themefinder 0.6.2__tar.gz → 0.6.3__tar.gz

{themefinder-0.6.2 → themefinder-0.6.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: themefinder
-Version: 0.6.2
+Version: 0.6.3
 Summary: A topic modelling Python package designed for analysing one-to-many question-answer data eg free-text survey responses.
 License: MIT
 Author: i.AI
@@ -49,9 +49,9 @@ ThemeFinder takes as input a [pandas DataFrame](https://pandas.pydata.org/docs/r
 - `response_id`: A unique identifier for each response
 - `response`: The free text survey response
 
-ThemeFinder is compatible with any instantiated [LangChain LLM runnable](https://python.langchain.com/v0.1/docs/integrations/llms/), but you will need to use JSON structured output.
+ThemeFinder now supports a range of language models through structured outputs.
 
-The function `find_themes` identifies common themes in response and labels them, it also outputs results from intermediate steps in the theme finding pipeline.
+The function `find_themes` identifies common themes in responses and labels them, it also outputs results from intermediate steps in the theme finding pipeline.
 
 For this example, import the following Python packages into your virtual environment: `asyncio`, `pandas`, `lanchain`. And import `themefinder` as described above.
 
@@ -81,7 +81,6 @@ load_dotenv()
 llm = AzureChatOpenAI(
     model="gpt-4o",
     temperature=0,
-    model_kwargs={"response_format": {"type": "json_object"}},
 )
 
 # Set up your data
@@ -97,18 +96,15 @@ question = "What do you think of ThemeFinder?"
 # Make the system prompt specific to your use case 
 system_prompt = "You are an AI evaluation tool analyzing survey responses about a Python package."
 
-# Run the function to find themes
-# We use asyncio to query LLM endpoints asynchronously, so we need to await our function
+# Run the function to find themes, we use asyncio to query LLM endpoints asynchronously, so we need to await our function
 async def main():
     result = await find_themes(responses_df, llm, question, system_prompt=system_prompt)
     print(result)
 
 if __name__ == "__main__":
     asyncio.run(main())
-
 ```
 
-
 ## ThemeFinder pipeline
 
 ThemeFinder's pipeline consists of five distinct stages, each utilizing a specialized LLM prompt:
@@ -145,6 +141,25 @@ The file `src/themefinder.core.py` contains the function `find_themes` which run
 **For more detail - see the docs: [https://i-dot-ai.github.io/themefinder/](https://i-dot-ai.github.io/themefinder/).**
 
 
+## Model Compatibility
+
+ThemeFinder's structured output approach makes it compatible with a wide range of language models from various providers. This list is non-exhaustive, and other models may also work effectively:
+
+### OpenAI Models
+- GPT-4, GPT-4o, GPT-4.1
+- All Azure OpenAI deployments
+
+### Google Models
+- Gemini series (1.5 Pro, 2.0 Pro, etc.)
+
+### Anthropic Models
+- Claude series (Claude 3 Opus, Sonnet, Haiku, etc.)
+
+### Open Source Models
+- Llama 2, Llama 3
+- Mistral models (e.g., Mistral 7B, Mixtral)
+
+
 ## License
 
 This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

{themefinder-0.6.2 → themefinder-0.6.3}/README.md

@@ -18,9 +18,9 @@ ThemeFinder takes as input a [pandas DataFrame](https://pandas.pydata.org/docs/r
 - `response_id`: A unique identifier for each response
 - `response`: The free text survey response
 
-ThemeFinder is compatible with any instantiated [LangChain LLM runnable](https://python.langchain.com/v0.1/docs/integrations/llms/), but you will need to use JSON structured output.
+ThemeFinder now supports a range of language models through structured outputs.
 
-The function `find_themes` identifies common themes in response and labels them, it also outputs results from intermediate steps in the theme finding pipeline.
+The function `find_themes` identifies common themes in responses and labels them, it also outputs results from intermediate steps in the theme finding pipeline.
 
 For this example, import the following Python packages into your virtual environment: `asyncio`, `pandas`, `lanchain`. And import `themefinder` as described above.
 
@@ -50,7 +50,6 @@ load_dotenv()
 llm = AzureChatOpenAI(
     model="gpt-4o",
     temperature=0,
-    model_kwargs={"response_format": {"type": "json_object"}},
 )
 
 # Set up your data
@@ -66,18 +65,15 @@ question = "What do you think of ThemeFinder?"
 # Make the system prompt specific to your use case 
 system_prompt = "You are an AI evaluation tool analyzing survey responses about a Python package."
 
-# Run the function to find themes
-# We use asyncio to query LLM endpoints asynchronously, so we need to await our function
+# Run the function to find themes, we use asyncio to query LLM endpoints asynchronously, so we need to await our function
 async def main():
     result = await find_themes(responses_df, llm, question, system_prompt=system_prompt)
     print(result)
 
 if __name__ == "__main__":
     asyncio.run(main())
-
 ```
 
