openai-python VS llama_index

In this article, we will explore the concepts behind building an AI agent in Python. We will use the OpenAI Python SDK (Responses API) for the examples, but the same ideas can be generalized to any other LLM SDK. We will use a low-level SDK with minimal abstractions so we can observe and implement most of the agent's behavior on our end.

While we work on these projects, we will integrate with the OpenAI API. I need you to hire an OpenAI integration expert. They should be fully aware of the OpenAI API https://developers.openai.com/api/docs and the cookbook https://github.com/openai/openai-cookbook. We will be using the Python SDK https://github.com/openai/openai-python so they should be experts in that.

Stack & Versions: Python 3.12, FastAPI 0.110.0, PostgreSQL 16, Redis 7.2, deployed on AWS EKS 1.29. All code hosted at https://github.com/openai/openai-python

There is no way to upload files as a part of context with Azure deployments, you have to use the OAI API [0], and without having an architecture diagram of the solution, I am not going to trust it based off of the known native limitations with Azure's OAI implementation.

[0] https://github.com/openai/openai-python/issues/2300

import requests response = requests.post( "https://arkforge.fr/trust/v1/proxy", headers={"X-Api-Key": "YOUR_API_KEY"}, json={ "target": "https://arkforge.fr/api/v1/scan-repo", "payload": {"repo_url": "https://github.com/openai/openai-python"}, }, ) proof = response.json()["proof"] print(f"Proof: {proof['proof_id']}") print(f"Verify: {proof['verification_url']}")

openai/openai-python

Know (or check) tells of older versions, such as the python sdk of OpenAI changing from a client with global state in v0.x.x, to a declared instance in v1.x.x, or numpy's change in how random generators are declared.

LLM / agentic frameworks: LangChain, LlamaIndex, LangGraph, AutoGen, MCP, RAG. (Fiddler and Razorpay both list these. "Hands-on counts, not just awareness," as Razorpay puts it.)

Open-source alternative: The DIY pattern is a LlamaIndex RAG pipeline over your CRM + filings + news. Real, but takes a quarter to ship and 18 months to make trust-grade.

The add_nodes method (line 217 in the source file) first validates that all nodes have embeddings, then batches them into chunks of 100 (configurable via the batch_size parameter). For each batch, it constructs a list of Pinecone vector dictionaries with id, values, and metadata fields, then calls self._pinecone_index.upsert with tenacity retry logic for PineconeRateLimitError and PineconeApiException.

Same minimal-first methodology will be applied to outlines, marvin, and llama_index — one paired comparison, no holes, real numbers. A PR at stanfordnlp/dspy referencing this work is open; find it linked from the semantix-ai repo.

The PR (#21014) followed their contribution guidelines. Tests passed. The integration worked.

Integration Strategy: To make this a reality, we must integrate these orchestration capabilities directly into the data and framework layers. For example, my work on introducing orchestration primitives in LlamaIndex is designed to give the “Data Layer” the “OS” capabilities it needs to manage complex agent states.

Check out the code in Pull Request #20644.

Step 2: Set up LlamaIndex and Chroma DB

Excerpt of the original code; This is a code recipe that uses OpenSearch, an open-source search and analytics tool, and the LlamaIndex framework to perform RAG over documents parsed by Docling. In this notebook, we accomplish the following: 📚 Parse documents using Docling’s document conversion capabilities 🧩 Perform hierarchical chunking of the documents using Docling 🔢 Generate text embeddings on document chunks 🤖 Perform RAG using OpenSearch and the LlamaIndex framework 🛠️ Leverage the transformation and structure capabilities of Docling documents for RAG