To build a RAG pipeline with Azure AI Search, prepare searchable source content, retrieve relevant passages, and pass those passages to a language model to generate a grounded response. For a new conversational system, Microsoft recommends evaluating agentic retrieval first; classic RAG remains a practical fit when you need a simpler single-query flow, GA-only capabilities, or orchestration you already control. The right choice depends on your queries, latency and cost requirements, and the maturity level you can deploy.
How does an Azure AI Search RAG pipeline work?
A RAG system separates finding evidence from writing an answer. Azure AI Search retrieves relevant content from your indexed data; your application or an agent supplies that grounding content to a language model. The model then generates a response using the evidence rather than relying only on its training data. Search results can include source references and execution activity, which your application can use to show citations or inspect retrieval behavior. Microsoft’s RAG overview describes the classic pattern and its alternatives.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
MINISFORUM MS-02 Ultra Workstation Mini PC, Intel Core Ultra 9 285HX (24C/24T, up to 5.5GHz), PCIe... | $1,659.00 | Buy on Amazon |
| 2 |
|
GMKtec EVO-X2 AI Mini PC Ryzen Al Max+ 395 Superchip 128GB LPDDR5X 2TB SSD | $3,649.99 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
Agentic retrieval adds a knowledge base and knowledge sources. It can plan multiple focused queries for a complex or conversational request, execute them concurrently, rerank the results, and merge grounding content. That is retrieval orchestration—not a guarantee that the final answer is correct. Your response layer still needs to handle evidence and uncertainty responsibly.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsShould you use classic RAG or agentic retrieval?
Choose based on how much retrieval orchestration your application needs and which capabilities are mature enough for your deployment. These are documented architectural distinctions; benchmark your own data and queries before selecting a production design.
#1 Best Overall
- High-Performance AI Processor：The MS-02 Ultra features an Intel Core Ultra 9 285HX (24C/24T, up to 5.5 GHz, 13 TOPS NPU), delivering fast and efficient performance for AI inference, algorithm development, and media workloads. A PCIe x16 expansion slot supports desktop-class GPU upgrades for advanced model training and accelerated computing tasks. It's ideal for creators, engineers, and teams handling intensive parallel workloads.
- 4 × M.2 PCIe 4.0 + 4 × DDR5 SODIMM slots:Four DDR5 SODIMM slots support up to 256 GB of memory, while ECC helps maintain data integrity in mission-critical environments. Four PCIe 4.0 M.2 slots support up to 24 TB of storage, supporting RAID 0/1/5/10, combining high-speed performance with data protection. It allows for the creation of independent scratch disks, media libraries, and project drives, providing high-throughput for production workflows.
- PCIe & USB 4.0 v2: Up to three PCIe slots can be equipped, including a dual-slot x16 GPU. The main slot supports PCIe 5.0, meeting the needs of high-bandwidth creative and computing workloads. USB 4.0 v2 (80Gbps) supports high-bandwidth external storage and displays.
- Ultra-fast Networking: Wi-Fi 7 further enhances wireless performance with next-generation speeds and low-latency stability. Intelligent bandwidth switching optimizes throughput in different network environments, ensuring optimal performance for enterprise or local networks. Dual 25GbE ports (providing up to approximately 3.125 GB/s bandwidth, about 25 times faster than traditional 1GbE), enabling seamless large-scale file transfers and parallel computing. 10GbE and 2.5GbE ports, with support for Intel vPro technology, ensure enterprise-grade remote management and deployment flexibility.
- Server-grade thermal architecture: Utilizing a dedicated CPU/GPU airflow design, equipped with a 6-pipe dual-fan cooler, it maintains stable performance even under sustained loads, delivering up to 140W Turbo power while maintaining a 100W TDP, and operating with noise levels as low as 36 dB. An integrated 350W power supply ensures stable and reliable output for demanding computing tasks and fully loaded extended configurations.
| Decision | Classic RAG | Agentic retrieval |
|---|---|---|
| Query behavior | Runs a single search query; the application manages model handoff and orchestration. | A knowledge base can plan focused subqueries and run them in parallel. |
| Components | Fewer components; works with application-owned orchestration. | Adds knowledge sources, a knowledge base, and retrieval orchestration. |
| Latency and cost factors | No LLM query-planning call; semantic ranking can still add usage and cost. | Planning and optional answer synthesis use model tokens; retrieval and semantic ranking also have Azure AI Search billing. |
| Good fit | Simple queries, GA-only requirements, existing orchestration, or a preference for a direct search path. | Conversational or complex queries, agent and chat applications, or structured grounding results. |
| Feature maturity | An established pattern. | Base capabilities and advanced features have different maturity levels; the API version matters. |
Microsoft recommends starting with agentic retrieval for a new conversational RAG build, but that is a starting point, not a mandate. A classic pipeline may be the better choice if you need to avoid preview features or already have a reliable orchestration layer. The agentic retrieval overview explains its capabilities and version distinctions.
How should you prepare and index content?
Design the index around retrieval, not just storage. Keep readable text and useful metadata alongside vectors: vector conversion is one-way, so a vector field cannot be converted back into readable text for search results. Large documents usually need chunking so their passages fit embedding-model input limits; very small inputs may not need it. Azure AI Search’s integrated vectorization documentation covers indexer-driven ingestion and query-time vectorization.
