-
-
Rapid Impact Partner Assistant UI: event intake form and generated donor, social, and checklist outputs.
-
Cloud Run Next.js frontend → FastAPI backend → Gemini API via Secret Manager, with Cloud Logging.
-
Cloud Run dashboard with frontend and backend services in us-central1.
-
Cloud Run backend logs showing autoscaling events and successful /generate + /health requests.
-
Cloud Run service details for the Next.js frontend, showing active revision and public URL.
-
Social media post: Rapid Impact Partner Assistant turns field notes into donor-ready updates in seconds using Cloud Run and Gemini.
-
YouTube thumbnail text: Rapid Impact Partner Assistant • Cloud Run + Gemini
-
Social media post: Rapid Impact Partner Assistant turns field notes into donor-ready updates in seconds using Cloud Run and Gemini.
Inspiration
Working with nonprofit partners, we kept hearing the same story: outreach staff spend hours writing donor recaps after every event, delaying updates and draining limited time. We wanted to prove an AI assistant could lift that burden without forcing teams to learn new tools.
What it does
Rapid Impact Partner Assistant accepts quick field notes and instantly generates three artifacts a donor newsletter paragraph, a social caption, and an internal follow-up checklist. Users can load demo data, tweak their highlights, and copy polished content in one click.
How we built it
We split the system into two Cloud Run services. The FastAPI backend structures prompts and calls Google Gemini through the google-generativeai SDK, normalizing responses to JSON. A Next.js 14 + Tailwind frontend collects inputs, calls /generate, and renders the outputs. Docker, Artifact Registry, and Secret Manager keep deployments reproducible and secure.
Challenges we ran into
We hit authentication hurdles switching from local API keys to Cloud Run secrets, plus model entitlement quirks that required fallback logic. The new AI Studio Build UI also hid the share controls, so we iterated on prompt tooling to document the workflow correctly.
Accomplishments that we're proud of
We shipped a fully hosted experience frontend and backend live on Cloud Run, complete with responsive UI, copy-to-clipboard utilities, and prompt templating resilient to Gemini’s free-form answers. The architecture diagram, README, and Terraform skeleton showcase a production-ready pattern.
What we learned
Prompt design matters: structuring JSON outputs and clarifying tone rules made Gemini far more consistent. On Cloud Run, listening on the injected PORT and targeting linux/amd64 builds avoided deployment failures. Documenting secrets via Secret Manager early kept the project audit-friendly.
What's next for Rapid Impact Partner Assistant
Upcoming iterations include templated tone presets, email/SMS integrations for one-click donor outreach, activity logging for compliance, and expanding AI Studio prompts to support multilingual summaries letting nonprofits serve diverse communities even faster.
Log in or sign up for Devpost to join the conversation.