services package

class services.TranscriptionService(model_size='base')[source]

Bases: object

Service for audio transcription using OpenAI’s Whisper model.

Handles the complete process of transcribing audio files to text, including preprocessing, transcribing, and post-processing of the result.

Variables:

model_name – Name of the Whisper model to use
model – Instance of the loaded Whisper model

Initializes the transcription service with a Whisper model.

Parameters:: model_size (str) – Size of the Whisper model (“tiny”, “base”, “small”, “medium”, “large”)

__init__(model_size='base')[source]

Initializes the transcription service with a Whisper model.

Parameters:: model_size (str) – Size of the Whisper model (“tiny”, “base”, “small”, “medium”, “large”)

transcribe_audio_file(audio_file, encounter_id, language='it')[source]

Transcribes an audio file and saves the result to the database.

Parameters:

audio_file (FileField) – Audio file to transcribe
encounter_id (str) – Unique ID of the encounter associated with the transcription
language (str) – Language code of the audio content (default: “it”)

Returns:

Transcription object saved in the database

Return type:

AudioTranscript

Raises:

Exception – If an error occurs during transcription

class services.ClinicalExtractionService[source]

Bases: object

Service for extracting structured clinical data from transcripts

Initialize the extraction service

__init__()[source]: Initialize the extraction service

extract_clinical_data(transcript)[source]

Extract structured clinical data from a transcript

Parameters:: transcript (AudioTranscript) – AudioTranscript object to process
Returns:: ClinicalData: Object with extracted clinical data
Return type:: ClinicalData
Raises:: Exception – If extraction fails

Submodules

services.clinical_extraction module

Unified service for extracting clinical entities Handles both LLM (NVIDIA NIM) and NER (Text2NER) extraction

class services.clinical_extraction.ExtractionMethod(value)[source]

Bases: Enum

Listing of available clinical entity extraction methods.

Variables:

LLM – Extraction using Large Language Model (NVIDIA NIM)
NER – Extraction using Named Entity Recognition

LLM = 'llm'

NER = 'ner'

class services.clinical_extraction.ClinicalExtractionService[source]

Bases: object

Unified service for extracting clinical entities from medical transcriptions.

Allows selection between different extraction methods (LLM and NER) while maintaining a consistent interface for usage.

Variables:

llm_service – Service for extraction using LLM
ner_service – Service for extraction using NER
default_method – Default extraction method

Initializes the clinical extraction service.

Configures the available LLM and NER services, handling any initialization errors and setting the default method.

__init__()[source]

Initializes the clinical extraction service.

Configures the available LLM and NER services, handling any initialization errors and setting the default method.

get_available_methods()[source]

Returns the available extraction methods with their operational status.

Returns:: Dictionary containing information about available methods and their status
Return type:: Dict[str, Dict[str, Any]]

extract_clinical_entities(transcript_text, method=None, usage_mode='')[source]

Extract clinical entities from a medical transcription using the specified method.

Parameters:

transcript_text (str) – Text of the medical transcription to analyze
method (str) – Extraction method (“llm” or “ner”), if None uses the default
usage_mode (str) – Usage mode to customize extraction (e.g. “Checkup”, “Emergency”)

Returns:

Dictionary containing the extracted clinical entities and metadata

Return type:

Dict[str, Any]

Raises:

ValueError – If the specified method is invalid

set_default_method(method)[source]

Setting of default extraction method

Parameters:: method (str) – Method to set as default (“llm” or “ner”)
Returns:: True if the method was set successfully
Return type:: bool

get_method_comparison(transcript_text, usage_mode='')[source]

Execute extraction with both methods for comparison WARNING: This is a costly operation, use only for testing/debugging

Parameters:

transcript_text (str) – Text of the medical transcription
usage_mode (str) – Usage mode

Returns:

Dictionary with results from both methods

Return type:

Dict[str, Any]

services.extraction module

Service for extracting clinical data using LLM

class services.extraction.ClinicalExtractionService[source]

Bases: object

Service for extracting structured clinical data from transcripts

Initialize the extraction service

__init__()[source]: Initialize the extraction service

extract_clinical_data(transcript)[source]

Extract structured clinical data from a transcript

Parameters:: transcript (AudioTranscript) – AudioTranscript object to process
Returns:: ClinicalData: Object with extracted clinical data
Return type:: ClinicalData
Raises:: Exception – If extraction fails

services.mongodb_service module

Service for MongoDB management of patient and visit data

class services.mongodb_service.MongoDBService[source]

Bases: object

Service for MongoDB management of patient and visit data

Initialize MongoDB connection

__init__()[source]: Initialize MongoDB connection

is_connected()[source]

Check MongoDB connection status

save_patient_visit_transcript_only(encounter_id, patient_id, doctor_id, audio_file_path, transcript_text, triage_code=None, symptoms=None, triage_notes=None)[source]

Save a new patient visit with ONLY transcript (without extracted clinical data)

Parameters:

encounter_id (str) – ID encounter Django
patient_id (str) – ID paziente Django
doctor_id (str) – ID medico Django
audio_file_path (str) – Path file audio
transcript_text (str) – Testo trascrizione
triage_code (str) – Codice triage (white, green, yellow, red)
symptoms (str) – Sintomi principali
triage_notes (str) – Note del triage

Returns:

ID del transcript MongoDB creato

Return type:

Optional[str]

save_patient_visit(encounter_id, patient_id, doctor_id, audio_file_path, transcript_text, clinical_data)[source]

Save a new patient visit with audio and clinical data

Parameters:

encounter_id (str) – ID encounter Django
patient_id (str) – ID paziente Django
doctor_id (str) – ID medico Django
audio_file_path (str) – Path file audio
transcript_text (str) – Testo trascrizione
clinical_data (Dict[str, Any]) – Dati clinici estratti

Returns:

ID del transcript MongoDB creato

Return type:

Optional[str]

get_patient_visits(patient_id)[source]

Retrieve all visits of a patient

Parameters:: patient_id (str) – ID paziente Django
Returns:: Lista delle visite del paziente
Return type:: List[Dict[str, Any]]

get_all_patients_summary()[source]

Retrieve the summary of all patients with their latest visits

Returns:: List of patients with aggregated data
Return type:: List[Dict[str, Any]]

get_visits_today()[source]

Count ALL visits today (created today, both completed and in progress)

Returns:: Number of visits created today
Return type:: int

get_waiting_patients_count()[source]

Count patients waiting for processing (visits not yet completed)

Returns:: Number of patients waiting for processing
Return type:: int

get_completed_visits_today()[source]

Count ALL completed visits (with status ‘extracted’ or ‘validated’)

Returns:: Number of completed visits
Return type:: int

get_unique_patients()[source]

Retrieve a list of unique patients grouped by fiscal code from all interventions

Returns:: List of unique patients with aggregated data
Return type:: List[Dict[str, Any]]

update_patient_data(patient_id, updated_data)[source]

Update the personal data of a patient in the latest visit

Parameters:

patient_id (str) – Django patient ID
updated_data (Dict[str, Any]) – Updated data fields

Returns:

True if update succeeded, False otherwise

Return type:

bool

generate_report_content(transcript_id)[source]

Generate the content for the PDF report

Parameters:: transcript_id (str) – ID transcript MongoDB
Returns:: Dictionary with structured content for PDF
Return type:: Optional[Dict[str, Any]]

update_transcript_text(transcript_id, new_text)[source]

Update the transcript text

Parameters:

transcript_id (str) – ID transcript MongoDB
new_text (str) – New transcript text

Returns:

True if update succeeded, False otherwise

Return type:

bool

update_clinical_data(transcript_id, clinical_dict)[source]

Update the clinical data associated with the transcript

Parameters:

transcript_id (str) – ID transcript MongoDB
clinical_dict (Dict[str, Any]) – Dictionary with clinical data fields to update

Returns:

True if update succeeded, False otherwise

Return type:

bool

get_all_visits_summary()[source]

Retrieve a summary list of all visits/interventions

Returns:: List of dictionaries with summary information
Return type:: List[Dict[str, Any]]

get_visit_data(transcript_id)[source]

Retrieve the complete data of a visit/transcript

Parameters:: transcript_id (str) – ID of the transcript
Returns:: Dictionary with visit data or None if not found
Return type:: Optional[Dict[str, Any]]

delete_visit(transcript_id)[source]

Delete a visit from MongoDB

Parameters:: transcript_id (str) – ID of the transcript to delete
Returns:: True if deletion was successful, False otherwise
Return type:: bool

services.ner_service module

Service for extracting clinical entities with NER model Manages extraction using the Text2NER model from pacovalentino/Text2NER

class services.ner_service.NERService[source]

Bases: object

Service for extracting clinical entities with NER model pacovalentino/Text2NER

Initialize the NER service

__init__()[source]: Initialize the NER service

test_connection()[source]

Test the availability of the NER model

Returns:: Dictionary with model information
Return type:: Dict[str, Any]

extract_clinical_entities(transcript_text, usage_mode='')[source]

Extract clinical entities from the transcribed text using the NER model

Parameters:

transcript_text (str) – Transcribed medical text
usage_mode (str) – Usage mode (e.g. “Checkup”, “Emergency”)

Returns:

Dictionary with extracted clinical entities

Return type:

Dict[str, Any]

services.nvidia_nim module

Service for integration with NVIDIA NIM API Handles LLM calls for extracting clinical entities

class services.nvidia_nim.NVIDIANIMService[source]

Bases: object

Service for integration with NVIDIA NIM API for clinical entity extraction.

Handles LLM calls for structured extraction of medical information from audio transcriptions, with support for local fallback when the service is not available.

Variables:

model – LLM model used for extraction
available – Flag indicating if the service is available
client – OpenAI client for API calls

Initializes the NVIDIA NIM service.

Configures the OpenAI client with credentials and base URL from Django settings. If credentials are not available, enables fallback mode.

__init__()[source]

Initializes the NVIDIA NIM service.

Configures the OpenAI client with credentials and base URL from Django settings. If credentials are not available, enables fallback mode.

test_connection()[source]

Tests the connection with NVIDIA NIM API.

Returns:: Dictionary containing information about the connection status
Return type:: Dict[str, Any]
Raises:: Exception – If an error occurs during connection testing

extract_clinical_entities(transcript_text, usage_mode='')[source]

Extract structured clinical entities from a medical transcription using NVIDIA NIM.

Parameters:

transcript_text (str) – Text of the medical transcription to analyze
usage_mode (str) – Service usage mode (e.g. “Checkup”, “Emergency”)

Returns:

Dictionary containing the extracted structured clinical entities

Return type:

Dict[str, Any]

Raises:

Exception – If an error occurs during extraction

services.nvidia_nim.get_nvidia_nim_service()[source]

Factory function per ottenere l’istanza del servizio NVIDIA NIM. Utilizzata per evitare problemi di importazione durante la generazione della documentazione.

Returns:: Istanza del servizio NVIDIA NIM
Return type:: NVIDIANIMService

services.pdf_report module

Enhanced service for generating emergency medical PDF reports

class services.pdf_report.PDFReportService[source]

Bases: object

Enhanced service for generating emergency medical PDF reports - Professional layout - Dynamic (adapts text space) - More readable and well-spaced

Initialize PDF report service with styles and settings

__init__()[source]: Initialize PDF report service with styles and settings

generate_medical_report(report_data, output_path)[source]

Generate a professional medical report PDF

get_report_path(encounter_id, report_type='medical', patient_name='', visit_date='')[source]

Generate a safe file path for the report PDF

Parameters:

encounter_id (str) – Unique encounter identifier
report_type (str) – Type of report (default “medical”)
patient_name (str) – Patient’s name for filename (optional)
visit_date (str) – Visit date for filename (optional)

Returns:

Full file path for the report PDF

Return type:

str

services.pdf_report.get_pdf_report_service()[source]

Factory function per ottenere l’istanza del servizio PDF Report. Utilizzata per evitare problemi di importazione durante la generazione della documentazione.

Returns:: Istanza del servizio PDF Report
Return type:: PDFReportService

services.transcription module

Service for audio transcription using Whisper

services.transcription.check_dependencies()[source]

Verify that all necessary dependencies for transcription are available.

Check for the presence of numpy, torch, whisper, and librosa in the system.

Returns:: True if all dependencies are available
Return type:: bool
Raises:: ImportError – If one or more dependencies are missing

class services.transcription.TranscriptionService(model_size='base')[source]

Bases: object

Service for audio transcription using OpenAI’s Whisper model.

Handles the complete process of transcribing audio files to text, including preprocessing, transcribing, and post-processing of the result.

Variables:

model_name – Name of the Whisper model to use
model – Instance of the loaded Whisper model

Initializes the transcription service with a Whisper model.

Parameters:: model_size (str) – Size of the Whisper model (“tiny”, “base”, “small”, “medium”, “large”)

__init__(model_size='base')[source]

Initializes the transcription service with a Whisper model.

Parameters:: model_size (str) – Size of the Whisper model (“tiny”, “base”, “small”, “medium”, “large”)

transcribe_audio_file(audio_file, encounter_id, language='it')[source]

Transcribes an audio file and saves the result to the database.

Parameters:

audio_file (FileField) – Audio file to transcribe
encounter_id (str) – Unique ID of the encounter associated with the transcription
language (str) – Language code of the audio content (default: “it”)

Returns:

Transcription object saved in the database

Return type:

AudioTranscript

Raises:

Exception – If an error occurs during transcription

services.whisper_realtime module

services.whisper_service module

Service for audio transcription with Whisper medium Stable version for production

class services.whisper_service.WhisperService[source]

Bases: object

Service for audio transcription with Whisper medium

__init__()[source]

transcribe_audio_file(audio_file_path, language='it')[source]

Transcription of audio file with Whisper

Parameters:

audio_file_path (str) – Path to the audio file
language (str) – Language for transcription (default: Italian)

Returns:

Dictionary with transcription results

Return type:

Dict[str, Any]

transcribe_audio_blob(audio_blob, format='wav', language='it')[source]

Transcription of audio blob with Whisper

Parameters:

audio_blob (bytes) – Audio data in bytes
format (str) – Audio format (wav, mp3, etc.)
language (str) – Language for transcription (default: Italian)

Returns:

Dictionary with transcription results

Return type:

Dict[str, Any]

test_transcription()[source]

Test the transcription service

Returns:: Dictionary with test results
Return type:: Dict[str, Any]

get_supported_formats()[source]

Supported audio formats

Returns:: List of supported audio formats
Return type:: list