Skip to main content
Collabase integrates AI throughout the platform under the Collabase Brain umbrella. All AI features share the same provider configuration set in Settings → AI — configure your provider once and every feature benefits.

AI in the editor

Generate, summarise, translate, and improve text anywhere you write in Collabase.

Semantic search

Ask questions in plain language and get answers grounded in your actual content.

AI in Automation

Use the AI/LLM node to generate text, summarise inputs, and classify data inside pipelines.

AI logs

See every AI request made across the platform — tokens used, latency, and errors.

​
Enabling Collabase Brain

All AI features are controlled by a single master switch: Collabase Brain.
  1. Go to Settings → AI.
  2. Toggle Collabase Brain to enabled.
  3. Select and configure your AI provider (see AI Configuration).
When Brain is disabled, the /ai command disappears from the editor, semantic search falls back to keyword search, and AI automation nodes return an error at runtime.

​
AI in the editor

The AI command is available anywhere you write content — pages in the Docs app, test case descriptions, automation notes, and Intranet posts.

​
Using the AI command

Type /ai on any line to open the AI action menu. The menu closes and the cursor returns to the text when the action completes.
CommandWhat it does
GenerateWrite new content from a prompt you provide
SummariseCompress the selected text or current page into a concise summary
Improve writingRefine grammar, clarity, and overall tone
Fix spelling & grammarCorrect errors only — tone and structure are preserved
Make shorterTrim the selection to its essential points
Make longerExpand the selection with more detail
Continue writingExtend the text from the current cursor position
TranslateTranslate selected text — you choose the target language
Selection behaviour: If text is selected, the command operates on the selection. If nothing is selected, it operates on the paragraph at the cursor.

​
AI writing in custom fields

In the Registry app, text-type custom fields also support the AI command. Use it to auto-fill descriptions, summaries, or structured text based on other field values.

​
Semantic search (RAG)

When RAG / Semantic Search is enabled in Settings → AI, Collabase indexes all pages as vector embeddings and uses them to answer natural-language questions in Brain chat.

​
How it works

  1. When a page is saved, its content is split into 512-token chunks.
  2. Each chunk is converted to a vector embedding by your configured embedding model.
  3. When you ask Brain a question, the engine retrieves the most relevant chunks and passes them to the language model as context.
  4. The model generates a response grounded in your actual content, with source links.

​
Embedding models

Configure the embedding source in Settings → AI → RAG / Semantic Search:
OptionDescription
Same as AI providerReuses the generation model for embeddings. Works well for most setups.
Custom providerUse a dedicated embedding model. Recommended for best accuracy.
Recommended dedicated models:
  • Ollama: nomic-embed-text, mxbai-embed-large
  • OpenAI: text-embedding-3-small

​
Indexing delay

Pages become searchable after indexing completes — typically a few seconds after saving. Newly installed instances may take a few minutes to index all existing content.

​
AI in Automation

The AI / LLM connector lets you run LLM actions inside any automation pipeline. This is distinct from the in-editor Brain — it uses a separately configured connection and can target a different model if needed.

​
Setting up the AI connector

  1. In Automation → Connections, click New Connection and select AI / LLM.
  2. Enter the baseUrl and apiKey for your provider.
  3. Save the connection.

​
Available actions

ActionUse caseKey inputs
Generate TextDraft content, classify input, transform datamodel, systemPrompt, userPrompt, temperature
Summarize TextCondense long inputs before passing them to other nodestext, model, maxSentences

​
Chaining AI with other connectors

AI nodes output content (Generate Text) or summary (Summarize Text) — reference them with {{previous.content}} in subsequent nodes. Example — auto-create a Jira issue from a failed test with an AI-generated description: 
[Trigger: Test Case Failed]
    ↓
[AI: Generate Text]
    systemPrompt: "Write a concise Jira bug report in English."
    userPrompt: "Test case '{{testCaseTitle}}' failed in run {{testRunId}}."
    ↓
[Jira: Create Issue]
    projectKey: QA
    summary: "Automated bug: {{testCaseTitle}}"
    description: {{previous.content}}
    issueType: Bug
    priority: High

​
AI logs

Settings → AI → Logs shows a time-ordered history of every AI request sent from your Collabase instance across all features (editor, RAG, automation nodes). Each log entry shows:
FieldDescription
TimestampWhen the request was made
FeatureWhich feature triggered the request — editor, RAG, or automation
ModelThe model that processed the request
Tokens usedTotal token count (prompt + completion)
LatencyTime from request sent to response received
StatusSuccess or error with the error message

​
Using logs to debug

If an AI action in an automation fails, find the corresponding log entry and check the error message. Common issues:
ErrorFix
Connection refusedThe Ollama server is not running or is unreachable from Collabase
model not foundThe specified model has not been pulled in Ollama. Run ollama pull <model-name> on the server.
invalid_api_keyThe API key for your cloud provider is incorrect or expired
context length exceededThe input is too long for the model. Reduce input size or switch to a model with a larger context window.

​
Privacy and data residency

All AI processing stays entirely on your own server. No content is sent to any external service. This is the recommended option for organisations with strict data residency requirements (e.g. Swiss financial or healthcare data).
Content sent to AI actions is transmitted to the respective cloud provider’s API. Review each provider’s data processing agreement to understand how they handle your data. API keys are encrypted at rest in Collabase and are never logged.