> ## Documentation Index
> Fetch the complete documentation index at: https://docs.oqoqo.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Oqoqo AI features
> Assign models for Ask Oqo, Rubric generation, Run evals, Run insights, and Experiment insights.
**Oqoqo AI features** are the AI roles that help you author work, score runs, and explain results.
Open **Organization settings → Oqoqo AI features**. Assign a model to each role your organization needs.
## Roles
| Role | What it does | Where you feel it |
| ----------------------- | --------------------------------------------------------------- | -------------------------------------- |
| **Ask Oqo** | Answers questions about your project's experiments and runs | Ask Oqo page, header, and Home |
| **Rubric generation** | Drafts rubric criteria from the task instructions | Generate / Regenerate on a task rubric |
| **Run evals** | Scores each run against the task rubric, criterion by criterion | Run → **Evals**; Matrix pass rates |
| **Run insights** | Finds frictions and patterns in each evaluated run | Experiment → **Frictions** |
| **Experiment insights** | Combines every run's insights into one experiment-level read | Experiment → **Frictions** |
## Prerequisites
1. Add a compatible provider under **Organization settings → Model providers**, or connect a supported personal Integration under **Profile → Integrations**. See [Model providers](/setup/model-providers).
2. Assign a model to each role you plan to use.
3. Connect your account under **Profile → Integrations** if a selected model uses a personal subscription.
4. For Ask Oqo, authorize project access when prompted.
Supported personal Integrations can power Oqoqo AI features, including **OpenAI (ChatGPT Subscription)**, **Anthropic (Claude Subscription)**, and **Nous Portal**. Oqoqo checks the initiating person's connection before it starts the work.
## Ask Oqo
Ask Oqo is the in-app assistant. It helps you inspect experiments, explain Matrix and Frictions, draft small plans, and find the right dashboard screen.
See the full guide: [Ask Oqo](/setup/ask-oqo).
## Rubric generation
On a task rubric, choose **Generate** or **Regenerate** to draft pass or fail requirements from the task instructions.
Review every draft. Edit the requirements until they match what "done" means for your product. Rubric generation does not save until you accept the draft.
See [Tasks and rubrics](/experiments/defining-experiments).
## Run evals
Run evals use a judge model to mark each rubric requirement pass or fail and explain why.
* Turn Evals on at [launch](/experiments/launch-options)
* Or trigger / re-trigger Evals later on saved recorded work
Evals are judge-based. They are not a separate deterministic checker.
See [Evals and Insights](/evaluation/evals-and-insights).
## Run insights and Experiment insights
Insights analyze traces and output.
* **Run insights** finds frictions in each run
* **Experiment insights** combines those findings into one experiment read
Turn Insights on at launch, or trigger them later. Findings appear on the experiment **Frictions** tab.
See [Frictions](/evaluation/frictions).
## Launch overrides
At launch, you can confirm or override the models used for Run evals and Insights for that experiment. Organization defaults remain the baseline for Ask Oqo and Rubric generation.
## Related pages
Use the in-app assistant.
How scoring and friction analysis work.
Connect the models these roles use.
Runs cover platform work. Model inference is separate.