}")
             raise ValueError(f"Failed to process model response: {e}")
 
-    def _build_prompt(self) -> str:
-        """
-        Builds the prompt to be sent to the model for repository analysis.
-
-        This method loads the prompt structure from a file and formats it with values
-        extracted from the repository's metadata and other relevant information like
-        the project name, presence of key files, and repository tree.
-
-        Returns:
-            str: The formatted prompt to be used in the model request.
-        """
-        with open(self.prompt_path, "rb") as f:
-            prompts = tomllib.load(f)
-
-        main_prompt = prompts.get("prompt", {}).get("main_prompt", "")
-        prompt = main_prompt.format(
-            project_name=self.metadata.name,
-            metadata=self.metadata,
-            repository_tree=self.sourcerank.tree,
-            presence_files=self._extract_presence_files(),
-            readme_content=extract_readme_content(self.base_path),
-        )
-        return prompt
-
     def _extract_presence_files(self) -> list[str]:
         """
         Extracts information about the presence of key files in the repository.

osa_tool/analytics/report_generator.py

@@ -21,17 +21,20 @@
 from osa_tool.analytics.report_generator import TextGenerator
 from osa_tool.analytics.sourcerank import SourceRank
 from osa_tool.config.settings import ConfigLoader
-from osa_tool.logger import logger
-from osa_tool.utils import osa_project_root
+from osa_tool.utils.logger import logger
+from osa_tool.utils.prompts_builder import PromptLoader
+from osa_tool.utils.utils import osa_project_root
 
 
 class ReportGenerator:
 
-    def __init__(self, config_loader: ConfigLoader, sourcerank: SourceRank, metadata: RepositoryMetadata):
+    def __init__(
+        self, config_loader: ConfigLoader, sourcerank: SourceRank, prompts: PromptLoader, metadata: RepositoryMetadata
+    ):
         self.config = config_loader.config
         self.sourcerank = sourcerank
         self.metadata = metadata
-        self.text_generator = TextGenerator(config_loader, self.sourcerank, self.metadata)
+        self.text_generator = TextGenerator(config_loader, self.sourcerank, prompts, self.metadata)
         self.repo_url = self.config.git.repository
         self.osa_url = "https://github.com/aimclub/OSA"
 

osa_tool/analytics/report_maker.py

@@ -2,7 +2,7 @@
 import re
 
 from osa_tool.config.settings import ConfigLoader
-from osa_tool.utils import get_repo_tree, parse_folder_name
+from osa_tool.utils.utils import get_repo_tree, parse_folder_name
 
 
 class SourceRank:

osa_tool/analytics/prompt_builder.py osa_tool/analytics/response_validation.pyosa_tool/analytics/prompt_builder.py renamed to osa_tool/analytics/response_validation.py

@@ -1,4 +1,4 @@
-[prompt]
+[prompts]
 main_prompt = """
 INPUT DATA: