0.2.0

What's Changed

• chore(deps): update aquasecurity/trivy-action action to v0.34.0 [security] by @renovate[bot] in #35
• chore(deps): update actions/upload-artifact action to v7 by @renovate[bot] in #34
• chore(deps): update docker/login-action action to v4 by @renovate[bot] in #36
• chore(deps): update aquasecurity/trivy-action action to v0.34.2 by @renovate[bot] in #30
• chore(deps): update aquasecurity/trivy-action action to v0.35.0 by @renovate[bot] in #40
• chore(deps): update docker/metadata-action action to v6 by @renovate[bot] in #39
• chore(deps): update docker/build-push-action action to v7 by @renovate[bot] in #38
• chore(deps): update docker/setup-buildx-action action to v4 by @renovate[bot] in #37
• chore(deps): update dependency python to 3.14 by @renovate[bot] in #22
• feat(transcription): implement start and stop control for realtime by @tommaso-ascani in #33

New Contributors

• @tommaso-ascani made their first contribution in #33

Full Changelog: 0.1.0...0.2.0

TTS

Release 0.1.0

Highlights

This release introduces Text-to-Speech (TTS) support via Deepgram, adding two new HTTP API endpoints and significantly expanding test coverage.

New Features

• POST /api/get_speech — Text-to-Speech endpoint that synthesizes text into MP3 audio using Deepgram TTS voices (da8a4d8, 599b6a4)

  • Accepts text (or input), model, language, sample_rate, bit_rate parameters
  • Automatic text chunking (≤2000 chars) via RecursiveCharacterTextSplitter for long inputs
  • MP3 chunk concatenation and volume boost (+8 dB) via ffmpeg post-processing
  • Language-based model auto-selection when model is omitted
  • Uses Deepgram Python SDK (deepgram-sdk) instead of raw HTTP calls
  • Returns StreamingResponse with audio/mpeg content type

• GET /api/get_models — Returns available Deepgram TTS voice models, with optional language filter (22d6dc3)

  • Ships with 100+ built-in Aura/Aura-2 voice models (en, it, es, de, fr, nl, ja)

Bug Fixes

• Accept audio/mpeg content type in POST /api/get_transcription for MP3 file uploads (ae50981)
• Enforce MP3-only output in TTS endpoint — reject non-MP3 encoding or container params with 400 (2151bb9)
• Use Deepgram SDK for TTS instead of raw HTTP/curl, improving error handling and timeout management (599b6a4, 3483948)

Tests

• Added comprehensive TestGetSpeech test class with 13 new tests covering: (8663f59, 3483948)
  • MP3 response format and filename validation
  • Text chunking and concatenation for long inputs
  • Unknown parameter filtering
  • Language-based model selection and explicit model precedence
  • Missing text → 400, unsupported language → 400, non-MP3 encoding/container → 400
  • Deepgram timeout → 504, Deepgram API error propagation
  • TTS → STT roundtrip integration test (Italian, requires DEEPGRAM_API_KEY)
• Updated existing test to use video/avi for rejected file type instead of audio/mp3

Documentation

• Updated README with TTS API documentation and Italian voice model table (d49071d, c16a922)

Chores

• Added AGENTS.md to .gitignore (3d36b35)
• Added ffmpeg to container image (Containerfile)

Fix memory leak whe Deepgram connection fails

NethServer/dev#7872

Fix: Outbound caller ID resolution

When Asterisk reports the trunk instead of the endpoint as the connected number/name on outbound calls, Satellite now falls back to CALLERIDNUMINTERNAL and CALLERIDNAMEINTERNAL.
If present, these values are used to correctly update the caller details.
NethServer/dev#7795

API Authentication

0.0.7

feature(db): Add deleted_at db field (#26)

Transcription API persistence

The get_transcription API now allow to persist transcriptions in a pgsql database

0.0.5

Full Changelog: 0.0.4...0.0.5

0.0.4

What's Changed

• fix: concurrent call handling and dynamic speaker mapping by @edospadoni in #6

New Contributors

• @edospadoni made their first contribution in #6

Full Changelog: 0.0.2...0.0.4

Fix delay and speaker information

Merge pull request #5 from nethesis/Stell0-patch-3

fix: add speaker counterpart information to transcription

0.0.2

What's Changed

• enhancement(ai):Update model and temperature in summarization by @Stell0 in #3
• fix: add speaker counterpart information to transcription by @Stell0 in #5

Full Changelog: 0.0.1...0.0.2