-
Clair-X – Unleashing filter-free vision right inside Salesforce
Inspiration
Most commercial AI services censor creative output. I wanted full artistic control without leaving the Salesforce environment I already use to run my business.
What Clair-X does
- Prompts are entered in an Agentforce chat panel.
- A Salesforce custom action calls an uncensored FLUX.1-dev model hosted on Hugging Face Spaces.
- The generated image is returned and attached to any lead, opportunity, or campaign.
- Every call is logged, so I can trigger Flows, reports, or marketing journeys.
Technical build
- Agent built with Salesforce Agent Builder.
- Secure HTTPS call from Salesforce Apex → Python FastAPI relay → FLUX.1-dev.
- Signed URLs ensure images upload directly to Salesforce Files.
- Optional Vercel v0 front-end mirrors the agent for public demos.
Challenges
- Passing Salesforce auth context to an external model endpoint.
- Mapping binary image data back into Salesforce files with correct record IDs.
- Keeping round-trip time under five seconds.
Accomplishments
- First filter-free image generator fully operational inside Salesforce.
- End-to-end audit trail for every generation, ready for automation.
- All built by one person in under 48 hours.
Lessons learned
- Agentforce custom actions can safely bridge Salesforce with any external AI.
- Even solo founders can deploy powerful AI pipelines by combining best-of-class tools.
Next steps
- Package for AppExchange so other solo builders can install in minutes.
- Add text-and-audio generation via Hugging Face multi-modal endpoints.
- Introduce voice prompts inside Agentforce.
Built With
- agentbuilder
- agentforce
- audited
- gcp
- huggingfacespaces
- right-inside-agentforce
- salesforce
- salesforcecustomactions
- vercelv0
- with-every-output-instantly-stored
