AI Supervisor Worker
Installing and using AI Supervisor Worker: connecting to MikoPBX, local models, job processing, voice analytics, and diagnostics on a Mac.
AI Supervisor Worker is a standalone macOS application that performs local AI analysis of completed MikoPBX transcripts. The application receives jobs from the AI Supervisor module, uses a local language model, processes the call recording, and sends a structured result back to the PBX.
The application does not select which calls to analyze and does not store the server-side queue. Transcript import, analysis components, model profile, result language, attention rules, and saved results are managed by the module in MikoPBX.
Requirements and compatibility
A Mac with Apple silicon.
macOS 14 or later.
MikoPBX 2025.1.1 or later.
Network access from the Mac to MikoPBX.
A local Ollama runtime or an OpenAI-compatible local endpoint.
Internet access for the initial runtime and model installation.
Application navigation
The sidebar contains:
Overview
Worker readiness, connection, current work, and summary statistics.
Activity
Current job and local processing history with filters and export.
Models
Ollama, the model selected by the PBX, analysis components, and local models.
Diagnostics
Connection, runtime readiness, automatic recovery, and logs.
Settings
MikoPBX connection, local behavior, and analysis parameters.
There is no longer a separate Registration screen. The PBX address, token, worker name, and Worker UID are under Settings.
Overview
Overview is the main AI Supervisor Worker status screen. It includes:
the overall status and primary available action;
MikoPBX connection status and last contact time;
calls analyzed today and the success rate;
current work and job progress;
recent local jobs.
Main states:
Setup required
Open the connection settings and enter the PBX address and worker token.
Checking local AI
Wait while Ollama and the models are checked.
Model download in progress
Do not close the application until the download finishes.
Local AI needs attention
Open Models.
Ready to start
Click Start worker.
Everything works
The worker is connected and checks the queue automatically.
Analyzing a call
The current job is being processed locally.
Connecting to MikoPBX
The worker automatically restores the connection.
Worker needs attention
Open Diagnostics.
Activity
Activity shows the current work and the job history saved on this Mac. The history persists after the application restarts, but it is not the MikoPBX server-side queue.
Available controls include:
periods of 24 hours, 7 days, and 30 days;
In Progress, Deferred, Completed, and Failed filters;
search by job and Call ID;
expandable rows with the model, analysis type, components, and error text;
export of visible rows to CSV.
The Deferred status means that the worker safely released the job for another processing attempt, for example when a local model or dependency was temporarily unavailable.
Models
Models shows the local runtime, the analysis profile from MikoPBX, and installed models.
The page includes:
an Ollama installation card if the runtime is not yet available;
endpoint and model readiness;
the model selected by MikoPBX;
the connected analysis profile: Call analysis, Voice analytics, Text emotion, and Acoustic emotion;
a list of installed local models;
Voice Analytics model status;
current download progress.
The Refresh button checks the runtime, model, and voice components again. The current interface has no separate Check Runtime or Check Model buttons.
The following actions are available for local models:
Show in Finder
Open the model or related files in Finder.
Delete
Delete the model from the managed local runtime.
Deletion is disabled while the worker is running or the model is in use. If MikoPBX selects the deleted model again, the worker can download it again.
Diagnostics
Diagnostics combines status checks and the local log.
The top section shows:
MikoPBX connection status;
the token and Worker UID;
the local endpoint and selected model;
recovery after a transient error;
Check Connection and Open Logs Folder buttons.
The log supports:
1 hour, Today, and All periods;
All, Warnings, and Errors levels;
event search;
copying visible entries;
exporting to a file.
Settings
Settings are divided into three groups.
Connection
This group contains the PBX address, token stored in Keychain, worker name, Worker UID, and connection test.
Required fields:
MikoPBX Address
Full address beginning with http:// or https://.
Worker token
Key created under AI Supervisor → Settings → Workers.
Worker name
A friendly name for this Mac in MikoPBX.
Worker UID
A stable technical identifier.
Click Check Connection. The application first requests GET /worker-api-contract, verifies ModuleAISupervisor, the API version, and the required capabilities, and then uses the token to register.
General
App language
Interface language. Changing it requires an application restart.
Start with macOS
Open the application after the user signs in.
Keep worker running
Resume automatically after the network or PBX connection is restored.
TLS verification
Verify the MikoPBX HTTPS certificate.
Custom CA file
Custom PEM/CRT/CER file for a private certificate.
Temporary files
Directory for temporary recordings and intermediate data.
Analysis
Analysis profile
Opens Models; active components are supplied by the MikoPBX job.
OpenAI-compatible endpoint
Local endpoint for structured analysis; the default is http://127.0.0.1:11434/v1.
Emotion model profile
Model used for text-based emotion analysis.
Max fragments
Maximum number of fragments used for emotion analysis.
Minimum call length
Shorter calls skip the optional emotion enrichment stage.
The primary model profile, context size, and analysis components are configured in MikoPBX and sent with every job.
Last updated
Was this helpful?