José Henrique Oliveira de Carvalho built a retrieval-augmented generation (RAG) assistant to answer questions about his background, experience, projects and technical choices. Its embeddings are generated locally, and PostgreSQL with pgvector stores and retrieves the content; response generation goes through Groq. That distinction matters: this is a partly local pipeline, not an entirely offline one.
The project is a concrete example, not a benchmark or a universal recipe. Carvalho’s main point is that a useful answer depends on the complete path from source documents through retrieval—not just on which language model writes the response.
As an Amazon Associate I earn from qualifying purchases.
How the portfolio assistant’s RAG pipeline works
The implementation described by Carvalho connects versioned Markdown files to an LLM through a retrieval pipeline. In order, it parses and chunks the documents, enriches them with likely questions, generates embeddings locally, stores the content and vectors in PostgreSQL, retrieves similar chunks for each question, filters them by distance, and sends accepted context to an LLM through Groq.
Crashes, 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 minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThe stack named in the project article includes Bun, Elysia, TypeScript, Drizzle ORM, LangChain’s text splitter, Transformers.js via @huggingface/transformers, the Xenova/multilingual-e5-small embedding model, PostgreSQL with pgvector, and openai/gpt-oss-120b for response generation through Groq. These are the author’s implementation choices, not requirements for RAG generally. Carvalho’s project article provides the implementation context.
#1 Best Overall
- Includes Raspberry Pi 5 with 2.4Ghz 64-bit quad-core CPU (8GB RAM)
- Includes 128GB Micro SD Card pre-loaded with 64-bit Raspberry Pi OS, USB MicroSD Card Reader
- CanaKit Turbine Black Case for the Raspberry Pi 5
- CanaKit Low Noise Bearing System Fan
- Mega Heat Sink - Black Anodized
What “local” means here
The author reports generating embeddings on the local CPU rather than calling an external embedding API. The response-generation step, however, uses Groq. So “local” describes the embedding stage, not every component or the entire flow. The project article does not establish that the full assistant can run offline.
How should documents be chunked?
The assistant’s knowledge lives in versioned Markdown files covering the author’s profile, experience and projects. The project uses LangChain’s RecursiveCharacterTextSplitter in Markdown mode, with a reported chunkSize of 800 and chunkOverlap of 50.
Rank #2
- Includes Raspberry Pi 4 4GB Model B with 1.5GHz 64-bit quad-core CPU (4GB RAM)
- Includes Pre-Loaded 32GB EVO+ Micro SD Card (Class 10), USB MicroSD Card Reader
- CanaKit Premium High-Gloss Raspberry Pi 4 Case with Integrated Fan Mount, CanaKit Low Noise Bearing System Fan
- CanaKit 3.5A USB-C Raspberry Pi 4 Power Supply (US Plug) with Noise Filter, Set of Heat Sinks, Display Cable - 6 foot (Supports up to 4K60p)
- CanaKit USB-C PiSwitch (On/Off Power Switch for Raspberry Pi 4)
Those figures describe this implementation; they are not demonstrated as optimal settings for other content. Chunking determines what a retrieval result can contain: a chunk that is too narrow may omit needed context, while a broad one may include material that is less relevant to a particular question. The project article does not report a comparative evaluation of alternative chunk sizes.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesWhy add likely questions to the source text?
Before embedding, Carvalho adds probable questions visitors might ask. The idea is to give a document wording that resembles the queries it should answer, potentially helping similarity search find it. This is an enrichment strategy in the project, not evidence that it will improve retrieval for every dataset or model.
Rank #3
- Not including the Raspberry Pi 5 (8GB), the Crowpi advanced version comes with the Raspberry Pi 5
- ELECROW Black Case for the Raspberry Pi 5, CrowPi is equipped with a 9-inch HD touchscreen along with a camera; All the regular components used in DIY electronics are packed into the CrowPi development board, such as LCD, LED matrix, buzzer, light sensor, PIR sensor, ultrasonic sensor, IR sensor, etc
- Raspberry Pi Sensors: The Crowpi raspberry pi 5 programming kit is jam-packed with lots of buttons such as 19 different sensors in a tidy easy to use package; You don't have to wait and wire things
- Build Quality: Solid ABS shell and well made components in one place make it strong and convenient to travel
- Programming Lessons: This raspberry pi 5 learning kit ships with step by step instructions and provides 21 lessons to take you through identifying components reading code and running it in the terminal
How are embeddings generated and matched?
