AI
Every AI tool, with the scope it needs, its arguments and the API operation it calls.
6 tools
Search AI documents
Changes dataatlas_ai_documents_search
Semantic search + cross-encoder rerank over caller-supplied candidate documents. Send a query and a list of { id, text } candidates; returns the best matches ranked by relevance. CPU-only and tenant-scoped. Requires the intelligence:read scope.
- Scope
intelligence:read- Calls
- Example prompt
- Rank these candidate passages against my question.
Arguments
| Field | Type | Required | Description |
|---|---|---|---|
query | string | Yes | The search query text. |
candidates | array of object | No | Documents to rank ({ id, text }). Reranked against the query. |
limit | integer | No |
Recommend AI items
Changes dataatlas_ai_items_recommend
Item-item collaborative-filtering recommendations / next-best-action. Send interaction events and a target user id; returns recommended item ids. Requires the intelligence:read scope.
- Scope
intelligence:read- Calls
- Example prompt
- Given these interactions, what should this user look at next?
Arguments
| Field | Type | Required | Description |
|---|---|---|---|
forUserId | string | Yes | The user to recommend items for. |
interactions | array of object | Yes | User↔item interaction events used to build the model. |
limit | integer | No |
Get AI tenant settings
Read onlyatlas_ai_tenant_settings_get
Read the workspace AI settings. Returns { provider, model, customKeyConfigured }: the model provider and model the workspace uses, or null for the Atlas default, and whether the workspace has its own key configured. The key itself is never returned. Use this to explain which model answers requests in this workspace.
- Scope
ai:read- Calls
- Example prompt
- Which AI model does this workspace use, and do we have our own key configured?
Arguments
This tool takes no arguments.
Capture AI text
Changes dataatlas_ai_text_capture
Natural-language capture: extract dates, entities, and mentions from free text so a host can turn a sentence into structured task/meeting data. Requires the intelligence:write scope.
- Scope
intelligence:write- Calls
- Example prompt
- Turn this sentence into a task with a due date.
Arguments
| Field | Type | Required | Description |
|---|---|---|---|
text | string | Yes | Free text to extract entities/dates from. |
Generate AI text
Changes dataatlas_ai_text_generate
Grammar-constrained text generation. Provide a prompt (and optional system prompt / JSON schema) to get valid-by-construction output from the local model. Requires the intelligence:read scope.
- Scope
intelligence:read- Calls
- Example prompt
- Generate a JSON object matching this schema from the prompt.
Arguments
| Field | Type | Required | Description |
|---|---|---|---|
prompt | string | Yes | |
system | string | No | |
maxTokens | integer | No | |
jsonSchema | object | No | Optional JSON schema to constrain generation. |
Summarize AI text
Changes dataatlas_ai_text_summarize
Summarise a block of text (abstractive when the local LLM is enabled, otherwise extractive). Optionally cap the number of sentences. Requires the intelligence:read scope.
- Scope
intelligence:read- Calls
- Example prompt
- Summarise this meeting note in three sentences.
Arguments
| Field | Type | Required | Description |
|---|---|---|---|
text | string | Yes | |
maxSentences | integer | No |