Upload a document and start processing it.

Ingest Document

Details

Error Code

Message

ErrorDetail

External Id

Path where the file is stored (null for embedding-based documents)

File Path

Task identifier for tracking progress (null for synchronous operations)

Task Identifier

UploadResponseData

Body_ingest_document_api_v1_documents_ingest_post

ApiKeyAuth

UploadResponse

Fieldwise.ai

Welcome to the Fieldwise API documentation. Our API offers a developer-friendly interface to enhance your document intelligence by providing real-time data retrieval and AI-driven insights.

Introduction

Welcome to the Fieldwise API! This guide will help you get started with authentication and making your first API request.

Quickstart

Ingest Document From Url

Ingest a document directly from a pre-computed embedding vector.

This endpoint allows you to create documents directly from embedding vectors,
bypassing the file upload and text extraction process. This is useful when:

- You already have pre-computed embeddings
- You want to ingest structured data or custom content
- You need to create documents from API responses or generated content
- You want to maintain embedding consistency across systems

The embedding vector must be exactly 1536 dimensions (OpenAI's text-embedding-ada-002 format).

Performance Benefits:
- No file processing overhead
- No text extraction or embedding generation
- Immediate availability for search
- Synchronous operation (completed immediately)

Retrieval:
- Documents can be retrieved using the returned task_identifier via GET /task/{task_identifier}
- External ID retrieval also works via GET /external/{external_id} if provided

Best Practices:
- Ensure embedding vector quality matches your use case
- Include meaningful content text for human readability
- Use consistent metadata structure for filtering
- Consider embedding source and generation method in metadata

Ingest Document From Embedding

 Answer a question based on the documents matching the provided metadata filter. <br /> The `metadata_filter` parameter supports MongoDB-like query syntax with operators: <br />
- Comparison: `$eq, $ne, $gt, $lt, $gte, $lte, $in, $nin`
- Array: `$contains, $containsAny, $notContains, $size`
- Logical: `$and, $or, $not`
- Type: `$type`
- Regex: `$regex, $iregex`
<br />
#### Example Metadata Filter
```json { "$and": [ { "category": "finance" }, { "tags": { "$contains": ["important"] } }, { "$or": [ { "priority": { "$gt": 2 } }, { "status": "active" } ] } ] } ```


Get Document Answer

List Documents

 Get the most relevant pages from documents matching the provided metadata filter. <br />
The `metadata_filter` parameter supports the same MongoDB-like query syntax as the [/answer](api-reference/documents/get-document-answer) endpoint. See [/answer](api-reference/documents/get-document-answer) endpoint documentation for detailed examples of metadata filtering.


Search document pages (deprecated)

Search through document pages using semantic similarity.

This endpoint performs a semantic search at the page level, which is best for:
- Finding broad context about a topic
- Getting an overview of document sections
- When you need more context around the matches

Performance Considerations:
- Fastest search performance of all levels
- Most efficient memory usage
- Best for initial broad searches
- Recommended for large document collections

Best Practices:
- Use for initial exploratory searches
- Combine with metadata filters for better performance
- Start with this before using more granular searches
- Ideal for queries needing broader context
- Use min_relevance to filter out low-quality matches

The search uses embeddings to find the most semantically similar pages to your query,
regardless of exact keyword matches.

You can provide either:
- A text query (which will be converted to embeddings)
- A pre-computed embedding vector (1536 dimensions)

Filter options:
- min_relevance: Set a threshold (0-1) to only return results above a certain relevance score

Search Document Pages

Search through document paragraphs using semantic similarity.

This endpoint performs a semantic search at the paragraph level, which is best for:
- Finding specific topics or concepts
- Getting more focused results than page-level search
- When you need a moderate amount of context

Performance Considerations:
- Moderate search performance
- Balanced memory usage
- Good for medium-sized document collections
- More vectors to search than pages

Best Practices:
- Use when page-level search is too broad
- Ideal for finding specific explanations
- Good for topic-focused research
- Consider using metadata filters to improve performance
- Use min_relevance to filter out low-quality matches

Paragraphs provide a natural unit of text that maintains coherent thoughts and ideas,
making this search level ideal for finding self-contained explanations or descriptions.

You can provide either:
- A text query (which will be converted to embeddings)
- A pre-computed embedding vector (1536 dimensions)

Filter options:
- min_relevance: Set a threshold (0-1) to only return results above a certain relevance score

Search Document Paragraphs

Search through document sentences using semantic similarity.

This endpoint performs a semantic search at the sentence level, which is best for:
- Finding very specific information or facts
- Getting precise answers to questions
- When you need the most granular search results

Performance Considerations:
- Slower search performance due to more vectors
- Highest memory usage
- Best for smaller, focused searches
- Consider impact on rate limits

Best Practices:
- Use only when specific facts are needed
- Keep document set focused using metadata filters
- Consider caching frequent queries
- Use smaller top_k values (5-10 recommended)
- Use min_relevance to filter out low-quality matches

Sentence-level search provides the highest precision but with minimal context.
Each result is a single, focused statement or fact from the document.

You can provide either:
- A text query (which will be converted to embeddings)
- A pre-computed embedding vector (1536 dimensions)

Filter options:
- min_relevance: Set a threshold (0-1) to only return results above a certain relevance score

Search Document Sentences

Retrieve the latest document by its external ID.

Get Document By External Id

Get Document By Id

Delete a single document by its internal ID.

Delete Document By Id

Retrieve a document by its task identifier.

Get Document By Task Id

Preview how many documents would be deleted by the metadata filter.

Preview Delete By Metadata

Delete all documents matching the provided metadata filter.

Delete Documents By Metadata

Retrieve a specific page by its ID, including all its paragraphs and sentences.
    The response is hierarchically structured with:
    - Page content and metadata
    - List of paragraphs (sorted by paragraph_index)
    - List of sentences within each paragraph (sorted by sentence_index)

Get page content with paragraphs and sentences

Retrieve a specific paragraph by its ID, including all its sentences.
    The response includes:
    - Paragraph content and metadata
    - List of sentences (sorted by sentence_index)

Documents

Content

Extractions

Tasks

Ingest Document

Authorizations

Headers

Body

Response