MaruOps · Infrastructure and AI notes
Linux, Kubernetes
and local AI.
Linux, networking, Kubernetes and local LLMs.
Configuration choices and troubleshooting, explained step by step.
Series guides
Kubernetes Lab
Prepare three VMs, build a cluster, then work through deployment, networking and recovery.
Local LLM
Run a first model, then explore APIs, document retrieval, tools and operations.
Git and GitOps
Follow repositories and commits through CI, deployment and GitOps recovery.
Latest articles
Browse the topics to find the guide you need.
-
[Local LLM 12] Set Up Web Chat with Open WebUI and Ollama
Connect Open WebUI to the correct Ollama endpoint and verify model discovery, container networking and persistent data.
-
[Local LLM 13] Call the Ollama API from Python
Send a request to the local Ollama API from Python, handle the response, and distinguish connection errors from model errors.
-
[Local LLM 14] Write Prompts with Clear Source Text and Constraints
Separate instructions, reference text and output constraints so that a model request can be checked and repeated.
-
[Local LLM 15] Validate Structured Responses with JSON Schema
Define the expected JSON structure, validate model responses and decide how the application handles invalid output.
-
[Local LLM 16] Search Documents with Embeddings and Cosine Similarity
Turn document passages and a query into embeddings, rank them by cosine similarity, and inspect retrieval quality.
-
[Local LLM 17] Build RAG Answers from Retrieved Documents
Retrieve relevant document passages, include them in the prompt, and check whether the answer is supported by those passages.
-
[Local LLM 18] Build an Evaluation Sheet for Models and RAG
Use a fixed question set and scoring criteria to compare model answers and retrieval results before changing the application.
-
[Local LLM 19] Understand Hermes Models and Chat Templates
Match a Hermes model to the runtime and conversation template, then inspect how messages and tool calls are represented.
-
[Local LLM 20] Connect Hermes Agent and Review Tools and Memory
Connect Hermes Agent to the intended model service and inspect the permissions and persistence of tools and memory.
About these notes
Configuration
and troubleshooting.
From first configuration to operational issues.
Find what to inspect, why it matters, and how to proceed.