[FEATURE]: Add Option to Disable AI Summarization (Transcription-Only Mode)

[BuddhiD-Workaholic]

Feature Description

Add an option to disable AI-powered summarization and use Meetily as a transcription-only tool.

Problem Statement

Currently, Meetily always runs an LLM to generate meeting summaries. While this is useful for many users, there are scenarios where users only need a high-quality transcript.

For users running Meetily locally, even a small language model still consumes CPU/GPU resources, memory, startup time, and power. Some users also have privacy or simplicity requirements and don't want to run any LLM at all.

I personally just want accurate transcription without any summarization.

Proposed Solution

Add a simple toggle or setting such as:

• Enable AI Summarization (On/Off)

When disabled:

• Skip loading any LLM.
• Do not generate summaries or action items.
• Only produce the transcript.
• Everything else should continue working normally.

This makes Meetily usable as a lightweight local transcription tool while keeping summarization available for users who want it.

User Story

As a user who only needs meeting transcripts,

I want to disable AI summarization,

So that I can use Meetily without running a local language model and save system resources.

Acceptance Criteria

• Users can enable or disable AI summarization from Settings.
• When summarization is disabled, no LLM is loaded or executed.
• Meetily still generates and saves complete meeting transcripts.
• Existing summarization functionality remains unchanged when the feature is enabled.

Technical Considerations

• This could simply bypass the summarization pipeline after transcription completes.
• The application should avoid initializing or downloading any LLM when summarization is disabled.
• The setting should persist across application restarts.

Alternatives Considered

An external API-based summarization option could work, but many users—including myself—don't want summarization at all. A simple on/off toggle is the most flexible and lightweight solution.

Additional Context

I love the project and mainly use it for local transcription. Having the ability to completely disable summarization would make Meetily an excellent lightweight transcription tool for users who don't need AI-generated summaries.

Checklist

• I have searched for similar feature requests
• I have provided all required information
• I have included any relevant screenshots/mockups
• I have described the problem and proposed solution clearly

[safvanatzack]

Hi,
if you're referring to the automatic summary that's generated after the recording stops, you can turn it off in the Settings.
Settings -> Summary -> Auto summary(OFF)

Hope it helps
Thanks

[dandeannie]

Hi @safvanatzack , thanks for pointing that out, but I think that setting only suppresses the summary output after it's generated, rather than skipping LLM initialization/execution entirely. The core ask here (and the resource/privacy concern) is avoiding loading or running the LLM at all when summarization is disabled, not just hiding the result.

@BuddhiD-Workaholic , can you confirm whether "Auto summary (OFF)" in Settings still spins up the LLM in the background, or does it skip that step entirely? That would clarify whether this is already solved or still open.

If it turns out the LLM is still being loaded/run behind the scenes, I'd like to work on this as part of a contribution project for my lab (ROSP Lab) and happy to look at gating the summarization pipeline initialization behind the toggle so no LLM is loaded when it's off. Could a maintainer assign this to me if that's the case?

[happyfamily9500-ux]

Hi have fied this issue
Issue fixed. No. 677

​Since this feature makes total sense—saving compute cycles, memory, and startup latency for transcript-only users—here is how this feature can be structured and implemented cleanly into Meetily's codebase.
​🛠️ Implementation Plan for Contributor / Maintainer
​1. Settings / Configuration
​Add a persistent toggle state (e.g., using your local storage / settings store like electron-store, tauri-plugin-store, or standard user config JSON).
​Default Value: true (maintains backward compatibility for existing users)
​2. UI Component (Settings Page)
​Add a toggle under Settings > AI / Summarization:
// Example UI Component Snippet

Enable AI Summarization

Disable to skip LLM initialization and run Meetily as a lightweight transcription tool.

updateSettings({ enableSummarization: checked })} />

3. Pipeline Logic Update ​Modify the post-recording workflow to conditionally bypass the LLM phase. ​Before: \text{Audio Recording} \longrightarrow \text{Transcription Engine (Whisper)} \longrightarrow \text{LLM Summarizer} \longrightarrow \text{Save Results} After: \text{Audio Recording} \longrightarrow \text{Transcription Engine (Whisper)} \longrightarrow \begin{cases} \text{LLM Summarizer} \rightarrow \text{Save Full Results} & \text{if } \mathbf{Summary = Enabled} \\ \text{Save Transcript Only} & \text{if } \mathbf{Summary = Disabled} \end{cases} Code Logic Example: async function processMeeting(audioFilePath: string) { // 1. Always run transcription const transcript = await transcriptionEngine.transcribe(audioFilePath);

// 2. Conditionally run LLM
let summary = null;
let actionItems = [];

if (userSettings.enableSummarization) {
// Lazy-load or initialize LLM model ONLY when needed
const llm = await getLLMInstance();
const result = await llm.generateSummary(transcript);
summary = result.summary;
actionItems = result.actionItems;
}

// 3. Save meeting record
return await saveMeeting({
transcript,
summary: summary ?? "Summarization was disabled for this meeting.",
actionItems,
createdAt: new Date(),
});
}
4. LLM Initialization Guard (Crucial Step)
​Ensure the local LLM runtime (Ollama, Llama.cpp, ONNX, etc.) is lazy-loaded so that disabling summarization prevents the model binaries from being loaded into RAM/VRAM on application startup.// Avoid initializing during app startup if summarization is OFF
if (settings.enableSum.marization) {
await initializeLLM();
}
Verification & Acceptance ChecklistTest CaseExpected OutcomeStatus
Toggle Off & RecordTranscript saved successfully; CPU/GPU usage stays low during post-processing.⬜
App Startup (Toggle Off)Memory footprint remains minimal; no LLM binaries loaded into RAM.⬜
Toggle On & RecordComplete meeting summary and action items are generated as normal.⬜
PersistenceToggle setting remains set after restarting the application.⬜ mearge it with meetup/meetily