Project Context & Knowledge Files

The Project Context section is where you define the non-code information that shapes how Origin understands your project. While the repository shows how the system is implemented, Context explains why it exists, what it should do, and which constraints it must follow.

In practice, this is where you attach product requirements, architectural documentation, schemas, and operational notes so agents reason with the same inputs your team uses.

What Context Solves

Code rarely captures full intent. For example:

• A database schema may enforce constraints, but it does not explain business rules.
• A feature implementation may exist, but acceptance criteria might live in a PRD.
• A rate limiter may be implemented, but the reason behind specific thresholds might only exist in documentation.

Without this information, task generation and reasoning can become misaligned. Context ensures agents evaluate changes against documented requirements rather than making assumptions from code structure alone.

The Document List

When you open Project Context, you'll see a searchable list of all context documents attached to the project.

A Search documents bar lets you filter by name. Two category tabs: PRD and Docs, let you narrow the list by document type, each showing a count of documents in that category. A status filter chip (Ready or Processing) further filters the list by indexing state.

Each document entry shows:

• A checkbox for bulk selection (use Select all to select the full list)
• File name
• File size (where available)
• Created date
• Category badge (PRD or Docs)
• Status badge (Ready or Processing)

Refresh and + Add Document controls sit at the top right.

Document Categories

Documents are organized into two categories:

PRD (Product Requirements Documents)

The PRD category captures product-level intent, what the system should do, independent of current implementation.

Typical PRDs include:

• Feature descriptions and expected behaviors
• Acceptance criteria
• Constraints and non-goals
• Business logic definitions

For example, if your PRD specifies that users can only access content belonging to their company, AI Task Discovery may surface tasks like validating company ID enforcement in API routes or hardening row-level security policies.

Docs (Documentation & Technical Guides)

The Docs category stores implementation-level documentation and internal guides:

• Architecture overviews
• Database schemas
• API usage documentation
• Workflow explanations
• Security guidelines

For example, uploading a database_schema.md helps agents understand table relationships before suggesting index optimizations. An authentication flow document ensures tasks involving middleware follow your existing model.

Opening a Document

Clicking a document opens a detail panel on the right side of the page.

The panel header shows:

• File name and size
• Navigation arrows to move between documents without closing the panel
• Category badge and Ready for AI status badge

Below the header, the panel shows:

• Created and Updated timestamps
• Action buttons: Open Original, Download, Rename, Delete
• The full rendered content of the document

For PRD documents that have finished indexing, an additional Generate Tasks button appears in the panel. This triggers AI Task Discovery against that document, surfacing tasks aligned with the requirements it contains.

Document States

When first added, a document shows a Processing status while indexing is underway. The detail panel shows a percentage progress during this phase. Once indexing completes, the status updates to Ready for AI and the document becomes available to all agent workflows.

The status filter chip in the document list updates in real time, showing counts for Ready and Processing documents separately.

Adding Documents

Click + Add Document to open the Add Document modal.

First, select the document Type: either PRD or Docs. This determines which category the document is filed under once added.

Then choose a method:

Upload File: drag and drop a file into the upload area or click to browse. Supported formats include PDF, Markdown, HTML, JSON, CSV, DOCX, and other common formats up to 10 MB. Use this when migrating existing PRDs from Notion or Google Docs, uploading formal architecture documents, or attaching compliance documentation. Click Upload to confirm.

Write Document: enter a Title and write content directly in the Content field, which supports Markdown formatting. The document will be saved as a .md file. This is useful for lightweight clarifications, architecture notes specific to this repository, or temporary planning documents. For example, after importing a legacy repository you might write: "This service is part of a larger event ingestion pipeline. It should not directly access the analytics database." Click Create Document to save.

Click Cancel at any point to dismiss the modal without saving.

Auto-Generated Code Index

When a repository is imported, Origin automatically generates a code index document using Repomix. This document is a merged representation of the entire codebase combined into a single file and appears in the Docs category.

It is structured as follows:

• File summary: an overview of the document's organization
• Directory structure: a listing of repository paths and files
• Repository files: the full contents of each file, with file path as an attribute

The code index enables repository-wide analysis without scanning individual files during each agent run. It should be treated as read-only, any changes should be made to the original repository files, not this document.

Refreshing Indexing

The Refresh button re-indexes context and repository documents. Use this after uploading new PRDs, updating documentation, or making significant repository changes.

Re-indexing affects only contextual understanding. It does not modify source files or tasks.

How Context Influences Agent Behavior

Once indexed, context documents are available across AI Task Discovery, task execution, and chat reasoning.

• If a PRD states strict tenant isolation, discovery may generate RLS hardening tasks.
• If documentation includes performance constraints, agents may prioritize index suggestions.
• If a schema document highlights sensitive fields, assessments may flag related exposures.

Context does not automatically change code. It shapes reasoning.

How Teams Typically Use Context

Teams use Project Context to:

• Keep product intent close to implementation
• Reduce repetitive explanations in chat sessions
• Align AI suggestions with real-world constraints
• Improve onboarding for new contributors
• Maintain consistent architectural decisions over time

When Context is maintained properly, task planning becomes more accurate, reviews require fewer corrections, and AI-assisted development stays aligned with product intent rather than drifting toward generic improvements.