8000 GitHub - cgpavlakos/genai_playground: Sample application for Oracle Cloud GenAI Agents service and Generative AI LLM Playground
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

cgpavlakos/genai_playground

Repository files navigation

About

This 4-in-1 demo showcases the power of Oracle's Cloud Platform and AI Services with real-world utility:

  • The LLM Playground shows Oracle Cloud Generative AI in action.
  • You can play with the hyperparameters and change models with the sidebar.
  • USSC Helpline Agent uses Generative AI Agents for a retrieval-augmented generation (RAG) chat.
    • The documents used are all publicly available on ussc.gov.
    • This was created originally as a tailored demo for the customer.
  • The Speech to Text page uses OCI Speech to transcribe, and Generative AI to provide a summary.
  • The Summarize Document page uses Generative AI to provide a summary of any text or PDF file.
  • The UI is Streamlit, an open-source Python framework running on OCI Compute. llm_screenshot

Live Demo App

https://genai.pavlakos.cloud (currently not so live)

Generative AI Agent Features

  • Secure & Private: All data remains confidential within your Oracle Cloud tenancy, benefiting from all of the built-in security features.
  • Chat with the GenAI Agent: Have a conversation - ask questions and get insightful answers.
  • View Citations: Explore the sources behind the agent's responses to validate the responses are grounded.

Underlying Architecture

diagram

  • Object Storage stores private data files for the RAG knowledge base and speech to text functionality with AES256 encryption.
  • Generative AI Agents provides the RAG pipeline as a PaaS service.
  • Open Search is a managed service for the knowledge base, automatically indexed for fast search.
  • Generative AI Can be either shared or dedicated hosting, with your choice of Cohere and Meta for Large Language Model (LLM).
  • Compute - a E5 Flex (1 OCPU) virtual machine hosts the Streamlit app to provide the UI.

Known Issues

  • Live Demo App is down
    • need to rotate SSL certs and update dns

Try it Out in Your Oracle Cloud Tenancy

Before you start

  • You must have an Oracle Cloud Account subscribed to the Chicago, Frankfurt, or London region
  • You must already have an Generative AI Agents endpoint available
    • this app only provides a front end
    • currently only available with whitelisting in Chicago region
  • You must set up oci config in order to authenticate to the agent endpoint.
  • You must update .streamlit/secrets.toml
    • agent_endpoint_id
    • compartment_id
    • other items as noted in comments

Oracle Cloud Free Tier

  • Did you know you can sign up for an Oracle Cloud Free Tier Account?
    • You get $300 in credits for 30 days to use on all available services
    • You get Always Free Resources for as long as you want them including but not limited to:
      • Oracle Autonmous Database with APEX
      • NoSQL Database
      • 2 AMD Compute Instances with 1/8 OCPU and 1 GB RAM each
      • Up to 4 ARM Compute Instances with a combined total of 4 OCPU and 24 GB of RAM
      • 200 GB Block Storage
      • 20 GB Object and Archive Storage
      • Security Services including Certificates, Valut and Bastion (managed SSH jumpbox)
      • Observability Services including Logging, Monitoring, Notifications
      • Networking Services including VCN, Load Balancer, Site-to-Site VPN, and 10 TB of outbound data transfer per month
  • If you are one of my customers I can get you up to 60 days and $500 in credits. Reach out to me directly and I will get you set up.

Get started

  1. Set up Generative AI Agents service and note the agent_endpoint_id
  2. Make sure you have port 8501 open on security list
  3. Launch a VM with Oracle Linux 8 image and attach setup.sh as cloud-init script
  4. SSH into your VM (opc@ipaddress) and check the log at /home/opc/genai_agent_setup.log
  5. Run setup.sh if you did not add it as cloud-init script
  6. Set up OCI CLI config
  7. Update .streamlit/secrets.toml with your agent_endpoint_id and compartment_id
  8. Use run.sh to run the demo
  9. Your application will be running on http://server-ip-address:8501
  10. Set it up as a service so you can restart it like sudo systemctl restart streamlit.service

Screenshots

RAG Agent:

agent_screenshot agent_screenshot-2 agent_screenshot-3

LLM Playground:

llm_screenshot

Speech to Text:

stt_screenshot_1 stt_screenshot_2

Summarize Document:

sum_screenshot_1

About

Sample application for Oracle Cloud GenAI Agents service and Generative AI LLM Playground

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published
0