-
 ## ThemeFinder pipeline
 
 ThemeFinder's pipeline consists of five distinct stages, each utilizing a specialized LLM prompt:
@@ -114,6 +110,25 @@ The file `src/themefinder.core.py` contains the function `find_themes` which run
 **For more detail - see the docs: [https://i-dot-ai.github.io/themefinder/](https://i-dot-ai.github.io/themefinder/).**
 
 
+## Model Compatibility
+
+ThemeFinder's structured output approach makes it compatible with a wide range of language models from various providers. This list is non-exhaustive, and other models may also work effectively:
+
+### OpenAI Models
+- GPT-4, GPT-4o, GPT-4.1
+- All Azure OpenAI deployments
+
+### Google Models
+- Gemini series (1.5 Pro, 2.0 Pro, etc.)
+
+### Anthropic Models
+- Claude series (Claude 3 Opus, Sonnet, Haiku, etc.)
+
+### Open Source Models
+- Llama 2, Llama 3
+- Mistral models (e.g., Mistral 7B, Mixtral)
+
+
 ## License
 
 This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

{themefinder-0.6.2 → themefinder-0.6.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "themefinder"
-version = "0.6.2"
+version = "0.6.3"
 description = "A topic modelling Python package designed for analysing one-to-many question-answer data eg free-text survey responses."
 authors = ["i.AI <packages@cabinetoffice.gov.uk>"]
 packages = [{include = "themefinder", from = "src"}]

{themefinder-0.6.2 → themefinder-0.6.3}/src/themefinder/__init__.py

@@ -5,6 +5,8 @@ from .core import (
     theme_generation,
     theme_mapping,
     theme_refinement,
+    theme_target_alignment,
+    detail_detection,
 )
 
 __all__ = [
@@ -13,6 +15,8 @@ __all__ = [
     "theme_generation",
     "theme_condensation",
     "theme_refinement",
+    "theme_target_alignment",
     "theme_mapping",
+    "detail_detection",
 ]
 __version__ = "0.1.0"

{themefinder-0.6.2 → themefinder-0.6.3}/src/themefinder/core.py

@@ -3,10 +3,17 @@ from pathlib import Path
 
 import pandas as pd
 from langchain_core.prompts import PromptTemplate
-from langchain_core.runnables import Runnable
+from langchain.schema.runnable import RunnableWithFallbacks
 
 from .llm_batch_processor import batch_and_run, load_prompt_from_file
-from .models import SentimentAnalysisOutput, ThemeMappingOutput
+from .models import (
+    SentimentAnalysisResponses,
+    ThemeGenerationResponses,
+    ThemeCondensationResponses,
+    ThemeRefinementResponses,
+    ThemeMappingResponses,
+    DetailDetectionResponses,
+)
 from .themefinder_logging import logger
 
 CONSULTATION_SYSTEM_PROMPT = load_prompt_from_file("consultation_system_prompt")
@@ -14,11 +21,12 @@ CONSULTATION_SYSTEM_PROMPT = load_prompt_from_file("consultation_system_prompt")
 
 async def find_themes(
     responses_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     target_n_themes: int | None = None,
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
     verbose: bool = True,
+    concurrency: int = 10,
 ) -> dict[str, str | pd.DataFrame]:
     """Process survey responses through a multi-stage theme analysis pipeline.
 
