Operational Checklists
Use these checklists before sending real work through Redbit. They are intentionally short and operational: each item changes whether a user, customer, or teammate should proceed.Who Should Read This
| Reader | Use this page when |
|---|---|
| Workspace owner | You are preparing a Redbit environment for users |
| First-time user | You are about to run a real generation task |
| Support or success team | A video or provider workflow failed and you need a structured first pass |
Before You Read
Know which route you are using: direct provider, relay provider, custom relay, or Local Core. Do not paste credentials into prompts, Cards, Workshop scripts, or Agent chat.Pre-Launch Configuration
| Check | Why it matters |
|---|---|
Confirm the workspace URL and route (/dashboard, /workshop, Settings) | Users need a stable entry point before model setup |
| Configure only the provider keys required for the planned workflow | Limits accidental exposure and reduces provider confusion |
| Confirm whether the route is direct provider, relay, or custom relay | Visible models and payload shape differ by route |
| Test the describe/Agent runtime profile if the Agent will be used | Tool calling, structured output, system prompt, and vision support vary by model |
| Pair Local Core only when local media, MCP, automation, FFmpeg, plugin, or storage bridge features are required | Browser-only generation does not need a local engine |
| Verify provider terms for confidential media | Redbit coordinates requests; provider privacy and retention policies still apply |
| Decide how assets will be backed up or exported | Browser profile deletion, site-data clearing, or quota pressure can remove local data |
Before the First Generation
| Check | Why it matters |
|---|---|
Select the correct tab: Describe, Image, Video, Audio, or Speech | The tab controls visible models, defaults, and input requirements |
| Confirm model family and variant | Different models produce different outputs from the same prompt |
| Confirm ratio, duration, resolution, and audio settings where relevant | Video and image providers reject unsupported combinations |
| Attach only the references the selected model accepts | Seedance, motion models, image editors, and describe models use different inputs |
| Keep source media out of Agent chat unless the task requires the Agent to inspect or route it | This avoids unnecessary exposure through assistant context |
| Save or pin outputs you intend to reuse | Temporary recent assets may be cleaned up |
| Review the card error before retrying | Retrying without changing credentials, inputs, or route usually repeats the same failure |
Video Failure Triage
| Symptom | First checks |
|---|---|
| Missing reference error | Confirm the mode and required reference role: first frame, last frame, image reference, video reference, or audio reference |
| Seedance reference mix error | Use either first/last-frame roles or multi-reference roles in one task, not both |
| Provider or relay error | Check API key type, base URL, relay capability, model ID mapping, quota, and provider status |
| Model disappears after enabling relay | The relay capability filter may not include that family or variant |
| Slow or stalled video | Try a shorter duration, lower resolution, fast variant, or a provider route with less queue pressure |
| Result does not match prompt | Compare provider/model capability, reference strength, prompt specificity, and whether the selected model supports that input type |
| Agent says it completed but no media exists | Check whether a generation tool actually returned { success: true }; text-only completion is not enough |