1
Send the prompt to your agent
Help me add the Higgsfield API to this project. First, inspect the codebase to understand its framework, backend or agent architecture, authentication, user or tenant identity model, file storage, and background job handling.
Read the current documentation starting at the [Higgsfield docs index](https://docs.higgsfield.ai/docs/llms.txt). Follow the relevant authentication, model, SDK, request lifecycle, polling, and webhook documentation. Verify endpoint availability and request schemas before implementing; do not assume sample code is current.
Based on the codebase, briefly explain where Higgsfield would fit. Then ask what I want my users or agent to create. Offer relevant options such as image generation, animating uploaded images, or generating videos from prompts. If it is unclear whether I want an application feature or a tool for my coding agent, clarify before making changes.
Once the use case is clear, implement the integration using the project’s existing conventions:
- Choose the documented official SDK or direct REST requests based on the project’s language and architecture.
- For REST, use `https://api.higgsfield.ai` with `Authorization: Key ${HF_API_KEY_ID}:${HF_API_KEY_SECRET}` and `Content-Type: application/json`. Keep credentials exclusively on the server, following the [authentication guidance](https://docs.higgsfield.ai/docs/authentication). Add placeholder environment variables and setup instructions without exposing secrets.
- For text-to-video, use `/bytedance/seedance-2.0/text-to-video` as the starting example, after confirming its current schema and availability. Use this request as the reference:
```bash
curl --request POST \
--url https://api.higgsfield.ai/bytedance/seedance-2.0/text-to-video \
--header "Authorization: Key ${HF_API_KEY_ID}:${HF_API_KEY_SECRET}" \
--header "Content-Type: application/json" \
--data '{
"prompt": "A cinematic tracking shot along a sunlit coastal road",
"resolution": "720p",
"generate_audio": true,
"duration": 5,
"aspect_ratio": "16:9"
}'
```
- Treat generation as asynchronous. Save the returned `request_id`, associate it with the requesting user or tenant, and enforce ownership when retrieving results.
- Follow the documented [polling lifecycle](https://docs.higgsfield.ai/docs/concepts/polling), including backoff, timeouts, and terminal states. Consider webhooks when the project supports background processing.
- Validate inputs, handle API errors and rate limits, avoid duplicate generation submissions, and expose clear progress and results.
Complete the agreed implementation, run relevant checks, and summarize what changed, which credentials I need to configure, and how to try the feature.