@@ -32,7 +40,7 @@ async def find_themes(
 
     Args:
         responses_df (pd.DataFrame): DataFrame containing survey responses
-        llm (Runnable): Language model instance for text analysis
+        llm (RunnableWithFallbacks): Language model instance for text analysis
         question (str): The survey question
         target_n_themes (int | None, optional): Target number of themes to consolidate to.
             If None, skip theme target alignment step. Defaults to None.
@@ -40,6 +48,7 @@ async def find_themes(
             Defaults to CONSULTATION_SYSTEM_PROMPT.
         verbose (bool): Whether to show information messages during processing.
             Defaults to True.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         dict[str, str | pd.DataFrame]: Dictionary containing results from each pipeline stage:
@@ -56,21 +65,28 @@ async def find_themes(
         llm,
         question=question,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
     theme_df, _ = await theme_generation(
         sentiment_df,
         llm,
         question=question,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
     condensed_theme_df, _ = await theme_condensation(
-        theme_df, llm, question=question, system_prompt=system_prompt
+        theme_df,
+        llm,
+        question=question,
+        system_prompt=system_prompt,
+        concurrency=concurrency,
     )
     refined_theme_df, _ = await theme_refinement(
         condensed_theme_df,
         llm,
         question=question,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
     if target_n_themes is not None:
         refined_theme_df, _ = await theme_target_alignment(
@@ -79,6 +95,7 @@ async def find_themes(
             question=question,
             target_n_themes=target_n_themes,
             system_prompt=system_prompt,
+            concurrency=concurrency,
         )
     mapping_df, mapping_unprocessables = await theme_mapping(
         sentiment_df[["response_id", "response"]],
@@ -86,6 +103,14 @@ async def find_themes(
         question=question,
         refined_themes_df=refined_theme_df,
         system_prompt=system_prompt,
+        concurrency=concurrency,
+    )
+    detailed_df, _ = await detail_detection(
+        responses_df[["response_id", "response"]],
+        llm,
+        question=question,
+        system_prompt=system_prompt,
+        concurrency=concurrency,
     )
 
     logger.info("Finished finding themes")
@@ -97,17 +122,19 @@ async def find_themes(
         "sentiment": sentiment_df,
         "themes": refined_theme_df,
         "mapping": mapping_df,
+        "detailed_responses": detailed_df,
         "unprocessables": pd.concat([sentiment_unprocessables, mapping_unprocessables]),
     }
 
 
 async def sentiment_analysis(
     responses_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     batch_size: int = 20,
     prompt_template: str | Path | PromptTemplate = "sentiment_analysis",
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
 ) -> tuple[pd.DataFrame, pd.DataFrame]:
     """Perform sentiment analysis on survey responses using an LLM.
 
@@ -117,7 +144,7 @@ async def sentiment_analysis(
     Args:
         responses_df (pd.DataFrame): DataFrame containing survey responses to analyze.
             Must contain 'response_id' and 'response' columns.
-        llm (Runnable): Language model instance to use for sentiment analysis.
+        llm (RunnableWithFallbacks): Language model instance to use for sentiment analysis.
         question (str): The survey question.
         batch_size (int, optional): Number of responses to process in each batch.
             Defaults to 20.
@@ -126,6 +153,7 @@ async def sentiment_analysis(
             or PromptTemplate instance. Defaults to "sentiment_analysis".
         system_prompt (str): System prompt to guide the LLM's behavior.
             Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         tuple[pd.DataFrame, pd.DataFrame]:
@@ -134,32 +162,33 @@ async def sentiment_analysis(
                 - The second DataFrame contains the rows that could not be processed by the LLM
 
     Note:
-        The function uses validation_check to ensure responses maintain
+        The function uses integrity_check to ensure responses maintain
         their original order and association after processing.
     """
     logger.info(f"Running sentiment analysis on {len(responses_df)} responses")
-    processed_rows, unprocessable_rows = await batch_and_run(
+    sentiment, unprocessable = await batch_and_run(
         responses_df,
         prompt_template,
-        llm,
+        llm.with_structured_output(SentimentAnalysisResponses),
         batch_size=batch_size,
         question=question,
-        validation_check=True,
-        task_validation_model=SentimentAnalysisOutput,
+        integrity_check=True,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
 
-    return processed_rows, unprocessable_rows
+    return sentiment, unprocessable
 
 
 async def theme_generation(
     responses_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     batch_size: int = 50,
     partition_key: str | None = "position",
     prompt_template: str | Path | PromptTemplate = "theme_generation",
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
 ) -> tuple[pd.DataFrame, pd.DataFrame]:
     """Generate themes from survey responses using an LLM.
 
@@ -168,7 +197,7 @@ async def theme_generation(
     Args:
         responses_df (pd.DataFrame): DataFrame containing survey responses.
             Must include 'response_id' and 'response' columns.
-        llm (Runnable): Language model instance to use for theme generation.
+        llm (RunnableWithFallbacks): Language model instance to use for theme generation.
         question (str): The survey question.
         batch_size (int, optional): Number of responses to process in each batch.
             Defaults to 50.
@@ -181,6 +210,7 @@ async def theme_generation(
             or PromptTemplate instance. Defaults to "theme_generation".
         system_prompt (str): System prompt to guide the LLM's behavior.
             Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         tuple[pd.DataFrame, pd.DataFrame]:
@@ -193,22 +223,24 @@ async def theme_generation(
     generated_themes, _ = await batch_and_run(
         responses_df,
         prompt_template,
-        llm,
+        llm.with_structured_output(ThemeGenerationResponses),
         batch_size=batch_size,
         partition_key=partition_key,
         question=question,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
     return generated_themes, _
 
