Project Context

Overview

Project Context is a project-level background document that the AI silently references when generating responses, running agents It acts as a persistent system briefing — you write it once, and every AI interaction in the project automatically benefits from it. Use Project Context to describe things the AI would otherwise not know: the domain your team operates in, how your systems are named, the workflows you follow, the terminology that is specific to your organization, and any constraints or policies that should always apply.

Project Context is only available in team (non-public) projects. It does not appear in the global project or personal project workspaces.

Consistent AI Behavior

Define once, applied everywhere. Every agent in the project automatically uses the same background context.

Domain-Aware Responses

Give the AI your terminology, workflows, and constraints so it speaks your language and avoids generic or incorrect answers.

Per-Agent Override

Agents can individually opt out of the Project Context when their task requires a neutral, context-free perspective.

Accessing Project Context

You need the models.project_context.view permission to see this page. Editing requires models.project_context.edit. Users with Viewer or Monitor roles typically have view-only access. Contact your project administrator if the page is missing from your Settings menu.

To access Project Context:

Click the Settings icon in the main navigation sidebar.
Select Project Context from the vertical settings drawer on the left.

The Project Context editor opens in the main content area, with two main components: the enable toggle card at the top, and the Project Background editor below it.

Project Background Editor

The Project Background editor is where you write and manage the context document for your project. It is a full Markdown editor with real-time character counting, a Build with AI button to generate content using AI, an import tool for loading content from a file, and a preview mode so you can verify formatting before saving. The editor is only visible when Project Context is enabled, or when content already exists (so it remains accessible for editing even while the feature is temporarily disabled).

Disabling Project Context does not delete the saved content. The content is preserved and can be re-enabled at any time. A banner is shown below the toggle when the context is disabled but content exists: “Project Context is turned off. The project background is not applied to AI responses or workflows.”

What to Write in Project Context

Project Context is written in Markdown and read by the AI as background information before processing any request. The content should be factual, concise, and focused on information that applies broadly across the project — not instructions specific to a single task.

Use the Edit mode () to write and update content directly in the editor.
Switch to Preview mode () at any time to see the rendered Markdown before saving. Toggle between them using the buttons in the top-right corner of the editor.

Recommended Content

Category	What to Include	Example
Domain & purpose	What the project or team does	”This project supports the QA automation team at Acme Corp. Our primary function is test planning, defect tracking, and release validation.”
Terminology & naming	Internal names, acronyms, system names	”Our CI system is called ‘RocketCI’. Environments: DEV, STAGING, PROD. ‘Ticket’ always refers to a Jira issue.”
Key workflows	How work gets done	”All code changes go through a PR review with at least 2 approvals. Hotfixes bypass staging and go directly to PROD after QA sign-off.”
Constraints & policies	Rules the AI should always respect	”Never suggest storing credentials in code. Always recommend secrets vault usage. All API responses must be paginated.”
Tech stack	Languages, frameworks, platforms in use	”Backend: Python 3.11 + FastAPI. Frontend: React 18 + TypeScript. Database: PostgreSQL 15. Deployed on AWS EKS.”
Team context	Roles, responsibilities, conventions	”The on-call engineer handles P1/P2 incidents. Bug severity: P1 (outage) → P4 (cosmetic). Sprint duration: 2 weeks.”

The editor enforces a maximum of 2,500 characters. A counter below the editor shows how many characters remain. When the limit is reached, the Save button is disabled and a warning message appears.

Build with AI

The Build with AI button lets you generate Project Context content automatically using AI. You describe your project in plain language, and the AI produces a structured Markdown draft ready to review and edit before applying it to the editor. The Build with AI button appears in the Project Background editor toolbar (next to the Import button) when the feature is enabled and you have edit permission. It is not visible to users with view-only access. To generate Project Context using AI:

Click the Build with AI button in the editor toolbar.
The Build with AI dialog opens.
In the text field, describe your project.
Click Generate.
The Review step opens, displaying the generated Project Background in an editable text field.
You can edit the generated text directly before applying it.
Click Apply to insert the content into the editor, or Back to prompt to return and refine your description.
After applying, the content appears in the editor as unsaved changes. Click Save to persist it.

