The Ragwalla Blog

A blog about building and using an OpenAI Assistants compatible service

March 14, 2025

Key Differences Between the OpenAI Responses API and the Assistants API

The OpenAI Responses API simplifies conversation state management by handling it server-side. Developers can maintain context across interactions by including the `previous_response_id` parameter in their requests, referencing the last response's ID. This approach eliminates the need to manually track conversation history, as the API retrieves and incorporates the entire conversation chain automatically. However, it's important to note that all previous input tokens in the conversation chain are billed as input tokens.

March 10, 2025

Managing Costs with the OpenAI Assistants API

Warning: using the OpenAI Assistants API can result in unexpectedly high costs because of inherent flaws in its architecture. Ragwalla Assistants addresses these flaws head-on.

March 10, 2025

Hybrid Search With The OpenAI Assistants API

The OpenAI Assistants API does not natively support hybrid search or metadata indexing, but Ragwalla's implementation does. Hybrid search allows developers to build smarter, faster, and more precise data retrieval systems. Whether manually setting keywords or automating the process with LLM Auto mode, this approach enhances the effectiveness and efficiency of vector-based applications.

March 04, 2025

Real Use Cases for Cosine Similarity, Dot Product, and Euclidean Distance

Understanding cosine similarity, dot product, and Euclidean distance can be much easier with real-world analogies. These measures each capture “similarity” or “distance” in different ways — direction vs. magnitude vs. straight-line distance . We explore two narrative-style scenarios for each measure, showing when to use them and why.

January 13, 2025

Choosing Between Cosine Similarity, Dot Product, and Euclidean Distance for RAG Applications

Choosing the right similarity measure depends on your use case. If you’re dealing with text embeddings and want to isolate direction over magnitude, go with Cosine Similarity. When overall vector size or weighting matters, Dot Product shines. And if your model benefits from real geometric distances like clustering points in space — Euclidean Distance is the way to go.

January 06, 2025

Use Multiple Vector Stores With The OpenAI Assistants API

Use multiple vector stores in an OpenAI Assistants API service to enable parallel querying across each store that each handle 5M vectors—500x OpenAI's current limit.

November 14, 2024

Introducing Ragwalla: A Wire Compatible OpenAI Assistant

The OpenAI Assistant was a major leap forward for developers, but pitfalls can sink your application and team. The Ragwalla Assistant addresses the challenges related to scale, transparency, support and cost.