Python
import txtai

# Basic semantic search
embeddings = txtai.Embeddings()
embeddings.index(["Correct result", "Not what we hoped", "Perfect answer"])
results = embeddings.search("positive outcome", 1)
print(results)  # [(2, 0.85), ...]

# RAG workflow with LLM
workflow = txtai.Workflow()
workflow.load({
    "embeddings": {"path": "sentence-transformers/all-MiniLM-L6-v2"},
    "llm": {"path": "microsoft/DialoGPT-medium"},
    "tasks": [
        {"action": "embeddings", "task": "search", "select": "data"},
        {"action": "llm", "template": "Answer based on context: {data}. Question: {input}"}
    ]
})
answer = workflow("What is the main topic?")

Progress:

• Install and configure txtai
• Define embeddings configuration
• Set up pipelines for specific tasks
• Create workflows to chain operations
• Deploy API endpoints
• Test and optimize performance

Step 1: Installation and Setup

Bash
pip install txtai
# For full functionality
pip install txtai[all]

Step 2: Configure Embeddings Database

Python
import txtai

# Basic configuration
embeddings = txtai.Embeddings({
    "path": "sentence-transformers/all-MiniLM-L6-v2",
    "content": True,  # Store original content
    "objects": True   # Enable object storage
})

# Index documents
documents = [
    {"id": 1, "text": "Machine learning algorithms"},
    {"id": 2, "text": "Natural language processing"},
    {"id": 3, "text": "Computer vision techniques"}
]
embeddings.index(documents)

Step 3: Build Pipelines

Python
# Question-answering pipeline
qa = txtai.Pipelines.create("extqa")
answer = qa("What is machine learning?", ["Machine learning is AI"])

# Summarization pipeline
summary = txtai.Pipelines.create("summary")
result = summary("Long text to summarize...")

# Translation pipeline
translate = txtai.Pipelines.create("translation")
translated = translate("Hello world", "es")

Step 4: Create Multi-Step Workflows

YAML
# workflow.yml
embeddings:
  path: sentence-transformers/all-MiniLM-L6-v2

tasks:
  - action: embeddings
    task: search
    query: input
    limit: 3
    select: data

  - action: summary
    text: data

  - action: llm
    template: "Summarize this content: {text}"

Step 5: Deploy API

Python
# app.yml
embeddings:
  path: sentence-transformers/all-MiniLM-L6-v2

# Run API
# CONFIG=app.yml uvicorn "txtai.api:app"