Evaluation Guide
Use this guide to evaluate Redbit as a product workflow, not only as a media-generation interface. A good evaluation checks the creative path, provider route, reuse model, Agent boundary, and team handoff artifact.Who Should Read This
| Reader | Use this page to |
|---|---|
| Potential user | Decide whether Redbit matches the kind of creative work you run |
| Team lead | Compare Redbit with a direct provider console, chat tool, or custom internal tool |
| Engineering or creative operator | Plan a small proof of fit without exposing unnecessary credentials or source media |
Before You Evaluate
Read Concepts and Glossary first if the product terms are new. Redbit coordinates cards, assets, providers, and optional local integrations. It does not own model quota, provider uptime, model policy, billing, or output quality.Evaluation Outcome
By the end of a useful evaluation, you should be able to answer:| Question | Good evidence |
|---|---|
| Can the team run its common input type? | A representative prompt, source image, video, audio, or script works in the selected Card or Workshop path |
| Is provider access clear? | Settings has the minimum direct provider, relay, or custom relay route needed for the test |
| Can outputs be reused? | Useful results are pinned or selected through Asset Dock and SmartPicker, not lost in a one-off thread |
| Is the Agent helpful without becoming unsafe? | The Agent handles bounded workspace actions, while credentials, billing, deletion, and external effects stay reviewed |
| Can another teammate continue the work? | Cards, Workshop project state, pinned assets, exports, and notes are understandable without private chat context |
45-Minute Evaluation Path
| Step | Action | Pass condition |
|---|---|---|
| 1. Pick one real scenario | Choose one input, one output, and one reviewer | The scope fits one short session |
| 2. Confirm provider route | Configure only the model family needed for the scenario | Settings shows a direct provider, relay, or custom relay route that matches the media type |
| 3. Run a manual baseline | Create a Card or Workshop project before using the Agent | The user can inspect prompt, model, status, result, and error state |
| 4. Reuse one output | Save a result to Asset Dock and select it again through SmartPicker | The asset is available as a later reference |
| 5. Try bounded Agent help | Ask for a multi-step workspace task with clear targets | Tool actions are reviewable and produce visible workspace changes |
| 6. Handoff the result | Download outputs, export a Workshop package, or record the selected assets and route | Another teammate can continue without guessing hidden state |
Fit Signals
| Redbit is a strong fit when | Redbit is a weak fit when |
|---|---|
| Work involves repeated prompts, references, model choices, or review steps | You only need one occasional generation from one provider console |
| Outputs need to become inputs for later Cards or Workshop scenes | Results do not need reuse, comparison, or project structure |
| The team wants BYOK routing with explicit provider and relay control | The team expects Redbit to supply every model credit, quota, or provider SLA |
| Operators need a visible Agent that calls bounded tools | The desired automation requires unrestricted shell access or unreviewed external posting |
| Handoff matters between creative, growth, support, and engineering roles | Work always stays with one person and one chat transcript |
Realistic Scenario Cards
Ecommerce product image set
Input: product photos, selling points, target ratios, and brand notes. Recommended path: Image Cards with Series
ecommerce, Asset Dock, SmartPicker, and a configured image provider. Output: reviewed hero, detail, usage, comparison, and social-ready images. Watch for product identity drift, provider policy, and human review before publishing.Short video storyboard
Input: campaign brief, script, reference image, and scene list. Recommended path: Workshop for script and scenes, then Video Cards or Seedance modes for clips. Output: scene images, video clips, optional voice/music assets, and export packages where configured. Watch for video queue time, duration, resolution, and reference-role rules.
Social growth report task
Input: objective, public links or approved data sources, and review criteria. Recommended path: Agent runtime with configured search, CMO, growth-report, MCP, or Local Core tools only when available. Output: reviewable findings, Cards, assets, or Workshop preparation. Watch for real-account actions, tool availability, and provider/runtime capability tests.
Team Handoff Checklist
| Artifact | Why it matters |
|---|---|
| Selected Cards and groups | Shows prompts, settings, status, results, and failures |
| Pinned Asset Dock items | Keeps reusable inputs visible for the next operator |
| Workshop project | Preserves script, scenes, consistency references, generated media, and export settings |
| Provider route note | Explains whether the result used direct provider, relay, custom relay, or Local Core |
| Known limits | Captures slow video, provider errors, model mismatch, or manual review needs |