Create a project
- UI
- SDK
- Navigate to your organization’s project list
- Click + Project
- Enter a project name
- Optionally add a description
- Click Create
Configure AI providers
Project-level AI provider keys override organization-level keys. Use project-level keys when:- Different projects need separate billing or rate limits
- You want to isolate API usage by project
- Projects require different provider accounts or credentials
- Navigate to your project.
- Go to Configuration.
- Under Project, select Project AI providers.
- Click the provider you want to configure.
- Enter your API key for that provider.
- Click Save.
Secrets are encrypted at rest using transparent data encryption with a unique 256-bit key and nonce.
Add custom providers
Braintrust supports custom AI providers at both the organization and project level. See Custom providers for details on configuring custom endpoints.Add tags
Tags help organize and filter logs, datasets, and experiments:- Go to Settings.
- Under Project, select Tags.
- Click Add tag.
- Enter tag details:
- Name: Tag identifier.
- Color: Visual indicator.
- Description: Optional explanation.
- Click Save.
Configure human review
Define scores for manual review by users or your team:- Go to Settings.
- Under Project, select Human review.
- Click + Human review score.
- Enter a name and description for your score. Descriptions support Markdown.
- Select a score type:
- Categorical score: Predefined options with assigned scores. Each option gets a unique percentage value between 0% and 100% (stored as 0 to 1). Use for classification tasks like sentiment or correctness categories. Also supports writing to the
exptectedfield instead of creating a score. - Continuous score: Numeric values between 0% and 100% with a slider input control. Use for subjective quality assessments like helpfulness or tone.
- Free-form input: String values written to the
metadatafield at a specified path. Use for explanations, corrections, or structured feedback.
- Categorical score: Predefined options with assigned scores. Each option gets a unique percentage value between 0% and 100% (stored as 0 to 1). Use for classification tasks like sentiment or correctness categories. Also supports writing to the
- Click Save.
Create aggregate scores
Combine multiple scores into a single metric:- Go to Settings.
- Under Project, select Aggregate scores.
- Click Add aggregate score.
- Define the aggregation:
- Name: Score identifier.
- Type: Weighted average, minimum, or maximum.
- Selected scores: Scores to aggregate.
- Weights: For weighted averages, set score weights.
- Description: Optional explanation.
- Click Save.
Set up online scoring
Define project-level scoring rules that automatically evaluate production logs as they arrive. These rules can be created here or when creating and editing scorers.- Go to Settings.
- Under Project, select Online scoring.
- Click Add rule.
- Configure the rule:
- Name: Rule identifier.
- Scorers: Select which scorers to run.
- Sampling rate: Percentage of logs to evaluate (1-100%).
- Filter: Optional SQL query to select specific logs.
- Span type: Apply to root spans or all spans.
- Click Save.
Configure span iframes
Customize how specific span fields render in the UI:- Go to Settings.
- Under Project, select Span iframes.
- Click Add iframe.
- Configure rendering:
- Field path: Which field to render (e.g.,
output.html). - iframe URL: Template for the iframe src attribute.
- Field path: Which field to render (e.g.,
- Click Save.
Set comparison key
Customize how experiments match test cases:- Go to Settings.
- Under Project, select Advanced.
- Enter a SQL expression (default:
input). - Click Save.
input.question- Match by question field only.input.user_id- Match by user.[input.query, metadata.category]- Match by multiple fields.
Edit project details
Update project name and description:- Navigate to your project.
- Click Edit project in the top-right.
- Modify name and description.
- Click Save.
Delete a project
- Navigate to Configuration.
- Scroll to the bottom of the page.
- Click Delete project.
- Confirm by typing the project name.
- Click Delete.
Next steps
- Control access to projects with permission groups
- Set up automations for project-specific alerts
- View logs filtered by tags and metadata
- Run evaluations on project datasets