Tanmay Deep SharmaandGitHub
21c0f4dc52
Two production-grade fixes to the existing audio transcription service. **Independent of the WhatsApp Calling work** — these affect every audio attachment that goes through Whisper (voice notes, call recordings, voicemails, etc.). ## Closes - [PLA-151 — PR-5: Recording Upload + Transcription Pipeline](https://linear.app/chatwoot/issue/PLA-151/pr-5-recording-upload-transcription-pipeline) ## Why this is needed ### 1. Whisper rejects payloads larger than 25 MB OpenAI's [Whisper API](https://platform.openai.com/docs/guides/speech-to-text) hard-caps file uploads at 25 MB. Long audio recordings — voice notes from chatty contacts, ~70+ min Opus call recordings — currently hit OpenAI with the full payload and 413 (\`Payload Too Large\`). The job retries via the existing \`Faraday::BadRequestError\` discard path, but the agent still sees a transcription failure for an attachment we knew was too big up front. This PR adds a pre-flight \`audio_too_large?\` check via the blob's \`byte_size\` and returns a controlled error without hitting OpenAI. The audio attachment is preserved (agents can still listen), only the transcription is skipped. ### 2. Whisper hallucinates on silence at non-zero temperature At \`temperature: 0.4\` (the previous value), Whisper produces well-documented hallucinated repeats on silence and near-silent segments — e.g. \`Oh, dear. Oh, dear. Oh, dear.\` filling the transcript. This shows up in real recordings whenever there's a hold or quiet moment. \`temperature: 0.0\` matches OpenAI's recommended default for transcription and eliminates the spirals. Reference: [openai/whisper#928](https://github.com/openai/whisper/discussions/928), [openai-python#1010](https://github.com/openai/openai-python/issues/1010). ## Are WhatsApp call recordings already handled? Yes — by the existing pipeline, **before this PR**: \`\`\` Browser MediaRecorder → upload_recording (PR-4) → @call.message.attachments.create!(file_type: :audio, ...) → Enterprise::Concerns::Attachment#enqueue_audio_transcription (after_create_commit hook) → Messages::AudioTranscriptionJob.perform_later(attachment.id) → Messages::AudioTranscriptionService → Whisper \`\`\` The \`after_create_commit\` hook already fires for every audio attachment regardless of source. PR-4's \`upload_recording\` endpoint creates the attachment; the existing job/service take it from there. No new wiring needed. This PR just makes the existing service more robust: - Calls longer than ~70 min (Opus 48 kbps) no longer 413 against OpenAI - Quiet recordings no longer produce hallucinated transcripts ## How to test \`\`\`ruby # In rails console with a real audio attachment: service = Messages::AudioTranscriptionService.new(Attachment.audio.last) # Normal-sized audio: unchanged behaviour service.perform # => { success: true, transcriptions: ... } # Large audio: new guard returns error instead of 413-ing OpenAI allow(attachment.file.blob).to receive(:byte_size).and_return(30.megabytes) service.perform # => { error: 'Audio too large for Whisper' } \`\`\` Existing transcription specs cover the happy path; one new spec exercises the byte-limit guard. ## Risk Low. Both changes are pre-flight guards or parameter values — they reduce the surface of OpenAI calls that can fail. Failure to transcribe is already non-fatal (the audio attachment is preserved either way).
Chatwoot
The modern customer support platform, an open-source alternative to Intercom, Zendesk, Salesforce Service Cloud etc.
Chatwoot is the modern, open-source, and self-hosted customer support platform designed to help businesses deliver exceptional customer support experience. Built for scale and flexibility, Chatwoot gives you full control over your customer data while providing powerful tools to manage conversations across channels.
✨ Captain – AI Agent for Support
Supercharge your support with Captain, Chatwoot’s AI agent. Captain helps automate responses, handle common queries, and reduce agent workload—ensuring customers get instant, accurate answers. With Captain, your team can focus on complex conversations while routine questions are resolved automatically. Read more about Captain here.
💬 Omnichannel Support Desk
Chatwoot centralizes all customer conversations into one powerful inbox, no matter where your customers reach out from. It supports live chat on your website, email, Facebook, Instagram, Twitter, WhatsApp, Telegram, Line, SMS etc.
📚 Help center portal
Publish help articles, FAQs, and guides through the built-in Help Center Portal. Enable customers to find answers on their own, reduce repetitive queries, and keep your support team focused on more complex issues.
🗂️ Other features
Collaboration & Productivity
- Private Notes and @mentions for internal team discussions.
- Labels to organize and categorize conversations.
- Keyboard Shortcuts and a Command Bar for quick navigation.
- Canned Responses to reply faster to frequently asked questions.
- Auto-Assignment to route conversations based on agent availability.
- Multi-lingual Support to serve customers in multiple languages.
- Custom Views and Filters for better inbox organization.
- Business Hours and Auto-Responders to manage response expectations.
- Teams and Automation tools for scaling support workflows.
- Agent Capacity Management to balance workload across the team.
Customer Data & Segmentation
- Contact Management with profiles and interaction history.
- Contact Segments and Notes for targeted communication.
- Campaigns to proactively engage customers.
- Custom Attributes for storing additional customer data.
- Pre-Chat Forms to collect user information before starting conversations.
Integrations
- Slack Integration to manage conversations directly from Slack.
- Dialogflow Integration for chatbot automation.
- Dashboard Apps to embed internal tools within Chatwoot.
- Shopify Integration to view and manage customer orders right within Chatwoot.
- Use Google Translate to translate messages from your customers in realtime.
- Create and manage Linear tickets within Chatwoot.
Reports & Insights
- Live View of ongoing conversations for real-time monitoring.
- Conversation, Agent, Inbox, Label, and Team Reports for operational visibility.
- CSAT Reports to measure customer satisfaction.
- Downloadable Reports for offline analysis and reporting.
Documentation
Detailed documentation is available at chatwoot.com/help-center.
Translation process
The translation process for Chatwoot web and mobile app is managed at https://translate.chatwoot.com using Crowdin. Please read the translation guide for contributing to Chatwoot.
Branching model
We use the git-flow branching model. The base branch is develop.
If you are looking for a stable version, please use the master or tags labelled as v1.x.x.
Deployment
Heroku one-click deploy
Deploying Chatwoot to Heroku is a breeze. It's as simple as clicking this button:
Follow this link to understand setting the correct environment variables for the app to work with all the features. There might be breakages if you do not set the relevant environment variables.
DigitalOcean 1-Click Kubernetes deployment
Chatwoot now supports 1-Click deployment to DigitalOcean as a kubernetes app.
Other deployment options
For other supported options, checkout our deployment page.
Security
Looking to report a vulnerability? Please refer our SECURITY.md file.
Community
If you need help or just want to hang out, come, say hi on our Discord server.
Contributors
Thanks goes to all these wonderful people:
Chatwoot © 2017-2026, Chatwoot Inc - Released under the MIT License.