[Suggestion] Haystack audio transcription pipeline (Python)

[deepgram-robot]

Integration: Haystack Audio Transcription Pipeline

What this should show

A Python example demonstrating how to use Deepgram as a custom Haystack 2.x component for audio transcription in a RAG pipeline. The example should:

• Create a custom Haystack @component that accepts audio file paths or URLs
• Transcribe audio via Deepgram Pre-recorded STT (Nova-3)
• Output Haystack Document objects with transcript text + metadata (speaker labels, timestamps, confidence)
• Include a retrieval demo using an in-memory document store
• Support both single-file and batch audio transcription

Credentials likely needed

• DEEPGRAM_API_KEY

---

Original request:

What to build

A working example demonstrating Deepgram as a Haystack component for audio transcription in a RAG pipeline — loading audio files, transcribing with Deepgram STT, and feeding transcripts into a Haystack retrieval pipeline.

Why this matters

Haystack (by deepset) is a leading enterprise NLP/RAG framework used by teams building production search and retrieval systems. Developers building audio-aware RAG pipelines need a reference integration showing how to use Deepgram as an audio ingestion source. There is currently no example of Deepgram + Haystack working together, despite Haystack's growing adoption for enterprise AI.

Suggested scope

• Language: Python
• Framework: Haystack 2.x (haystack-ai package)
• Deepgram APIs: Pre-recorded STT (Nova-3)
• What it does: Custom Haystack @component that accepts audio file paths or URLs, transcribes via Deepgram, and outputs Haystack Document objects with transcript text + metadata (speaker labels, timestamps, confidence)
• Includes: Pipeline YAML config, example audio file, retrieval demo with an in-memory document store
• Complexity: Medium — single Python file + pipeline config

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the pattern clearly
• Uses current SDK version and Haystack 2.x
• Demonstrates both single-file and batch audio transcription
• Output documents include Deepgram metadata (speakers, timestamps)

---

Raised by the DX intelligence system.

[github-actions]

@deepgram/devrel — new suggestion from @deepgram-robot (no write access). Add the approved label to process it, or close to decline.

[github-actions]

Ooo, we'll get right on that! 🎉

The Engineer will pick this up shortly and build a Haystack + Deepgram audio transcription pipeline example — a custom Haystack 2.x component that transcribes audio with Deepgram STT and feeds transcripts into a retrieval pipeline.

I'll keep this issue open so you can track progress — we'll close it automatically when the PR merges.

If you have any extra context (preferred language, specific API, or credentials you're already using), drop it here — it helps the build go faster!

[github-actions]

Built in #190