Clicking Apply does not automatically save the Project Context. The content is placed into the editor as an unsaved draft. You must click Save to persist the changes.

When using Append to existing content, the combined length of the existing content and the generated draft must not exceed 2,500 characters. If it does, the character counter turns red and the Apply button is disabled until you shorten the generated text.

Importing Content

Click the Import button in the editor toolbar to upload a .md (Markdown) file from your local machine. The file content replaces the current editor content.

Files exceeding 2,500 characters are rejected with an error. Trim the file or paste a condensed version directly into the editor if needed.

Example Project Context

## Project: ACME Platform — Backend Team

**Domain:** E-commerce order management and fulfillment platform.

**Tech Stack:**
- Backend: Python 3.11, FastAPI, Celery
- Database: PostgreSQL 15 (primary), Redis (cache/queue)
- Infrastructure: AWS EKS, Terraform

**Key Terminology:**
- "Order" = a confirmed customer purchase (not a cart)
- "Fulfillment" = warehouse picking, packing, and shipping
- "SLA breach" = any order not fulfilled within 48h of placement

**Workflows:**
- All services follow the hexagonal architecture pattern
- New endpoints must include OpenAPI docstrings and integration tests
- Database migrations use Alembic; never modify existing migration files

**Constraints:**
- Never suggest raw SQL queries; always use the ORM (SQLAlchemy)
- All secrets must be stored in AWS Secrets Manager, not environment variables
- API responses follow the `{data, meta, errors}` envelope format

Keep the Project Context focused on stable facts — things that rarely change. Avoid adding dynamic information (e.g., current sprint goals or who is on-call this week), since outdated context can mislead the AI.

Per-Agent Override: Ignore Project Context

By default, every agent in the project receives the Project Context. If a specific agent should operate without it — for example, a general-purpose assistant whose responses should not be shaped by your domain terminology, or a utility agent that works across multiple projects — you can disable the context for that agent individually using the Ignore Project Context option. The Ignore Project Context option is available both when creating a new agent and when editing an existing one. In both cases it is located in the ADVANCED section of the agent configuration form. To enable the override:

Open the agent configuration — either click Create New Agent from the + menu in the chat toolbar, or open an existing agent for editing.
Expand the ADVANCED section.
Check Ignore Project Context.

“When enabled, this agent will not use the project background context in its responses.”

The Ignore Project Context checkbox only appears in the agent’s ADVANCED section when the feature is enabled by the platform. It has no effect if Project Context is already disabled at the project level.

Troubleshooting

'Project Context' tab is not visible in Settings

Confirm you have the models.project_context.view permission. Ask your project administrator to check your role.

Save button stays disabled after editing

Check the character counter below the editor. If the content exceeds 2,500 characters, the Save button is disabled until you reduce the length.
If the character count is within the limit, verify that you have actually made a change — the Save button is only enabled when unsaved changes exist.

Imported file was rejected

Only .md (Markdown) files are accepted by the Import button.
The file content must not exceed 2,500 characters. If the file is larger, trim it before importing or paste a condensed version directly into the editor.

Project Context is saved but AI responses haven't changed

Confirm the enable toggle is on. If it is disabled, the context is not applied even though content is saved.
Verify the agent being used does not have Ignore Project Context checked in its ADVANCED settings.
Start a new conversation — existing conversations may not retroactively pick up context changes mid-session.

'Ignore Project Context' checkbox is missing in agent Advanced settings

The checkbox is only shown when the feature is enabled at the platform level. Contact your platform administrator if it is expected to appear but is missing.
Also verify you are editing a draft version of the agent — published versions are read-only.

Settings Overview — All available Settings sections and navigation
AI Configuration — Configure LLM models and AI configurations for the project
How to Create and Edit Agents from Canvas — Agent configuration including the Ignore Project Context option

Welcome to ELITEA

Key Concepts

Reference

Platform Overview

Overview

Consistent AI Behavior

Domain-Aware Responses

Per-Agent Override

Accessing Project Context