https://github.com/armando71988padilla-cloud/lab-rescue-agent
AI agent system that turns failed certification labs into safe recovery plans, readiness gaps, and grounded practice checks.
https://github.com/armando71988padilla-cloud/lab-rescue-agent
ai-agents azure-functions certification hackathon microsoft-foundry python reasoning-agents safety
Last synced: 24 days ago
JSON representation
AI agent system that turns failed certification labs into safe recovery plans, readiness gaps, and grounded practice checks.
- Host: GitHub
- URL: https://github.com/armando71988padilla-cloud/lab-rescue-agent
- Owner: armando71988padilla-cloud
- Created: 2026-06-06T08:53:13.000Z (25 days ago)
- Default Branch: main
- Last Pushed: 2026-06-06T09:13:16.000Z (25 days ago)
- Last Synced: 2026-06-06T10:21:13.493Z (25 days ago)
- Topics: ai-agents, azure-functions, certification, hackathon, microsoft-foundry, python, reasoning-agents, safety
- Language: Python
- Homepage:
- Size: 28.3 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Lab Rescue Agent
[](https://github.com/armando71988padilla-cloud/lab-rescue-agent/actions/workflows/smoke-test.yml)
Lab Rescue Agent is a multi-agent certification lab recovery system for Microsoft AI Agents League.
It helps a technical learner recover from a failed hands-on certification lab, maps the failure to certification readiness, generates grounded practice questions, and verifies the final output for safety and reliability.
## Core idea
From failed lab to certification readiness in one safe, explainable multi-agent workflow.
## Judge quick path
1. Watch the demo video.
2. Run the smoke test: `PYTHONPATH="$PWD/src" python3 tests/test_demo_workflow.py`.
3. Run the quality gate: `PYTHONPATH="$PWD/src" python3 -m lab_rescue_agent evaluate`.
4. Export a scenario: `PYTHONPATH="$PWD/src" python3 -m lab_rescue_agent export az204-functions-storage-error`.
5. Open the static HTML dashboard under `demo/output/`.
6. Inspect the agent trace ledger JSON and manager executive brief.
7. Review `docs/foundry_readiness.md` and `docs/foundry_live_probe.md` for the Microsoft Foundry path.
## Current demo status
- Local deterministic Python demo works
- Seven specialized agents are wired into one enterprise readiness workflow
- Synthetic data only
- Grounded synthetic knowledge docs are cited in outputs
- Stdlib smoke test passes without extra test dependencies
- Evaluation harness validates 25 quality checks across all scenarios
- Static HTML dashboards, agent trace ledgers, and manager executive briefs export for each scenario
- Optional Microsoft Foundry live-probe guide is documented without requiring cloud access for judging
## Agent workflow
1. Lab Triage Agent diagnoses the failed lab pattern.
2. Recovery Planner Agent creates fix, rollback, and verification steps.
3. Learning Path Agent maps the failure to certification readiness.
4. Study Plan Agent creates a capacity-aware recovery schedule.
5. Assessment Agent generates grounded practice questions.
6. Manager Insights Agent summarizes aggregate team readiness signals.
7. Safety Verifier Agent checks citations, rollback, verification, synthetic data posture, and secret hygiene.
## Demo scenario
Synthetic learner L-1001 is preparing for AZ-204 and fails an Azure Functions HTTP trigger lab.
The default scenario uses an AZ-204 Azure Functions failure where AzureWebJobsStorage is missing or empty. The project now also includes AZ-400 pipeline approval and AZ-104 VM NSG connectivity recovery scenarios.
## List demo scenarios
```bash
PYTHONPATH="$PWD/src" python3 -m lab_rescue_agent scenarios
```
## Run the default demo
```bash
PYTHONPATH="$PWD/src" python3 -m lab_rescue_agent demo
```
## Run a specific scenario
```bash
PYTHONPATH="$PWD/src" python3 -m lab_rescue_agent demo az400-pipeline-approval-blocked
```
## Run the smoke test
```bash
PYTHONPATH="$PWD/src" python3 tests/test_demo_workflow.py
```
## Export demo reports
```bash
PYTHONPATH="$PWD/src" python3 -m lab_rescue_agent export
```
This writes scenario-specific Markdown reports, JSON summaries, static HTML dashboards, agent trace ledgers, and manager executive briefs under `demo/output/`.
Expected result:
```text
DEMO_WORKFLOW_SMOKE_OK
```
## Documentation
- Architecture: docs/architecture.md
- Microsoft integration plan: docs/microsoft_integration.md
- Microsoft Foundry readiness: docs/foundry_readiness.md
- Microsoft Foundry live probe guide: docs/foundry_live_probe.md
- Submission notes: docs/submission_notes.md
- Demo video script: docs/demo_video_script.md
- Demo prompts: demo/sample_prompts.md
## Hackathon alignment
- Track focus: reasoning agents
- Architecture: local-first Python multi-agent workflow
- Microsoft IQ layer focus: Foundry IQ style grounding over synthetic lab and certification knowledge
- Safety posture: synthetic data only, no secrets, rollback required, verification required
- Next integration target: Microsoft Agent Framework or Azure AI Foundry project endpoint
## Data policy
This repository uses synthetic learner data, synthetic lab logs, and synthetic knowledge documents only.
No real employee data, customer data, connection strings, secrets, or private logs belong in this repository.
## Current limitations
- The current version is deterministic and local-first.
- Live Microsoft Foundry integration is planned next.
- The first scenario covers Azure Functions storage configuration recovery only.
## Project structure
- src/lab_rescue_agent/agents: specialized agent modules
- src/lab_rescue_agent/core: deterministic report orchestration
- data/synthetic: synthetic lab failure scenarios, multi-scenario catalog, and team learning signals
- knowledge: approved synthetic grounding documents
- demo: sample demo prompts, commands, and exported report outputs
- tests: no-dependency smoke tests