Features used in this repository are in preview. Preview versions are provided without a service level agreement, and they are not recommended for production workloads. Certain features might not be supported or might have constrained capabilities. For more information, see Supplemental Terms of Use for Microsoft Azure Previews.
This repository contains a copilot getting sample that can be used with the the Azure AI Studio preview.
The sample walks through creating a copilot enterprise chat API that uses custom Python code to ground the copilot responses in your company data and APIs. The sample is meant to provide a starting point that you can further customize to add additional intelligence or capabilities. Following the below steps in the README, you will be able to: set up your development environment, create your Azure AI resources and project, build an index containing product information, run your co-pilot, evaluate it, and deploy & invoke an API.
NOTE: We do not guarantee the quality of responses produced by this sample copilot or its suitability for use in your scenario, and responses will vary as development of this sample is ongoing. You must perform your own validation the outputs of the copilot and its suitability for use within your company.
To get started quickly, you can use a pre-built development environment. Click the button below to open the repo in GitHub Codespaces, and then continue the readme!
Once you've opened in Codespaces you can proceed to the next step.
First, clone the code sample locally:
git clone https://github.com/azure/aistudio-copilot-sample
cd aistudio-copilot-sample
Create a new Python virtual environment where we can safely install the SDK packages:
- On MacOS and Linux run:
python3 -m venv .venv source .venv/bin/activate
- On Windows run:
py -3 -m venv .venv .venv\scripts\activate
Now that your environment is activated, install the SDK packages
pip install -r requirements.txt
Finally, install the Azure AI CLI. On Ubuntu you can use this all-in-one installer command:
curl -sL https://aka.ms/InstallAzureAICLIDeb | sudo bash
To install the CLI on Windows and MacOS, follow the instructions here.
Run ai init to create and/or connect to existing Azure resources:
ai init
- This will first prompt to you to login to Azure
- Then it will ask you to select or create resources, choose New Azure AI Project and follow the prompts to create an:
- Azure AI resource
- Azure AI project
- Azure OpenAI Service model deployments (we recommend ada-embedding-002 for embedding, gpt-35-turbo-16k for chat, and gpt-35-turbo-16k or gpt4-32k evaluation)
- Azure AI search resource
- This will generate a config.json file in the root of the repo, the SDK will use this when authenticating to Azure AI services.
Note: You can open your project in AI Studio to view your projects configuration and components (generated indexes, evaluation runs, and endpoints)
Run the following CLI command to create an index using that our code can use for data retrieval:
ai search index update --files "./data/3-product-info/*.md" --index-name "product-info"
The 3-product-info
folder contains a set of markdown files with product information for the fictitious Contoso Trek retailer company. You can run this command using a different folder, or replace the contents in this folder with your own documents.
Note: if you've previously done this step and already have an index created, you can instead run ai config --set search.index.name <existing-index-name>
.
Now that we've created an index, we can generate a .env file that will be used to configure the running code to use the resources we've created in the subsequent steps
ai dev new .env
To run a single question & answer through the sample co-pilot:
python src/run.py --question "which tent is the most waterproof?"
You can try out different sample implementations by specifying the --implementation
flag with promptflow
, semantickernel
, langchain
or aisdk
. To try running with semantic kernel:
python src/run.py --implementation semantickernel --question "what is the waterproof rating of the tent I just ordered?"
To try out the promptflow implementation, check deployment names (both embedding and chat) and index name (if it's changed from the previous steps) in src/copilot_promptflow/flow.dag.yaml
match what's in the .env
file.
python src/run.py --question "which tent is the most waterproof?" --implementation promptflow
The --implementation
flag can be used in combination with the evaluate command below as well.
You can also use the ai
CLI to submit a single question and/or chat interactively with the sample co-pilots, or the default "chat with your data" co-pilot:
ai chat --interactive # uses default "chat with your data" copilot
ai chat --interactive --function src/copilot_aisdk/chat:chat_completion
To run evaluation on a copilot implementations:
python src/run.py --evaluate --implementation aisdk
You can change aisdk
to any of the other implementation names to run an evaluation on them.
You can also use the ai
CLI to do bulk runs and evaluations:
ai chat evaluate --input-data src/tests/evaluation_dataset.jsonl # uses default "chat with your data" copilot
ai chat evaluate --input-data src/tests/evaluation_dataset.jsonl --function src/copilot_aisdk/chat:chat_completion
You can also run all of the evaluations using pytest, and where tests will fail if the metrics are less than 4:
pytest
This will run the tests named src/test_copilot_<implementation>.py
using the evaluation_dataset.jsonl
as a test dataset. This will compute a set of metrics calculated by chatgpt on a 1-5 scale, and will fail that metric if the average score is less than 4. Not all tests are currently passing (this is expected as we work to improve the sample copilot implementations).
To deploy one of the implementations to an online endpoint, use:
python src/run.py --deploy
To test out the online enpoint, run:
python src/run.py --invoke
You can pip install packages into your development environment but they will disappear if you rebuild your container and need to be reinstalled (re-build is not automatic). You may want this, so that you can easily reset back to a clean environment. Or, you may want to install some packages by default into the container so that you don't need to re-install packages after a rebuild.
To add packages into the default container, you can update the Dockerfile in .devcontainer/Dockerfile
, and then rebuild the development container from the command palette by pressing Ctrl/Cmd+Shift+P
and selecting the Rebuild container
command.
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.
This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.