title	Thoughtful AI Support Agent
emoji	🤖
colorFrom	blue
colorTo	green
sdk	gradio
app_file	app.py
pinned	false

Thoughtful AI Support Agent (FAQ + LLM fallback)

A simple customer support agent that:

• answers from a hardcoded Thoughtful AI FAQ using fuzzy retrieval
• falls back to an LLM when no FAQ match is found (optional)

Run locally

python -m venv .venv
# Windows: .venv\Scripts\activate
source .venv/bin/activate

pip install -r requirements.txt
python app.py

Open the URL printed in your terminal (usually http://127.0.0.1:7860).

Enable LLM fallback (optional)

Set an API key:

export OPENAI_API_KEY="YOUR_KEY"
# optional:
export OPENAI_MODEL="gpt-5.2"
python app.py

If OPENAI_API_KEY is not set, the app will still run and will gracefully respond when no FAQ match is found.

Deploy on Replit (preferred by prompt)

1. Create a new Replit (Python)
2. Add these files
3. Add OPENAI_API_KEY (Secrets) if you want LLM fallback
4. Run: python app.py

Replit quick-start

Option A: Import this zip

1. Create a new Repl → choose Python
2. Upload and extract this project zip into the Repl
3. Click Run (the .replit file runs python app.py)

Option B: Import from GitHub

1. Create a new Repl → Import from GitHub
2. Paste your repo URL
3. Click Run

Secrets (optional, for LLM fallback)

• Add OPENAI_API_KEY in Replit → Secrets (lock icon)
• Optionally add OPENAI_MODEL (defaults to gpt-5.2)

Notes:

• Replit sets a PORT environment variable automatically; app.py uses it for Gradio.
• If installs ever fail, open the Shell and run: pip install -r requirements.txt

Hugging Face Spaces deploy

1. Create a new Space on Hugging Face and choose Gradio.
2. Upload these files (or push via git).
3. (Optional) Add OPENAI_API_KEY and OPENAI_MODEL under Settings → Secrets.
4. Wait for the build and open the Space URL.

Notes for Hugging Face Spaces

This build avoids ChatInterface(type=...) for compatibility with older Gradio versions on some hosts.

Free-key LLM fallback (recommended: Groq)

For a lightweight demo without entering payment details, you can use Groq's free tier by setting:

• GROQ_API_KEY
• (optional) GROQ_MODEL (default: llama-3.3-70b-versatile)

On Hugging Face Spaces: Settings → Secrets.

If you set both GROQ_API_KEY and OPENAI_API_KEY, Groq is used first.

Improving Groq fallback answers about Thoughtful AI

The LLM fallback includes a short, hardcoded context summary of Thoughtful AI (derived from thoughtful.ai), plus the FAQ Q/A pairs, to keep responses grounded when a user's question doesn't match an FAQ entry.

Edit THOUGHTFUL_AI_CONTEXT in llm_fallback.py if you want to add more product/company details.