A clean, minimal AI chat interface powered by Ollama Cloud Models

DataVyn Labs · Ollama · Built with Streamlit

---

Features

	Feature	Details
🤖	Cloud Models	19 verified Ollama cloud models from OpenAI, DeepSeek, Qwen, Gemini, Mistral, Kimi, GLM, MiniMax and more
⚡	Streaming	Responses stream token by token in real time
🎨	Model Avatars	Each AI shows its company logo in chat (ChatGPT, DeepSeek, Gemini, Mistral...)
📎	File Upload	Attach .txt .pdf .json .py .csv — content sent to the model
🎙	Audio Input	Record voice via mic, auto-transcribed to text
🔐	Secure Login	API key stored in session only — never saved to disk
🌑	Dark UI	Clean Claude.ai-style dark theme with Inter font

---

Quick Start

1. Clone the project

git clone https://github.com/DataVyn-labs/ollama-agent
cd ollama-agent

2. Install dependencies

pip install -r requirements.txt

3. Run

streamlit run app.py

Open http://localhost:8501 in your browser.

---

Getting an Ollama API Key

1. Go to ollama.com and create a free account
2. Navigate to Settings → API Keys
3. Click Create new key
4. Paste it into the app login screen

---

Available Cloud Models

Model	ID	Company
GPT-OSS 120B	gpt-oss:120b	OpenAI (open weights)
GPT-OSS 20B	gpt-oss:20b	OpenAI (open weights)
DeepSeek V3.2	deepseek-v3.2	DeepSeek
DeepSeek V3.1 671B	deepseek-v3.1:671b	DeepSeek
Qwen3-Coder 480B	qwen3-coder:480b	Alibaba
Qwen3-Coder-Next	qwen3-coder-next	Alibaba
Qwen3-Next 80B	qwen3-next:80b	Alibaba
Kimi K2.5	kimi-k2.5	Moonshot AI
Kimi K2 Thinking	kimi-k2-thinking	Moonshot AI
Gemini 3 Flash Preview	gemini-3-flash-preview	Google
MiniMax M2.5	minimax-m2.5	MiniMax
MiniMax M2.1	minimax-m2.1	MiniMax
MiniMax M2	minimax-m2	MiniMax
GLM-5	glm-5	Zhipu AI
GLM-4.7	glm-4.7	Zhipu AI
Devstral 2 123B	devstral-2:123b	Mistral
Devstral Small 2 24B	devstral-small-2:24b	Mistral
Cogito 2.1 671B	cogito-2.1:671b	Essential AI
Nemotron 3 Nano 30B	nemotron-3-nano:30b	NVIDIA

Full list → ollama.com/search?c=cloud

---

Configuration

Edit these constants at the top of app.py (~line 337):

TEMPERATURE = 0.6    # 0.0 = focused  |  1.0 = creative
MAX_TOKENS  = 1200   # max tokens per response (hard cap: 2000)

---

Project Structure

datavynlabs_agent/
├── app.py            # Main Streamlit application
├── logo.png          # DataVyn Labs logo
├── requirements.txt  # Python dependencies
└── README.md         # This file

---

Requirements

streamlit>=1.43.0
requests>=2.31.0

---

Notes

• API key is stored in session memory only — cleared on sign out or tab close
• Audio input (accept_audio=True) requires Streamlit 1.43+ — run pip install --upgrade streamlit if needed
• Uploaded file content is truncated to 4 000 characters before being sent to the model
• All model requests go through https://ollama.com/api/chat
• System prompt instructs the model to give concise, complete answers within the token limit

---

About DataVyn Labs

DataVyn Labs builds intelligent data automation and AI agent tools for modern teams.

🔗 github.com/DataVyn-labs

---

© 2026 DataVyn Labs