 
 async def theme_condensation(
     themes_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     batch_size: int = 75,
     prompt_template: str | Path | PromptTemplate = "theme_condensation",
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
     **kwargs,
 ) -> tuple[pd.DataFrame, pd.DataFrame]:
     """Condense and combine similar themes identified from survey responses.
@@ -219,7 +251,7 @@ async def theme_condensation(
     Args:
         themes_df (pd.DataFrame): DataFrame containing the initial themes identified
             from survey responses.
-        llm (Runnable): Language model instance to use for theme condensation.
+        llm (RunnableWithFallbacks): Language model instance to use for theme condensation.
         question (str): The survey question.
         batch_size (int, optional): Number of themes to process in each batch.
             Defaults to 100.
@@ -228,6 +260,7 @@ async def theme_condensation(
             or PromptTemplate instance. Defaults to "theme_condensation".
         system_prompt (str): System prompt to guide the LLM's behavior.
             Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         tuple[pd.DataFrame, pd.DataFrame]:
@@ -247,10 +280,11 @@ async def theme_condensation(
         themes_df, _ = await batch_and_run(
             themes_df,
             prompt_template,
-            llm,
+            llm.with_structured_output(ThemeCondensationResponses),
             batch_size=batch_size,
             question=question,
             system_prompt=system_prompt,
+            concurrency=concurrency,
             **kwargs,
         )
         themes_df = themes_df.sample(frac=1).reset_index(drop=True)
@@ -263,10 +297,11 @@ async def theme_condensation(
     themes_df, _ = await batch_and_run(
         themes_df,
         prompt_template,
-        llm,
+        llm.with_structured_output(ThemeCondensationResponses),
         batch_size=batch_size,
         question=question,
         system_prompt=system_prompt,
+        concurrency=concurrency,
         **kwargs,
     )
 
@@ -276,11 +311,12 @@ async def theme_condensation(
 
 async def theme_refinement(
     condensed_themes_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     batch_size: int = 10000,
     prompt_template: str | Path | PromptTemplate = "theme_refinement",
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
 ) -> tuple[pd.DataFrame, pd.DataFrame]:
     """Refine and standardize condensed themes using an LLM.
 
@@ -292,7 +328,7 @@ async def theme_refinement(
     Args:
         condensed_themes (pd.DataFrame): DataFrame containing the condensed themes
             from the previous pipeline stage.
-        llm (Runnable): Language model instance to use for theme refinement.
+        llm (RunnableWithFallbacks): Language model instance to use for theme refinement.
         question (str): The survey question.
         batch_size (int, optional): Number of themes to process in each batch.
             Defaults to 10000.
@@ -301,6 +337,7 @@ async def theme_refinement(
             or PromptTemplate instance. Defaults to "theme_refinement".
         system_prompt (str): System prompt to guide the LLM's behavior.
             Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         tuple[pd.DataFrame, pd.DataFrame]:
@@ -319,22 +356,24 @@ async def theme_refinement(
     refined_themes, _ = await batch_and_run(
         condensed_themes_df,
         prompt_template,
-        llm,
+        llm.with_structured_output(ThemeRefinementResponses),
         batch_size=batch_size,
         question=question,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
     return refined_themes, _
 
 
 async def theme_target_alignment(
     refined_themes_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     target_n_themes: int = 10,
     batch_size: int = 10000,
     prompt_template: str | Path | PromptTemplate = "theme_target_alignment",
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
 ) -> tuple[pd.DataFrame, pd.DataFrame]:
     """Align themes to target number using an LLM.
 
@@ -346,7 +385,7 @@ async def theme_target_alignment(
     Args:
         refined_themes_df (pd.DataFrame): DataFrame containing the refined themes
             from the previous pipeline stage.
-        llm (Runnable): Language model instance to use for theme alignment.
+        llm (RunnableWithFallbacks): Language model instance to use for theme alignment.
         question (str): The survey question.
         target_n_themes (int, optional): Target number of themes to consolidate to.
             Defaults to 10.
@@ -357,6 +396,7 @@ async def theme_target_alignment(
             or PromptTemplate instance. Defaults to "theme_target_alignment".
         system_prompt (str): System prompt to guide the LLM's behavior.
             Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         tuple[pd.DataFrame, pd.DataFrame]:
@@ -376,23 +416,25 @@ async def theme_target_alignment(
     aligned_themes, _ = await batch_and_run(
         refined_themes_df,
         prompt_template,
-        llm,
+        llm.with_structured_output(ThemeRefinementResponses),
         batch_size=batch_size,
         question=question,
         system_prompt=system_prompt,
         target_n_themes=target_n_themes,
+        concurrency=concurrency,
     )
     return aligned_themes, _
 
