#AI
Posts tagged #AI · 6 posts
- Wiring Domino IQ RAG to a Remote Model: certstore Trust, the API Key, and the HTTPS Endpoint
Domino 14.5.1 FP1 added a Remote mode to RAG — the site's FP1 roundup already covered what changed. This piece doesn't repeat that; it fills in the wiring: how the Remote-mode configuration document is filled out, the HTTPS-only endpoint and API Key, and the step that trips people up — the remote AI server's trusted roots go into certstore.nsf, not the JVM truststore that the DRAPI path consumes.
2026.09.23 - Domino IQ Requests Timing Out at 5 Minutes? No Setting to Raise — Switch to Streaming
You build a long-document summary on Domino IQ; in testing the prompts are short and it's instant. In production someone feeds in a huge thread on a modest GPU, and you hit a 'network processing did not complete in time' timeout. You go hunting for a timeout setting to raise — there isn't one. Domino IQ requests time out at about 5 minutes, by design, and it can't be changed. This piece walks a real scenario: why it times out, how to switch from Completion to CompletionStream to get past it, and one important caveat — streaming keeps it from timing out, but doesn't make it faster.
2026.09.05 - Field Notes from OpenNTF's Domino IQ RAG Webinar: the 30-Char Gotcha, Vectorization in 3 Steps, and the FP1 Readers Fix
OpenNTF hosted HCL's Brian Arnold for a deep session on Domino IQ RAG. We've already covered how the pipeline works under the hood; this is the field-report layer — the 30-character limit for RAG-enabling short fields and the concat-field way around it, vectorization in three steps, the three knobs to tune, the fact that Readers/Authors-field security only really started working in Domino 2026 FP1, and the citations, side panel, and MCP still under development.
2026.08.04 - NotesLLMRequest: Call an LLM from LotusScript in 4 Lines
Domino 14.5 introduced NotesLLMRequest and NotesLLMResponse — two LotusScript classes that expose Domino IQ's local LLM as a synchronous API for app developers. This article walks the full surface: the session.CreateLlmRequest() factory method, the Completion three-parameter signature, the CompletionStream streaming variant, the IsCommandAvailable / GetAvailableCommands defensive helpers, and the Content / FinishReason / Role fields on NotesLLMResponse. Why the API takes a commandName rather than a raw prompt — the Command document abstraction lets admins and devs each control their own layer. Includes two practical examples (auto-reply agent + conditional summarization) and the Java mapping to LLMReq / LLMRes.
2026.05.23 - Domino IQ RAG: A Built-In Pipeline That Wires Your NSFs Straight Into a Local LLM
Domino 14.5.1 adds RAG (Retrieval-Augmented Generation) support to Domino IQ, running the LLM, embedding model, and vector database all on the Domino server itself — local execution, with NSF ACL and Readers fields enforced natively. This guide walks through prerequisites, the two-phase dominoiq.nsf configuration, updall vectorization, calling LLMReq from LotusScript, and why this is a different species from the OpenAI + Pinecone pipeline.
2026.05.05 - Domino IQ: What It Means to Run an LLM Inside the Domino Server
Domino 14.5 introduces Domino IQ — an AI inference engine baked into the Domino server backend, callable from LotusScript via NotesLLMRequest / NotesLLMResponse without ever leaving the box. This guide covers the architecture, hardware requirements, install flow, the two-phase dominoiq.nsf configuration, the Command and System Prompt document model, and why this trade-off works for existing Domino shops where bolting on OpenAI doesn't.
2026.05.05