Get Started – Backoffice

Set up your Clarifeye backoffice with APIs and pipelines for data and engineering teams

Who is this for?

Data teams and engineers integrating Clarifeye programmatically. The backoffice of the teammate is also known as the warehouse.

Quickstart steps

1. Connect your data

   Upload documents through the Backoffice UI or programmatically via the API/SDK.

2. Parse documents

   Create a parsing task to convert files into LLM‑readable text. Review parsed documents, tables, and figures; inspect blocks and applied enrichments.

3. Chunk content

   Split documents into retrieval‑ready chunks using Visual Extractors or custom code. Configure boundaries (headings, tables, figures) and set size/overlap as needed.

4. Define tags and objects

   Use Visual Extractors or the API to add structure:

   • Document tags: metadata to classify and navigate documents (e.g., type, year).
   • Chunk tags: labels for sections within a document (e.g., financials, legal, competition).
   • Objects: structured entities (pydantic classes) capturing precise fields (e.g., exact numerical values from a 10‑K).
5. Create tools

   Build tools that leverage your tags and objects: retrieve by tag, search specific sections, or fetch domain objects by field/value to ground LLM answers.

6. Create your agent

   Select expert artifacts (playbooks, briefs, query sets) and attach the tools to assemble an expert‑level agent. Test in the playground and iterate on prompts, tools, and routing.