Outskill Workshop Monorepo

Workshop code for Outskill GenAI tracks: CrewAI agents (beginner → advanced), RAG experiments, and small multimodal / diffusion scripts. This is a reference and teaching repo, not a versioned library—expect per-module dependencies and entrypoints documented in place.

---

Contents

• Quick start
• What’s inside
• Configuration
• Layout
• Dependencies
• Running agents

---

Quick start

git clone https://github.com/ishandutta0098/outskill.git && cd outskill

# Pick one:
conda create -n agents python=3.11 -y && conda activate agents
# or
python3 -m venv .venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate

Install what you need for the path you’re using (there is no root requirements.txt). Examples:

pip install crewai crewai-tools python-dotenv          # agent tracks
pip install -r rags/myntra_rag/requirements.txt      # Myntra RAG subproject

Create a .env at the repo root (see Configuration). Then run from the relevant folder, e.g. agents/advanced, so local imports (agents, tasks, …) resolve.

---

What’s inside

Module	Purpose
agents/	CrewAI curricula: single-agent, sequential multi-agent, parallel crews + custom tools. LLM calls in examples use OpenRouter. Curriculum detail: agents/README.md.
rags/	LangChain / LlamaIndex / Haystack-style experiments; Myntra RAG sample with pinned deps.
multimodal/	Image–text demos (e.g. CLIP via transformers).
diffusion/	OpenAI image generation samples.
docs/, assets/	Supporting PDFs, papers, resumes, static assets.
task_outputs/	Generated outputs; safe to regenerate or gitignore-heavy.

Not in scope: a unified CLI, Docker compose for every service, or a single lockfile for all subprojects—by design.

---

Configuration

Root .env (gitignored). Set only what you use:

OPENAI_API_KEY=
OPENROUTER_API_KEY=
EXA_API_KEY=
ADZUNA_APP_ID=
ADZUNA_API_KEY=

Most agent examples expect OPENROUTER_API_KEY. Other keys back optional tools (search, job APIs, etc.)—see the script you run.

---

Layout

outskill/
├── agents/
│   ├── beginner/           # single agent, OpenRouter LLM
│   ├── intermediate/       # sequential crew, built-in tools
│   ├── advanced/           # parallel crews, custom tools, optional API helpers
│   ├── dummy_logs/
│   └── README.md           # per-level objectives
├── rags/
├── multimodal/
├── diffusion/
├── docs/
├── assets/
└── task_outputs/           # + agents/*/task_outputs/ when scripts write files

---

Dependencies

Location	Notes
rags/myntra_rag/requirements.txt	Pinned LangChain-related stack for that subproject.
Everything else	Declared in file headers or notebook cells (pip install …). Roll your own requirements.txt if you want one fat environment.

---

Running agents

1. Activate the env (conda activate agents or .venv).
2. cd agents/<beginner|intermediate|advanced>.
3. python main.py (or the script documented in that directory).

Imports assume you run from the track directory (or adjust PYTHONPATH accordingly). If something fails on import, you’re usually in the wrong working directory.

---

License

MIT License

---

Acknowledgements

Workshop material for Outskill; upstream tools (CrewAI, OpenRouter, Hugging Face, etc.) are cited in individual scripts and notebooks.