Hushh Intelligence Portal — Post-Submit Results Guide (Agentic APIs)

The Hushh Intelligence Portal is an internal workflow in this repo that:

1. Collects a lead’s Email, Full Name, and Phone
2. Calls multiple agentic APIs in sequence
3. Merges all agent responses into one “best profile” (with Gemini prioritized)
4. Renders the Analysis Complete results screen with a structured breakdown of user details

This guide documents exactly what is rendered after the user enters details and how to reproduce the same behavior via APIs.

Portal entry (opens in a new tab): hushh.ai/intelligence-portal

1) What happens after the user clicks “Analyze Profile”

The UI collects:

• email
• fullName
• countryCode (e.g., +91, +1)
• phoneNumber (digits only; 10–15 digits)

Then the portal generates a single detailed prompt asking an agent to return a strict JSON profile that includes (examples):

• user ID, full name, email, phone
• address, age, gender
• demographics (marital status, children)
• professional (occupation, education, income)
• lifestyle (diet, cuisine, gym)
• technology + interests
• needs/wants/desires
• intents for 24h / 48h / 72h

2) Which agents are called (and in what order)

The portal calls these agents sequentially (so progress can be displayed):

1. brand — Brand/CRM Agent
2. hushh — Hushh (Supabase Query) Agent
3. public — OpenAI Public Data Agent
4. gemini — Gemini Public Data Agent

Note: In the results merge logic, agents are merged in order brand → hushh → public → gemini so Gemini overwrites/fills last (i.e., Gemini is prioritized).

3) How the portal calls agents (the exact API path)

From the client, the portal calls the local proxy route:

POST /api/a2a/{agent}
Content-Type: application/json

So for example:

• /api/a2a/brand
• /api/a2a/hushh
• /api/a2a/public
• /api/a2a/gemini

If you’re calling from outside the app, use the fully-qualified site URL: https://hushh.ai/api/a2a/{agent}

Request body

The proxy accepts:

{
  "text": "<prompt>",
  "sessionId": "session-1735...",
  "id": "task-abc123"
}

sessionId and id are optional.

4) Reproducing the portal with cURL (agent-by-agent)

Use the same prompt text you’d send via the UI.

Public Data Agent

curl -X POST "https://hushh.ai/api/a2a/public" \
  -H "Content-Type: application/json" \
  -d "{\"text\":\"Can you provide me with a detailed JSON profile of Sundar Pichai having email sundar.pichai@example.com and phone +1 6505559001. The output should strictly be in JSON format.\"}"

Gemini Public Data Agent

curl -X POST "https://hushh.ai/api/a2a/gemini" \
  -H "Content-Type: application/json" \
  -d "{\"text\":\"Can you provide me with a detailed JSON profile of Sundar Pichai having email sundar.pichai@example.com and phone +1 6505559001. The output should strictly be in JSON format.\"}"

Hushh (Supabase Query) Agent

curl -X POST "https://hushh.ai/api/a2a/hushh" \
  -H "Content-Type: application/json" \
  -d "{\"text\":\"Can you fetch all the details of the user having email sundar.pichai@example.com and phone +1 6505559001? Return JSON.\"}"

Brand (CRM) Agent

curl -X POST "https://hushh.ai/api/a2a/brand" \
  -H "Content-Type: application/json" \
  -d "{\"text\":\"Create a JSON profile for Sundar Pichai (email sundar.pichai@example.com, phone +1 6505559001) and summarize top preferences and intents.\"}"

5) How the portal merges the agent outputs (important)

After all agent calls finish, the portal:

1. Extracts a text blob from common JSON-RPC response shapes (e.g., result.status.message.parts[0].text)
2. Strips markdown fences like ```json
3. Parses JSON
4. If the JSON has a userProfile wrapper, it merges userProfile
5. Merges agents in this order: brand → hushh → public → gemini

This means:

• If multiple agents provide the same field, the value from Gemini wins (because it merges last)
• If a field is missing from some sources, later sources can fill it

6) What the “Analysis Complete” screen renders

After merging, the portal renders these sections and fields.

BASIC INFORMATION

• Full Name
• Email
• Phone
• Age
• Gender

LOCATION

• Address
• City
• State
• Postal Code
• Country

DEMOGRAPHICS

• Marital Status
• Household Size
• Children Count

PROFESSIONAL

• Occupation
• Education
• Income Bracket
• Home Ownership
• City Tier

LIFESTYLE

• Diet Preference
• Favorite Cuisine
• Coffee or Tea
• Fitness Routine
• Gym Membership
• Shopping Preference
• Grocery Store Type
• Fashion Style
• Transport
• Sleep Chronotype
• Eco-Friendliness

TECHNOLOGY

• Tech Affinity
• Primary Device
• Favorite Social Platform
• Social Media Usage Time
• Content Preference

INTERESTS

• Sports Interest
• Gaming Preference
• Travel Frequency

INTENT ANALYSIS

• Needs (array)
• Wants (array)
• Desires (array)
• Intent Timeline:
  • 24 hours: category, budget, timeWindow, confidence
  • 48 hours: category, budget, timeWindow, confidence
  • 72 hours: category, budget, timeWindow, confidence

ANALYSIS METADATA

• Confidence Score (from confidence_score / confidence)
• Accuracy Score (from accuracy_score / accuracy)
• Total Fields (count of keys in merged JSON)
• Response Time (average across agents)

7) Viewing Raw Data (debugging)

The portal also provides a Show Raw Data modal that displays:

• The submitted user input
• The merged/parsed profile JSON
• The full per-agent result payloads
• Metadata

Next steps

• Open the Intelligence Portal: hushh.ai/intelligence-portal