- Choose sources and refresh behavior. Decide which content is in scope and how changes or deletions will be reflected in the index.
- Preserve traceable metadata. Include stable identifiers and source or document details so retrieved passages can be associated with their origin.
- Store chunk text and embeddings together. Each searchable chunk needs readable text for grounding and an associated embedding for vector retrieval.
- Configure ingestion components. For integrated indexer-driven vectorization, configure an indexer, index, and skillset. A skillset can split text using a suitable approach, such as Text Split, Azure Content Understanding, or document parsing modes, and then create embeddings.
- Select an embedding route. Documented options include an Azure OpenAI embedding skill, a custom skill, and other supported model integrations.
Chunk size and overlap should be tested against your documents and questions: the material here does not establish one universally best setting. The practical constraint is that large raw documents may exceed the embedding model’s input limit unless they are split.
How do you retrieve useful evidence?
For a classic RAG baseline
Start with hybrid retrieval: run keyword and vector search together. Keyword matching helps preserve exact terms such as product names or identifiers, while vector search can find semantically similar passages. Azure AI Search fuses the result lists, and semantic ranking can rerank a text or hybrid result set. Tune scoring profiles or vector parameters only after evaluating results against representative questions. Microsoft’s RAG overview describes hybrid retrieval, and its integrated vectorization overview covers vector search configuration.
For agentic retrieval
Create knowledge sources and a knowledge base, then choose an API version that matches the behavior you require. With advanced query planning, the system can break a complex request into focused subqueries, run them concurrently, apply semantic reranking, and merge evidence with source references. The production API path with minimal retrieval reasoning does not provide the same LLM-based planning behavior as the preview path.
How do you connect retrieved content to a response model?
- Retrieve grounding content. Call Azure AI Search using your chosen classic or agentic retrieval pattern.
- Pass evidence to the response layer. Give the retrieved passages and any source references to your application’s language model or agent.
- Generate and present the answer. Ask the model to answer from the supplied evidence and present source references in the way your product requires.
- Inspect retrieval behavior. Use returned execution activity and references to diagnose what evidence was retrieved when an answer is weak.
Microsoft’s end-to-end agentic retrieval tutorial uses Azure AI Search for the knowledge base and Microsoft Foundry for model deployments and an agent; a Foundry project connection calls the knowledge base MCP endpoint. The agentic retrieval quickstart demonstrates a conversational flow with a deployed embedding model and LLM.
Answer synthesis inside the agentic retrieval flow is a preview capability in the documented API guidance, not a requirement for RAG. You can instead keep answer generation in your application or agent. If you choose preview synthesis, account for preview terms and possible schema changes.
Recommended Free Tools
Rank #2
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Which API version should you deploy?
Microsoft’s documentation checked on October 7, 2026 distinguishes these REST API paths:
2026-04-01: Production use of generally available knowledge source types with minimal, extractive retrieval.2026-08-01-preview: Preview knowledge source types and advanced features, including LLM-based query planning, answer synthesis, non-minimal retrieval reasoning effort, and multi-turn messages.
Portal experiences expose preview-only agentic retrieval capabilities. Objects created there can use preview schemas and may need migration to work with the GA REST API. Verify the current API guidance before implementation, and pin the version in your deployment rather than assuming all agentic features share the same maturity.
How should you secure the integration?
Use Microsoft Entra ID role-based access and managed identities where they fit your deployment. The quickstart configures a system-assigned managed identity for the search service, assigns relevant Search roles, and grants Cognitive Services User to the search identity for the Foundry resource. It also documents key-based authentication as an alternative when role assignment is infeasible. The quickstart’s Basic-tier managed identity prerequisite applies to that tutorial’s configuration, not universally to every Azure AI Search design. Follow the quickstart’s access setup for its specific walkthrough.
For a network-isolated Foundry resource with public network access disabled, the tutorial says to use the newer Foundry portal, SDK, or CLI to build agents; the classic portal does not support that scenario. See the tutorial’s network-access guidance when applying that setup.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhat limits and costs should you plan for?
Capacity depends on tier, region, request patterns, API version, semantic ranking, and model use. The figures below are examples from Microsoft’s service-limits documentation checked in 2026, not a complete tier comparison or a latency promise.
- Knowledge sources: The documented limits include 3 per service on Free and 50 on S1. Other tiers and legacy service cases have different limits. Check the current service limits by tier for your service.
- Retrieve request runtime: The documented minimum is 10 seconds, the default is 90 seconds, and the maximum is 600 seconds. These are Azure AI Search retrieve-request limits, not end-to-end chatbot latency targets.
- Serverless Developer preview: Microsoft says billing for this preview tier began September 13, 2026. It has no SLA, is region-limited, and its limits, features, and pricing may change before general availability. The pricing model does not support migration to or from other tiers; check the service limits and tier guidance before choosing it.
- Agentic retrieval billing: There is a free monthly token allowance; Standard billing is pay-as-you-go beyond that allowance and requires Basic tier or higher. From REST API
2026-04-01, billing consent for agentic retrieval is separate from semantic ranker. See Microsoft’s instructions to enable or disable agentic retrieval billing.
Budget for the services your design actually invokes: Azure AI Search retrieval and semantic-ranking charges, plus model token use for query planning and any answer synthesis. A fixed dollar estimate would require your region, tier, usage, and model configuration; check current pricing and limits for those details.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
