README.md

metadata

title: Intern Problem Solver API
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: false
app_port: 7860

Intern Problem-Solving API

A multi-agent FastAPI backend that helps interns understand real-world problems and build structured thinking. Five specialized agents run in sequence — each feeding context to the next — to deliver a complete analysis with actionable next steps.

Agent Pipeline

Input → [1] Problem Analyst → [2] Root Cause → [3] Solutions → [4] Action Plan → [5] Thinking Coach

Agent	Job
Problem Analyst	Identifies core problem, stakeholders, known vs unknown
Root Cause Analyst	Breaks down WHY the problem exists across technical, process, and people dimensions
Solution Brainstorm	Generates 7 diverse ideas — quick wins, medium-term, strategic, and unconventional
Action Planner	3 concrete next steps the intern can start immediately
Thinking Coach	Honest feedback on the intern's framing — are they a problem solver or task executor?

---

Endpoints

Method	Endpoint	Returns
GET	/	API info and endpoint list
GET	/health	Health check
POST	/analyze	Full analysis as JSON
POST	/analyze/stream	Server-sent events — streams each agent live
POST	/analyze/pdf	Downloadable PDF report

---

Request Body

All three POST endpoints accept the same JSON body:

{
  "content": "Paste your problem, transcript, or blocker here",
  "intern_name": "Your Name",
  "intern_role": "AI Developer Intern",
  "intern_goal": "What you are trying to achieve this week"
}

Field	Required	Default
content	✅ Yes	—
intern_name	No	"Intern"
intern_role	No	"AI Developer Intern"
intern_goal	No	""

---

Response — /analyze

{
  "problem_statement": "## Core Problem\n...",
  "root_causes": "## Root Cause Analysis\n...",
  "solutions": "## Solution Ideas\n...",
  "action_plan": "## Your Next Steps\n...",
  "thinking_feedback": "## Thinking Feedback\n..."
}

---

Response — /analyze/stream

Server-sent events. Each event is a JSON object on a data: line:

data: {"event": "agent_start", "agent": "analyst", "label": "Problem Analyst"}

data: {"event": "token", "agent": "analyst", "text": "## Core Problem\n"}

data: {"event": "agent_done", "agent": "analyst"}

data: {"event": "done"}

Event types:

Event	Fields	Meaning
agent_start	agent, label	Agent has started running
token	agent, text	Next streamed token from this agent
agent_done	agent	Agent finished
done	—	All 5 agents complete

---

Setup

Environment Variable

Set this as a Secret in your HuggingFace Space settings:

ANTHROPIC_API_KEY=sk-ant-...

Do not put it in the Dockerfile or commit it to the repo.

Files Required

main.py
requirements.txt
Dockerfile
README.md

---

Local Development

# Install dependencies
pip install -r requirements.txt

# Run with your API key
ANTHROPIC_API_KEY=sk-ant-... uvicorn main:app --reload --port 7860

API docs available at http://localhost:7860/docs once running.

---

Client

The frontend (client.html) is a standalone HTML file — no build step, no framework. Host it anywhere: a second HF Space (static SDK), GitHub Pages, or just open it locally in a browser.

Point the Backend API URL field in the client to this Space's URL:

https://your-username-intern-problem-solver.hf.space

---

Stack

• FastAPI — API framework
• Anthropic SDK — claude-sonnet-4-6 for all agents
• ReportLab — PDF generation
• Uvicorn — ASGI server