Daft Examples

An examples hub for running Multimodal AI Workloads on Daft

Getting Started

This repository is organized into three sections:

1. Usage Patterns - Small atomic demonstrations of core features.
2. Use Cases - Entire Pipelines built in Daft.
3. Notebooks - End to End tutorials on working with Daft in an interactive Jupyter Notebook

To get started, run the following in your terminal.

git clone https://github.com/Eventual-Inc/daft-examples.git
cd daft-examples
make setup

Running Examples

This project leverages uv scripts for dependency management isolation. This is because many of the scripts in this repository have different requirements. The venv that is created by the setup make target is just for type hints and notebooks.

You can run any script like:

uv run usage_patterns/prompt/prompt.py

If you don't have uv, check out this installation guide.

System dependencies

• Some examples leverage libraries like soundfile or PyAv to process audio and video files which requires ffmpeg.

Usage Patterns

Prompt

• prompt/prompt.py - Basic prompt example using OpenRouter with anime quote classification
• prompt/prompt_chat_completions.py - Use chat completions with personas from Nemotron dataset via LM Studio
• prompt/prompt_files_images.py - Multimodal prompting with images and PDFs using Gemini 2.5 Flash via OpenRouter
• prompt/prompt_github.py - Analyze markdown files from GitHub repositories with GPT-5
• prompt/prompt_openai_web_search.py - Perform web searches using OpenAI GPT-5 with web_search tools and store results in Supabase
• prompt/prompt_pdfs.py - Analyze PDF documents with GPT-5-nano reasoning models
• prompt/prompt_personas.py - Multi-persona business strategy analysis with web search and collective intelligence synthesis
• prompt/prompt_session.py - Use custom OpenRouter provider through Daft sessions
• prompt/prompt_structured_outputs.py - Get structured outputs using Pydantic models with OpenRouter
• prompt/prompt_synthetic_customer_discovery.py - Generate synthetic customer discovery responses from diverse personas using structured outputs

Embeddings

• embed/embed_images.py - Generate image embeddings using Apple's AIMv2 model
• embed/embed_pdf.py - Extract text from PDFs, chunk with spaCy, and embed sentences
• embed/embed_text_providers.py - Compare text embeddings across providers (Transformers, OpenAI, LM Studio)
• embed/embed_video_frames.py - Extract and embed video frames from YouTube videos
• embed/shot_boundary_detection.py - Detect scene cuts and dissolves in videos using frame embeddings
• embed/similarity_Search.py - Semantic search using cosine distance between text embeddings

Classification

• classify/classify_image.py - Classify images using CLIP model with custom labels
• classify/classify_text.py - Multi-label text classification using BART model

Common Crawl

• commoncrawl/chunk_embed.py - Process Common Crawl data: chunk text with spaCy and embed sentences
• commoncrawl/show.py - Query and display MIME types from Common Crawl datasets

UDFs

• udfs/cls_with_types.py - Class-based UDFs with TypedDict, Pydantic, batch processing, and async functions
• udfs/udf.py - Simple UDF example to extract file names from File objects

I/O

• io/read_audio_file.py - Read and resample audio files, write to MP3 format
• io/read_pdfs.py - Discover and download PDF files from URLs
• io/read_video_files.py - Extract video metadata, keyframes, and audio from video files

Use Cases

Transcription

• transcribe/faster_whisper_schema.py - Schema definitions for Faster Whisper transcription results
• transcribe/key_moments_extraction.py - Extract key moments from transcripts and clip audio for short-form content
• transcribe/transcribe_faster_whisper.py - Transcribe audio with Faster Whisper and Voice Activity Detection (VAD)

Voice AI Analytics

• voice_ai_analytics/voice_ai_analytics.py - Complete pipeline: transcription, summarization, translation to Chinese, segment embeddings, and RAG Q&A
• voice_ai_analytics/voice_ai_analytics_openai.py - Voice analytics with OpenAI Whisper, GPT-5 summarization, and Spanish translation
• voice_ai_analytics/voice_ai_tutorial.py - Voice AI analytics tutorial: transcription, summarization, Chinese translation, and embeddings

Notebooks

• notebooks/getting_started_with_common_crawl.ipynb - Interactive Common Crawl tutorial
• notebooks/voice_ai_analytics.ipynb - Voice AI analytics walkthrough
• notebooks/window_functions.ipynb - Window functions examples and usage