 
 async def theme_mapping(
     responses_df: pd.DataFrame,
-    llm: Runnable,
+    llm: RunnableWithFallbacks,
     question: str,
     refined_themes_df: pd.DataFrame,
     batch_size: int = 20,
     prompt_template: str | Path | PromptTemplate = "theme_mapping",
     system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
 ) -> tuple[pd.DataFrame, pd.DataFrame]:
     """Map survey responses to refined themes using an LLM.
 
@@ -402,7 +444,7 @@ async def theme_mapping(
     Args:
         responses_df (pd.DataFrame): DataFrame containing survey responses.
             Must include 'response_id' and 'response' columns.
-        llm (Runnable): Language model instance to use for theme mapping.
+        llm (RunnableWithFallbacks): Language model instance to use for theme mapping.
         question (str): The survey question.
         refined_themes_df (pd.DataFrame): Single-row DataFrame where each column
             represents a theme (from theme_refinement stage).
@@ -413,6 +455,7 @@ async def theme_mapping(
             or PromptTemplate instance. Defaults to "theme_mapping".
         system_prompt (str): System prompt to guide the LLM's behavior.
             Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
 
     Returns:
         tuple[pd.DataFrame, pd.DataFrame]:
@@ -432,17 +475,70 @@ async def theme_mapping(
         )
         return transposed_df
 
-    mapping, _ = await batch_and_run(
+    mapping, unprocessable = await batch_and_run(
         responses_df,
         prompt_template,
-        llm,
+        llm.with_structured_output(ThemeMappingResponses),
         batch_size=batch_size,
         question=question,
         refined_themes=transpose_refined_themes(refined_themes_df).to_dict(
             orient="records"
         ),
-        validation_check=True,
-        task_validation_model=ThemeMappingOutput,
+        integrity_check=True,
+        system_prompt=system_prompt,
+        concurrency=concurrency,
+    )
+    return mapping, unprocessable
+
+
+async def detail_detection(
+    responses_df: pd.DataFrame,
+    llm: RunnableWithFallbacks,
+    question: str,
+    batch_size: int = 20,
+    prompt_template: str | Path | PromptTemplate = "detail_detection",
+    system_prompt: str = CONSULTATION_SYSTEM_PROMPT,
+    concurrency: int = 10,
+) -> tuple[pd.DataFrame, pd.DataFrame]:
+    """Identify responses that provide high-value detailed evidence.
+
+    This function processes survey responses in batches to analyze their level of detail
+    and evidence using a language model. It identifies responses that contain specific
+    examples, data, or detailed reasoning that provide strong supporting evidence.
+
+    Args:
+        responses_df (pd.DataFrame): DataFrame containing survey responses to analyze.
+            Must contain 'response_id' and 'response' columns.
+        llm (RunnableWithFallbacks): Language model instance to use for detail detection.
+        question (str): The survey question.
+        batch_size (int, optional): Number of responses to process in each batch.
+            Defaults to 20.
+        prompt_template (str | Path | PromptTemplate, optional): Template for structuring
+            the prompt to the LLM. Can be a string identifier, path to template file,
+            or PromptTemplate instance. Defaults to "detail_detection".
+        system_prompt (str): System prompt to guide the LLM's behavior.
+            Defaults to CONSULTATION_SYSTEM_PROMPT.
+        concurrency (int): Number of concurrent API calls to make. Defaults to 10.
+
+    Returns:
+        tuple[pd.DataFrame, pd.DataFrame]:
+            A tuple containing two DataFrames:
+                - The first DataFrame contains the rows that were successfully processed by the LLM
+                - The second DataFrame contains the rows that could not be processed by the LLM
+
+    Note:
+        The function uses response_id_integrity_check to ensure responses maintain
+        their original order and association after processing.
+    """
+    logger.info(f"Running detail detection on {len(responses_df)} responses")
+    detailed, _ = await batch_and_run(
+        responses_df,
+        prompt_template,
+        llm.with_structured_output(DetailDetectionResponses),
+        batch_size=batch_size,
+        question=question,
+        integrity_check=True,
         system_prompt=system_prompt,
+        concurrency=concurrency,
     )
-    return mapping, _
+    return detailed, _