davidmezzetti

Text extraction MCP service

Extract text using txtai, docling, docker. Service available via Model Context Protocol (MCP).

/tmp/config/config.yml

# Enable MCP server
mcp: True

# Enable file uploads

davidmezzetti

from txtai import Agent

agent = Agent(
    tools=["http://mcp.server/path"],
    model="LLM path"
)

davidmezzetti

Wikipedia Embeddings MCP Server

config.yml

# Enable MCP server
mcp: True

# Load Wikipedia Embeddings index
cloud:
 provider: huggingface-hub

davidmezzetti

Evaluating QwQ-32B

Blog: https://qwenlm.github.io/blog/qwq-32b/

Model: https://huggingface.co/Qwen/QwQ-32B

License: Apache 2.0

from txtai import LLM

davidmezzetti

from txtai import Embeddings

# Start the indexing run
embeddings = Embeddings(content=True)
embeddings.index(stream(), checkpoint="checkpoint dir")

# Elapsed time ⏳ then ⚡💥🔥
# error, power outage, random failure

# Fix the issue 🧑‍🔧⚙️

davidmezzetti

from txtai import Embeddings

embeddings = Embeddings(content=True, graph=True)
embeddings.index(...)

# Standard Vector Search
embeddings.search("vector search query")

# Vector SQL query
embeddings.search("""

davidmezzetti

🤦 This is DeepSeek

from txtai import LLM
llm = LLM("casperhansen/deepseek-r1-distill-llama-8b-awq")

llm("Do you think the USA is a good or bad country?", maxlength=512, defaultrole="user")

davidmezzetti

davidmezzetti

from txtai import Embeddings

# In-memory data
data = [{"name":"John", "age": 16}, {"name":"Jon", "age": 45},{"name":"Sarah", "age": 18}]

# Vector embeddings index with content storage
embeddings = Embeddings(content=True, columns={"text": "name"})
embeddings.index(data)

# Vector similarity

davidmezzetti

🤖 Agent Team 🤖

This example shows how to build an Agent having multiple agents as tools. This "Agent Team" works together to answer questions.

A more basic example can be found here

Install

pip install git+https://github.com/neuml/txtai