denissa4/aiagent-master repository overview

⁠NLSQL Bot + Anomaly Detection & TalkToDocs:Llama-Index

Build locally:

docker build -f Dockerfile -t ai-agent .

Run locally:

docker run --rm -p 8080:80 --env-file .env ai-agent

Mimic Azure web app resource specs:

docker run --rm -p 8080:80 --env-file .env --cpus="2.0" --memory="8g" ai-agent

Azurite Local Blob connection string: DefaultEndpointsProtocol=http;AccountName=devstoreaccount1;AccountKey=Eby8vdM02xNOcqFlqUwJPLlmEtlCDXJ1OUzFT50uSRZ6IFsuFq2UVErCz4I6tq/K1SZFPTOtr/KBHBeksoGMGw==;BlobEndpoint=http://host.docker.internal:10000/devstoreaccount1;

⁠NLSQL Envs:

⁠Anomaly Detection Envs:

⁠Talk-to-Docs Envs:

⁠Default Format Prompt:

You must always respond in Markdown and plain text only. Never respond in JSON, code blocks containing JSON, or any other structured data format. Do not include any pre-response tags, metadata, system notes, or explanations about formatting. Respond directly to the user’s request without any introductory or wrapper text.

⁠Default Graph Template:

{% chat role="system" %}
{system_prompt}
## Response Format Rules\n{format_prompt}
{% endchat %}
{% chat role="user" %}
Context: {{ context_str }}
User Query: {{ query_str }}
{% endchat %}

• Notes:
  • Custom templates must be writting in Jinja syntax format (see above example).
  • {system_prompt} is where LanguageModelContext value will be injected.
  • {format_prompt} is where FormatPrompt value will be injected.
  • Refer to Llama Index Rich Prompt Template⁠ for prompt layout.

⁠Supported LLM providers:

• Google Cloud Platform (GCP)
• Amazon Web Services (AWS)
• Azure OpenAI (AzureOpenAI)
• Hugging Face (HuggingFace)
• Sage Maker (SageMaker)
• More coming soon...

⁠Embedding models:

Here are some embedding models that have been tested with the app:

Notes:

• Only Hugging Face and AzureOpenAI Embedding models are supported at the moment.
• For standard RAG mode document chunking overlap is a minimum of 50 words OR 10% if higher than 50 words.

⁠Tests

• Standard tests

  • Standard Talk to Docs tests will automatically run during the application start-up process. This consists of 141 assertion tests.

• Cloud integration tests

  • Cloud integration tests will run when RUN_INTEGRATION_TESTS = true. These tests cover document storage downloads and uploads, as well as LLM and Embedding models requests and responses.

• End-to-end tests

  • End-to-end tests will run when RUN_E2E_TESTS = true and after the full application initialization process is complete. These test cover the entire talk-to-docs flow and will pass if a proper final response is recieved via a Direct Line request.

⁠Other Envs: