docs/class_diagram.md

Class diagram - V7

Modules are shown as facade classes over their public functions. GitHub renders this Mermaid diagram directly. Compared to V3, Database.export_xml() is generalized into export_records(fmt, limit), dispatching to one exporter function per format (xml, json, csv - listed in db.EXPORT_FORMATS). V5 only changed the dashboard's model picker (dropdown instead of radio buttons) and the export button (dropdown menu) - no backend classes changed there. V6 added a batched method to NERProcessor, TopicClassifier and SentimentAnalyzer (*_batch, taking a list of texts), and Pipeline. process_batch calls those directly instead of looping over process_message - see the README section "Batched inference for large /ingest batches". V7 adds a new JobTracker class (jobs.py) and makes API.ingest_file start a background thread instead of processing synchronously - see "Background ingest jobs (progress bar + timer)".

classDiagram
    class API {
        <<main.py>>
        +home() FileResponse
        +health() dict
        +list_models() dict
        +analyze(payload) dict
        +predict(payload) dict
        +ingest_file(file, sentiment_model, ner_model, topic_model) dict
        +ingest_status(job_id) dict
        +records(limit) list
        +export_records(format, limit) Response
        +get_stats() dict
    }

    class IngestService {
        <<ingest.py>>
        +parse_csv(raw) list
        +parse_json(raw) list
        +parse_upload(filename, raw) list
    }

    class JobTracker {
        <<jobs.py>>
        -jobs: dict~job_id, dict~
        +create_job(total) str
        +progress_callback(job_id) Callable
        +finish_job(job_id, result) void
        +fail_job(job_id, error) void
        +get_job(job_id) dict
    }

    class Pipeline {
        <<pipeline.py>>
        +process_message(text, topic_labels, sentiment_model, ner_model, topic_model) dict
        +process_batch(messages, topic_labels, sentiment_model, ner_model, topic_model, on_progress) list
    }

    class NERProcessor {
        <<processors/ner.py>>
        -default_model_name: str = "dslim/bert-base-NER"
        +extract_entities(text, model_id) list
        +extract_entities_batch(texts, model_id) list~list~
        +is_ready() bool
    }

    class TopicClassifier {
        <<processors/topics.py>>
        -default_model_name: str = "facebook/bart-large-mnli"
        +classify_topic(text, labels, model_id) dict
        +classify_topic_batch(texts, labels, model_id) list~dict~
        +is_ready() bool
    }

    class SentimentAnalyzer {
        <<processors/sentiment.py>>
        -default_model_name: str = "vojmahdal/roberta-sentiment-3labels"
        +analyze_sentiment(text, model_id) dict
        +analyze_sentiment_batch(texts, model_id) list~dict~
        +is_ready() bool
    }

    class ModelRegistry {
        <<processors/model_registry.py>>
        -cache: dict~"task::model_id", Pipeline~
        -max_cached_pipelines: int = 6
        +get_pipeline(model_id, task, **kwargs) Pipeline
        +cached_models(task) dict
    }

    class Anonymizer {
        <<processors/anonymizer.py>>
        -backend: presidio | regex-fallback
        +anonymize_text(text) str
        +detect_pii(text) list
        +backend_name() str
    }

    class Database {
        <<db.py>>
        -export_formats: dict = {xml, json, csv}
        +save_record(result, source) void
        +get_records(limit) list
        +stats() dict
        +export_records(fmt, limit) bytes
    }

    API --> IngestService : parses uploaded files
    API --> Pipeline : runs analysis (in a background thread)
    API --> JobTracker : creates/polls ingest jobs
    API --> Database : reads/writes/export records
    API --> ModelRegistry : lists cached models
    Pipeline --> JobTracker : reports progress via on_progress
    Pipeline --> NERProcessor
    Pipeline --> TopicClassifier
    Pipeline --> SentimentAnalyzer
    Pipeline --> Anonymizer
    NERProcessor --> ModelRegistry : loads non-default HF models
    TopicClassifier --> ModelRegistry : loads non-default HF models
    SentimentAnalyzer --> ModelRegistry : loads non-default HF models