davidmezzetti

from txtai import Agent

# Define tools
tools = [
    "websearch",  # Runs a websearch using default engine
    "webview",  # Loads a web page
]

# Define LLM
model = "Qwen/Qwen3-4B-Instruct-2507"

davidmezzetti

import numpy as np
from txtai.ann import ANNFactory

# Index 10M vectors using llama.cpp style quants
ann = None
for _ in range(1000):
  # Generate batch of vectors
  batch = np.random.rand(10000, 768).astype(np.float32)

  if not ann:

davidmezzetti

# RAG Quick Start
# Easy to use way to get started with RAG using YOUR data
#
# For a complete application see this: https://github.com/neuml/rag
#
# TxtAI has 70+ example notebooks covering everything the framework provides
# Examples: https://neuml.github.io/txtai/examples
#
# Install TxtAI
# pip install txtai[pipeline-data]

davidmezzetti

import soundfile as sf

from txtai.pipeline import TextToAudio

# Create and run pipeline (Note that model is CC-BY-NC)
tta = TextToAudio("facebook/musicgen-stereo-medium")
speech, rate = tta("Happy 80s rock and synth for a fun startup company")

# Write to file
sf.write("out.wav", speech.T, rate)

davidmezzetti

import soundfile as sf

from txtai.pipeline import TextToSpeech

# Build pipeline
tts = TextToSpeech("neuml/kokoro-int8-onnx")

# Generate speech
speech, rate = tts(
    """Have you ever considered having a snooty British accent?

davidmezzetti

Image to parse

from txtai.pipeline import Textractor

textractor = Textractor(backend="docling", headers={"user-agent": "Mozilla/5.0"})
textractor("https://miro.medium.com/v2/resize:fit:720/format:webp/1*HHPVwIrcxYcLRvjDpwLQyQ.png")

davidmezzetti

from smolagents import WebSearchTool
from txtai import LLM

def webrag(query):
    prompt = f"""
Answer the following question using ONLY the context below.

Query: {query}
Context: {search(query)}
"""

davidmezzetti

from txtai.pipeline import Textractor

# Docling backend, split text by sections
textractor = Textractor(sections=True, backend="docling")

# BERT Paper
textractor("https://arxiv.org/pdf/1810.04805")

# PDF converted to Markdown, split on Markdown sections
# ['## BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding...

davidmezzetti

from datasets import load_dataset
from sklearn.metrics import accuracy_score
from transformers import AutoModelForSequenceClassification, AutoTokenizer
from txtai.pipeline import HFTrainer

def metrics(pred):
    labels, preds = pred.label_ids, pred.predictions.argmax(-1)

    # Calculate accuracy
    return {"accuracy": accuracy_score(labels, preds)}

davidmezzetti

from txtai import Embeddings
from txtai.pipeline import Textractor

urls = "https://github.com/neuml/txtai"
textractor = Textractor(chunker="semantic")

embeddings = Embeddings(backend="ggml", ggml={"quantize": "q4_0"})
embeddings.index((url, x) for x in textractor(url))
embeddings.save("gguf")