> For clean Markdown of any page, append `.md` to the page URL. > For a complete documentation index, see https://docs.sarvam.ai/llms.txt. > For full documentation content in one file, see https://docs.sarvam.ai/llms-full.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server. # Create a Config > Build reusable Extract or Digitise Configs from templates, duplicates, or scratch — name, prompt, language, and output schema. A **Config** is a saved recipe: a name, a description of what you want to extract or digitise, a language or output format, and any sharing settings. Build once, re-run against every document of that type. There are three ways to create a Config. Pick the one that fits your situation. Templates are the fastest path to a working project. Sarvam Pages ships with **12 ready-made templates** (9 Extract + 3 Digitise) covering the document types users ask for most, including KYC, Credit Card Statement, Invoice Screener, Resume Screener, GST forms, Land Deeds Digitisation, and Digitise Handwritten Document. **From Home:** Scroll past the Extract and Digitise cards to **Or start with a predefined config**. Pick a template card from the shelf (filter by **All / Extract / Digitise**), preview it, and start a project in one click. Home page showing Or start with a predefined config with template cards for Insurance Claim Form Extraction, Credit Card Statement Extraction, and KYC Document Extraction. **From Workspace:** Open **Workspace → Configs → Templates** and click a template to preview it. Each preview shows a sample document with the fields it extracts (and a confidence score per field), so you can see exactly what you'll get. Click **Use this template** to start. Configs page Templates tab showing the 12 shipped templates with Operation, Usage, and Last used columns, and All / Extract / Digitise filter pills. Background Verification template preview with a sample payslip on the left and extracted fields (document type, employee name, month and year, net salary payable, company name and address) with confidence scores on the right, plus a Use this template button. The template opens a New Project dialog with the **prompt prefilled** for that document type. Upload your document (PDF, JPEG, or PNG, up to 50 MB) and click **Extract**. New Project dialog with an upload dropzone, a Project name field, and a Describe what to extract field prefilled from the template with detailed extraction instructions. You can review and tweak the template's schema in the **Define your output format** dialog before running, exactly as you would for a custom Config. Define your output format dialog showing the template's AI-drafted schema with nested objects (documents, payslip_details) and typed fields, plus Regenerate and Clear all controls. To make a template your own, open its `⋯` menu and click **Duplicate**. The copy lands under **Your configs**, where you can rename it and edit the prompt, language, or schema. Already have a Config that's close to what you need? Duplicate it instead of starting over. Sidebar → **Workspace → Configs** → **Your configs** tab. Open its `⋯` menu and click **Duplicate**. A copy lands under **Your configs**. Give the copy a clear new name, then tweak the prompt, language, or schema for the new use case. Duplicating is the safe way to experiment: the original Config keeps running unchanged for every project that already uses it, while you iterate on the copy. If no template fits, or you want a purpose-built recipe for your team, build one from scratch. On the **Configs** page, click **Create config** (or **+ New config** in the top-right). Configs page empty state reading No custom configs yet, with Browse templates, Create config, and + New config buttons. On the **Create a new workflow** screen, fill in: A clear identifier like *Invoice Extraction* or *Land Deed Digitiser v2*. Shows up in the Configs list and in every Project that uses this recipe. Pick **Extract** for structured-field pulls or **Digitise** for full-document text conversion. Describe what you want in plain English. Copy this into **Prompt**, then edit for your use case: ```text Extract the name, bill number and the amount from the invoices. ``` Click **View / Edit** to open the schema builder. Create a new workflow screen with Name (Invoice Extraction), Pipeline type (Extract), Prompt, and an Output schema showing 1 field defined with a View / Edit button. In the **Define your output format** dialog, the workbench AI-drafts a schema from your prompt. Each field has a type (`string`, `number`, `array of objects`, and more), and you can add or remove fields, **Regenerate** from the prompt, or switch to the **JSON** tab. Edit until it matches what your downstream system expects. Define your output format dialog with the prompt on the left and an AI-drafted schema on the right: invoices (array of objects) containing name (string), bill_number (string), and amount (number). Click **Save**. The Config appears under **Your configs**, and also surfaces on the **Home** dashboard under **Start with your custom configs** for one-click reuse. ## Where Configs live Configs live under **Workspace → Configs** in the sidebar. The page has two tabs: * **Your configs.** Anything your workspace has built or duplicated. * **Templates.** The 12 shipped templates. Filter by **All / Extract / Digitise** using the pills. Your configs tab showing the workspace's saved Configs with Operation, Usage, and Last used columns. Configs page with Your configs and Templates tabs, All / Extract / Digitise filter pills, and a Usage column. Each row shows a `Usage` count (how many projects have used the Config) and a `Last used` timestamp, a quick signal for which recipes your team relies on. Next: run your first project with [Extract structured fields](/pages/how-to/extract-structured-fields) or [Digitise a document](/pages/how-to/digitise-a-document).