AI Agent Framework — Complete Scaffold

This archive contains a full scaffold for the AI Agent Framework with Django control plane, orchestration, messaging, OpenVINO optimization hooks, observability, and example agents. All files are placeholders and meant to be extended.

Build-Your-Own AI Agent Framework

Build an AI Agent framework (not just an app) that can orchestrate agentic workflows from input to output without using existing agent frameworks like crew.ai, AutoGen, or n8n. The framework should support allowing the definition of agentic workflows as a composition of task flows. These must then be executable. Must be able to monitor and audit them. Apache projects can be used for orchestration, messaging, and storage.

High-Level Guidelines

• Core Features: • Define and execute task flows (DAG or state machine). • Support input handlers, tools/actions, and output actions. • Include memory, guardrails, and observability (logs, metrics).

• Architecture: • Ingress (REST/queue) → Orchestrator → Executors → State/Memory. • Use Apache components (Kafka, Airflow, Camel, etc.) for messaging and orchestration.

The badge above points to DevPrasath6/Intel on GitHub. If you use a different repo path, update the badge URL accordingly.

# AI Agent Framework — Complete Scaffold

This archive contains a full scaffold for the AI Agent Framework with Django control plane, orchestration, messaging, OpenVINO optimization hooks, observability, and example agents. All files are placeholders and meant to be extended.

Build-Your-Own AI Agent Framework

Build an AI Agent framework (not just an app) that can orchestrate agentic workflows from input to
output without using existing agent frameworks like crew.ai, AutoGen, or n8n. The framework should
support allowing the definition of agentic workflows as a composition of task flows. These must then be
executable. Must be able to monitor and audit them. Apache projects can be used for orchestration,
messaging, and storage.

High-Level Guidelines

• Core Features:
• Define and execute task flows (DAG or state machine).
• Support input handlers, tools/actions, and output actions.
• Include memory, guardrails, and observability (logs, metrics).

• Architecture:
• Ingress (REST/queue) → Orchestrator → Executors → State/Memory.
• Use Apache components (Kafka, Airflow, Camel, etc.) for messaging and orchestration.

• Intel Tech:
• Develop and benchmark on Intel® DevCloud.
• Optimize any ML models (LLMs, re-rankers, OCR) with Intel® OpenVINO™.

• Deliverables:
• Framework SDK with APIs for flows, tools, and policies.
• At least two reference agents demonstrated real workflows.
• Design doc + performance benchmarks (pre/post optimization).

• Performance Targets:
• Reliable execution with retries and timeouts.
• Demonstrate Intel optimizations if ML is involved.

• Stretch Goals:
• Multi-agent collaboration, reflection loops, human-in-the-loop steps.

Make the format-on-push workflow run on PRs too (not only pushes to main). Make ruff configuration explicit (pyproject.toml) with your preferred rules. Add pre-commit to requirements-dev.txt or to a developer make target. Revert or tweak any of the earlier helper files I added/removed (e.g., conftest.py, ai_framework shim) if you want a different test discovery approach.

Repository templates and docs

• Issue/PR templates are available under .github/ISSUE_TEMPLATE/ and .github/PULL_REQUEST_TEMPLATE.md.
• CODE_OF_CONDUCT.md and SECURITY.md were added; update the contact email addresses as appropriate.

Running tests

To run the test-suite locally as a contributor, run the following from the repository root:

# create and activate a virtualenv (recommended)
python -m venv .venv; .\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
pip install -r requirements.txt
if (Test-Path requirements-dev.txt) { pip install -r requirements-dev.txt }

# Run the Django tests (uses the bundled test settings which use SQLite)
cd django_app
python manage.py test

# Or run the top-level pytest suite
cd ..
python -m pytest -q --ds=ai_framework.test_settings
CI & Codecov
-------------

The CI uploads coverage artifacts and will attempt to send coverage to Codecov if a
`CODECOV_TOKEN` secret is set in the repository settings. To enable Codecov uploads,
add the `CODECOV_TOKEN` secret (value from your Codecov account) to the repository
secrets in GitHub. If the secret is not set the CI will skip the Codecov upload step.

Notes and troubleshooting

• We attempt to keep test discovery robust across environments. The CI workflow contains an early check that runs python -c "import ai_framework" to surface import/path issues on GitHub Actions. If you see import errors locally, ensure you're running from the repo root or install the package in editable mode (e.g. pip install -e .) so ai_framework is importable.
• If you rely on IDE-run tests, configure the working directory to the repository root so imports resolve the same way CI and the command-line runner do.