Feature Suggestions: OpenAI Proxy Support, Customizable Summarization Prompts, and ElevenLabs Transcription Integration

[benzik]

Hi Meetily team! 👋

First of all, I want to express my genuine appreciation for this incredible project. Meetily is truly outstanding! The privacy-first approach, local AI processing, and the elegant architecture you've built are exactly what the market needs. As someone who works extensively with automation platforms, Docker deployments, and AI integrations, I can really appreciate the engineering quality and thoughtfulness behind Meetily. You've created something special here, and the 8.3k stars speak for themselves! 🎉

I've been using Meetily and exploring its capabilities, and I have a few feature suggestions that I believe would make it even more powerful for power users and enterprise scenarios.

1. OpenAI API Proxy Support

Problem: Currently, Meetily supports direct connections to AI providers. However, many users and organizations utilize OpenAI-compatible proxy servers for various reasons:

• Cost optimization and billing control.
• Compliance requirements and audit logging.
• Regional restrictions/firewalls.
• Accessing custom models served via standard endpoints.

Suggestion: Add a configuration option in the AI provider settings to specify a custom Base URL (API Endpoint). This would allow users to point Meetily to their own proxy infrastructure while maintaining compatibility with the OpenAI API format.

Implementation Concept:

AI Provider: [OpenAI (Custom / Proxy)]
Base URL:    [https://my-proxy.example.com/v1]  <-- New Field
API Key:     [sk-proj-...]
Model Name:  [gpt-4o]

2. Editable System Prompt (Custom Instructions)

Problem: Currently, the summarization system prompt is hardcoded/fixed. This makes it difficult to adapt the output for specific use cases. A medical consultation requires a completely different summary structure than a product planning meeting, a technical code review, or a sales call.

Suggestion: The priority here is not to build a complex template system immediately, but simply to allow the user to edit the current system prompt.
Giving users access to the raw system instruction allows us to tweak the personality, format, and focus of the summary to fit our specific needs.

Future evolution: Once the prompt is editable, you could eventually add a "Saved Templates" feature, but the core requirement right now is just the ability to customize the instruction string.

User Flow Idea:

Settings -> Summarization

[x] Use Custom System Prompt

System Instruction:
+---------------------------------------------------------------+
| You are an expert technical writer. Summarize the following   |
| transcript focusing on:                                       |
| 1. Architectural decisions                                    |
| 2. API contract changes                                       |
| 3. Docker configuration updates                               |
|                                                               |
| Output format: Markdown list.                                 |
+---------------------------------------------------------------+

3. Cloud Transcription API Support (Low Priority / Future Idea)

Context: This is a "nice-to-have" feature for the future. Whisper and Parakeet are excellent for local privacy, but they consume local hardware resources.

Suggestion: Add support for third-party Cloud STT (Speech-to-Text) APIs (such as ElevenLabs, OpenAI Whisper API, Google, etc.).

Why: While local processing is great for privacy, some users might prefer to offload the processing power to the cloud to keep their local system fast, or they might require specific models provided by services like ElevenLabs. Giving users the choice between "Local (Privacy)" and "Cloud (Performance/Convenience)" would make the tool more versatile.

Configuration Concept:

Transcription Engine:
 (o) Whisper (Local - Privacy First)
 ( ) Parakeet (Local - Fast)
 ( ) Cloud API (Offload Processing)
     
     Provider: [ElevenLabs / OpenAI / Other ▼]
     API Key:  [sk-.........................]

---

Why These Features Matter

These enhancements would position Meetily as an even more flexible solution:

• Proxy support enables enterprise deployment scenarios with centralized API management.
• Custom Prompts give users full control over how their meetings are summarized.
• Cloud Transcription offers flexibility for hardware-constrained environments.

All of this aligns perfectly with Meetily's core philosophy: giving users control and flexibility.

I'd be happy to discuss these ideas further or even contribute to the implementation if you're interested. Once again, amazing work on this project! Looking forward to seeing how Meetily evolves. 🚀

Best regards!

[richardchen874-sys]

For meeting tools, I would separate transcription cost from summarization/reasoning cost. The high-volume part is usually transcription and cleanup, while the higher-quality model is only needed for final summaries, action items, or domain-specific reasoning. I am testing an OpenAI-compatible multi-model API layer around official Chinese models, and meeting workflows are a useful cost/quality comparison case.