For embeddings, the author reports using Xenova/multilingual-e5-small through Transformers.js on CPU, with mean pooling and normalization. The resulting vectors have 384 dimensions in this implementation. Stored content receives the passage: prefix and incoming questions the query: prefix. These details belong to this model’s setup; the article does not establish that the same prefixes or processing choices apply to other embedding models.
At query time, the application compares the question vector with stored vectors using pgvector’s <=> cosine-distance operator and asks for the five closest results, ordered by ascending distance. The pgvector documentation describes cosine distance and the operator. pgvector’s documentation also explains its available nearest-neighbor search approaches.
Rank #4
- Fully assembled for plug-and-play operation
- Includes Raspberry Pi 5 with 8GB RAM
- 256 GB PCIe Pi NVMe SSD (Pre-loaded with Pi 64-Bit OS)
- M.2 HAT+
- CanaKit Turbine Black Case for the Pi 5
When is a retrieved chunk relevant enough to use?
Carvalho’s application accepts retrieved results only when their cosine distance is below 0.35. That cutoff is a project-specific filter, not a generally validated relevance boundary. A distance value depends on the embedding model, data, and retrieval setup; the project article does not provide an independent quality evaluation for this threshold.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Filtering can keep weak matches out of the prompt, but it cannot guarantee that an LLM will never invent information. The author’s approach is to avoid injecting arbitrary context when no result passes the cutoff and to give the model a basic instruction not to make up information. That is a sensible fallback for this assistant, but it does not establish a guarantee of factual answers.
Best Value
- 【What you Get】You will get 1*Pi 5 8GB Single Board,1*RasTech Case,1*Active Cooler,1*Screwdriver,1*Installation instructions,12-month free warranty, lifetime service, 24-hour prompt and friendly response.
- 【More Connectors】There are two USB 3.0 ports(5Gbps simultaneously) and two USB 2.0 ports, which triple total bandwidth ,support any combination of up to two cameras or displays. Peak SD card performance is doubled through support for the SDR104 high-speed mode. It provides a smooth desktop experience for you. Offer Gigabit Ethernet and a PCIe interface, along with dual-band Wi-Fi and Bluetooth 5.0/BLE wireless capability. The RasTech Pi 5 Kit use the new 27W 5.1V 5A USB-C power connector.
- 【 Support Dual 4Kp60 Display 】Each of the two microHDMI sockets can control a 4K display at 60 Hertz, now support HDR, offering super HD video for media streaming projects. RPi 5 is the first RPi model that comes with a PCI Express port (PCIe 2.0 x1 with 500 MB/s) to attach SSDs (requires separate M.2 HAT).
- 【 Excellent Chips And Applications】Pi 5 is a full-size Pi computer using silicon built in-house at Pi. The RP1 “southbridge” provides the bulk of the I/O capabilities for Pi 5. Pi 5 is more friendly and convenient in the development of Internet of Things, Web development, machine identification, automatic control and other electronic equipment applications and network.
- 【 Faster CPU, Better GPU 】 Pi 5 features a Broadcom BCM2712 64-bit quad-core Arm Cortex-A76 processor running at 2.4GHz, it delivers a 2–3× increase in CPU performance relative to RaspberryPi 4. The 800MHz VideoCore VII GPU is compatible to OpenGL ES 3.1 and Vulkan 1.2, substantial uplift in graphics performance. Pi 5 Offers lightning-fast CPU speed, a PCI Express interface, a Real Time Clock (RTC) and a power button and runs significantly cooler than Pi 4.
Do you need a dedicated vector database?
For this personal portfolio, the author says PostgreSQL with pgvector was sufficient. It keeps ordinary application data and vector search within PostgreSQL, which can be a practical fit when a project already uses that database. The project does not claim that PostgreSQL is the right choice for every workload.
pgvector performs exact nearest-neighbor search by default. Its documentation also describes HNSW and IVFFlat indexes for approximate search; approximate methods can improve search speed while trading off recall. Whether that tradeoff or a dedicated vector database makes sense depends on workload scale and complexity. Neither source establishes a universal size threshold or a benchmark winner.
What this example does—and does not—show
The pipeline illustrates how document preparation, chunk boundaries, question wording, embedding choices, retrieval, and relevance filtering all shape what reaches the model. As Carvalho puts it, “The most important lesson for me was that the LLM is not the whole system.” That is the author’s takeaway from his project, not the result of a comparative benchmark.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The implementation is presented as adequate for a personal portfolio. It does not supply independent hardware tests, quality measurements, cost comparisons, or scale benchmarks. Treat its settings as a worked example to understand and adapt, rather than evidence that a particular combination is the best architecture for a different application.
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.
