Data flow¶

Enqueueing (UI or CLI)¶

sequenceDiagram
    participant U as User
    participant UI as FastAPI
    participant J as jobs.py
    participant DB as queue.db

    U->>UI: Upload + Enqueue
    UI->>J: enqueue_job()
    J->>J: new_job_id()
    J->>J: copy source into job folder
    J->>DB: INSERT job queued
    UI->>U: Success message

Processing¶

sequenceDiagram
    participant W as Worker
    participant J as jobs.py
    participant P as pipeline
    participant T as transcribe
    participant S as summarize

    W->>J: claim_next_job()
    J->>J: status = running
    W->>P: run_pipeline(job_id)
    P->>T: transcribe()
    T->>T: load Parakeet from models/
    P->>P: export TXT + SRT
    alt summary requested
        P->>S: summarize()
        P->>P: export riassunto.txt
    end
    P->>J: status = completed

Reading results (UI)¶

load_index() → job list from SQLite
User selects job → get_job(id)
job.txt_path().read_text() → content from disk

The database does not store transcribed text — only metadata and paths.

Path variables¶

Variable	Local default	Docker
`SBOBINATOR_DATA`	`./data`	`/data`
`NEMO_CACHE_DIR`	`./models`	`/models`

See Configuration.