Tenovos Content Intelligence Platform (3.0)

Copy

• Copy for LLM

  Copy page as Markdown for LLMs
• View as Markdown

  Open this page as Markdown
• Open in ChatGPT

  Get insights from ChatGPT
• Open in Claude

  Get insights from Claude
• Connect to Cursor

  Install MCP server on Cursor
• Connect to VS Code

  Install MCP server on VS Code

Executes a structured query against all documents in the organization, returning matching results with full control over which fields and traits are returned, how results are filtered, and how pages are navigated.

Query Language

Queries are expressed as a single string using a structured query language with up to four clauses:

fetch(<fields>)[.orderBy(<path>, asc|desc)][.limit(<n>[, <cursor>])]
from(<kind>[, <kind>...])
[match('<keywords>')]
[filter(<expression>)]

Only fetch and from are required. match and filter are optional and may appear in any order after from.

fetch() — Field and Trait Selection

Specifies which fields or traits to return. Multiple items are comma-separated.

Ordering — append .orderBy(<path>, asc|desc) to fetch():

fetch(/).orderBy(/name, asc)
fetch($summary).orderBy(/createdAt, desc)

Pagination — append .limit(<count>) or .limit(<count>, <cursor>) to fetch():

fetch(/).limit(25)
fetch(/).limit(25, nextPageToken)

The cursor value is the opaque pageCursor string returned in the previous response.

from() — Document Kind Selection

Specifies one or more document kinds to query. Kind names must be lowercase letters and underscores only (3–62 characters).

from(brand)
from(asset, brand, region)

Default platform kinds: media, brand, campaign, category, toolkit, section, block. Custom kinds defined in your organization are also supported.

match() — Full-Text Search

Performs a full-text keyword search across indexed document content. Keywords are provided as a single-quoted string. Results are ranked by relevance when orderBy is not specified.

match('acme global')
match('Q4 campaign hero')

filter() — Field-Level Conditions

Applies field-level conditions to narrow results. Expressions can be simple or compound.

Comparison operators:

Logical operators: && (AND), || (OR), (...) (grouping)

Literal value types:

filter(/status == 'active')
filter(/rating >= 4 && /region != null)
filter((/status == 'active' || /status == 'pending') && /owner != null)
filter(/tags ... 'featured')
filter(/$$id == 'assetAbc123')

Strings in filter conditions must be single-quoted. Supported escape sequences: \\, \', \n, \r, \t.

Property Paths

Property paths reference fields within a document and always begin with /. Nested fields are separated by /.

Range selectors on link properties — link fields (relationships to other documents) support a range selector [start..end] to limit which linked items are returned. Both bounds are zero-based; start is inclusive, end is exclusive.

Ranges can appear at multiple levels of a path:

fetch(/sections[0..3]/media[..10]) from(toolkit)

Traits

Traits define a standard contract for a set of fields a document kind can return. They are referenced in fetch() using the $ prefix.

Sub-traits apply a secondary trait to each item returned by the parent:

fetch($children$summary)           # fetch children, return $summary for each
fetch($children$summary(limit=5))  # limit to 5 children

Complete Examples

# Fetch all fields from a single kind
fetch(/) from(brand)

# Fetch specific fields
fetch(/name, /status, /region) from(asset)

# Fetch a trait
fetch($summary) from(brand)

# Fetch multiple traits and fields
fetch($summary, $tile, /metadata) from(brand)

# Fetch a trait at a nested path
fetch(/heroImage/$preview(w=800, h=600)) from(campaign)

# Full-text search
fetch($summary) from(asset) match('product launch 2024')

# Filter by a field value
fetch(/) from(asset) filter(/status == 'approved')

# Filter with AND
fetch($summary) from(asset) filter(/status == 'active' && /owner != null)

# Filter with OR and grouping
fetch($summary) from(asset)
filter((/status == 'active' || /status == 'pending') && /region != null)

# Contains / not-contains
fetch(/) from(asset) filter(/tags ... 'featured')
fetch(/) from(brand) filter(/name !.. 'draft')

# Query multiple kinds
fetch($summary) from(brand, region, asset) filter(/status == 'active')

# Order and paginate
fetch($summary).orderBy(/name, asc).limit(20) from(brand)

# Continue from a cursor
fetch($summary).orderBy(/name, asc).limit(20, nextPageCursorToken) from(brand)

# Fetch linked items with a range
fetch(/items[0..5]) from(category)

# Nested ranges
fetch(/sections[0..3]/media[..10]) from(toolkit)

# Limit children with trait attribute
fetch($children(limit=5)) from(category)

# Combined: search, filter, order, paginate
fetch($summary, $preview(w=400, h=300)).orderBy(/name, asc).limit(10, abc123)
from(asset)
match('campaign hero')
filter(/status == 'active' && /region != null)

Request Parameters

Response

Performs a keyword search within the content of a specific document, returning a relevance-scored series of matching positions. This is distinct from the query endpoint, which searches across documents — query-within searches inside a single document's indexed content.

This is most useful for long-form content where you want to locate specific passages, pages, or moments, for example:

• Finding the page in a PDF where a keyword appears
• Locating the timestamp in a video where a term is spoken
• Searching within a large document's indexed text body

Request Parameters

Response

Each point in the series contains:

Example

To find all moments in a video where "product launch" is mentioned, with a minimum relevance score of 0.5:

{
  "id": "doc01abc123",
  "keywords": "product launch",
  "minScore": 0.5
}

The response series would contain entries like:

{
  "series": [
    { "score": 0.92, "ref": "00:01:34" },
    { "score": 0.71, "ref": "00:08:12" }
  ]
}

For a PDF, ref values would be zero-based page indices (0, 4, 11, etc.).

Applies an ordered sequence of field-level patch operations to one or more existing documents. Patches are applied in the order supplied; if any patch in the sequence is invalid, the entire update for that document is rejected.

Request: updates - List of documents to update. Each entry includes: id - The document ID to target. Either id or externalId must be supplied. externalId - The external system ID to target (alternative to id, useful when the document was created with an externalId for cross-system referencing). patches - Ordered list of patch operations. Field paths use forward-slash notation matching the document's field structure (e.g. /name or /metadata/status). Each patch is one of: Add(path, value) — sets a field that does not yet exist. Remove(path) — removes the field at the given path. Replace(path, value) — overwrites an existing field's value.

Response: ids - IDs of the documents that were successfully updated.