Whisper
Transcripts speech to text via a Whisper-Server.
Domains to whitelist
github.com — whisper-server binary releases (ggml-org/whisper.cpp) & ffmpeg (BtbN/FFmpeg-Builds)
objects.githubusercontent.com — GitHub release asset CDN
huggingface.co — Whisper GGML model downloads
DSGVO / GDPR Compliance
All audio data is processed locally on the server (using a local server only). No data is transmitted to any cloud service (Google, Microsoft, OpenAI, etc.). This makes it the ideal solution for organizations that require DSGVO-compliant speech recognition.
Plugin Properties
| Property | Type | Default | Description |
|---|---|---|---|
Active_AI | String | — | Must contain whisper to activate |
AI_Remove | — | — | If contains whisper, clean up all whisper files |
AI_Whisper_ModelUrl | URL | ggml-small | GGML model URL. Alternatives: ggml-base, ggml-medium, ggml-large-v3-turbo-q5_0 |
AI_Whisper_Port | Int | 8393 | Local port for whisper-server |
AI_Whisper_Release | String | v1.7.6 | whisper.cpp release tag for binary downloads |
AI_Whisper_ReleaseBaseUrl | URL | GitHub release | Base URL prefix used to download whisper.cpp release assets (release tag is appended automatically). |
AI_Whisper_NoGpu | Bool | false | Set true to disable GPU and force CPU-only |
AI_Whisper_Threads | Int | physical cores | CPU threads for whisper-server |
AI_Whisper_MaxRAMPercent | Double | 101.0 | RAM usage threshold (%) — blocks requests when exceeded |
AI_Whisper_MaxComputePercent | Double | 101.0 | Compute usage threshold (%) — gates on GPU% (CUDA) or CPU% (fallback). Blocks requests when exceeded |
AI_Whisper_MaxCPUPercent | Double | — | Legacy alias for MaxComputePercent (accepted as fallback) |
AI_Whisper_ExternalUrl | URL | — | Base URL of an OpenAI-compatible speech-to-text API (e.g. https://api.openai.com). When set, the local whisper-server is NOT started. |
AI_Whisper_ExternalApiKey | String | — | Bearer token / API key for the external API. |
AI_Whisper_ExternalModel | String | whisper-1 | Model name to send in the model field of the external API request. |
Functions
Initializes the AI components by reading configuration properties. Specifically, it acquires the msExpirationIDedImages from the plugin property "AI_CachedImageExpiration". Subclasses should call this method at the beginning of their own initialize implementation.
Initiates a task that removes unused images that're expired (msExpirationIDedImages) from the cache (cacheIDedImages).
Rejects tenant-level installation. CodBi must be installed as a system plugin because its AI services (Whisper, LLAMA) bind local server ports and manage heavyweight processes that would conflict when instantiated once per tenant.