Introducing the AICredits Cookbook: Runnable LLM Examples for Indian Developers

The gap between tutorials and real projects

Most AI tutorials assume you already have an OpenAI account, an international credit card to fund it, and separate accounts for every provider you want to compare. For a student or a new AI engineer in India, that's three blockers before you write a line of code.

We built the AICredits Cookbook to remove them. It's a public GitHub repo of runnable, beginner-friendly example projects that use AICredits as the single API — one key, INR billing via UPI/Razorpay, and access to 300+ models across OpenAI, Anthropic, Google Gemini, DeepSeek, Mistral, and xAI.

Repo: github.com/chetanrakheja/aicredits-cookbook

What's in the cookbook today

Three full recipes ship at launch, each a standalone Python 3.10+ project with a README, sample data, and an honest cost estimate for running the demo:

| Recipe | What it teaches | Estimated demo cost | |--------|------------------|----------------------| | Resume/JD Matcher | JSON-mode structured output, evidence-grounded scoring instead of vague claims | ~₹0.10–₹0.50 | | Exam MCQ Generator | Rubric-based generation and evaluation, batch prompting | ~₹0.10–₹0.75 | | Multi-Model Eval Harness | Comparing cost and quality across models through one endpoint | ~₹0.25–₹1.50 |

The Resume/JD Matcher is a good place to start. It scores a candidate against a job description and asks the model to quote a specific resume line as evidence for every strength it claims, so the output isn't just a confident-sounding guess:

from openai import OpenAI
import os
 
client = OpenAI(
    base_url="https://api.aicredits.in/v1",
    api_key=os.environ["AICREDITS_API_KEY"],
)
 
response = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}],
    response_format={"type": "json_object"},
)
print(response.choices[0].message.content)

That's the entire integration surface: one base URL, one sk-... key from your AICredits dashboard, and <provider>/<model-name> in the model field to pick whichever provider fits the task.

Twelve more ideas if you want to build your own

The repo's PROJECT_IDEAS.md lists 12 problem statements at increasing difficulty, each with what you'll learn and why AICredits helps:

• WhatsApp order-taking bot for a kirana store (function calling, conversation state)
• UPI expense categorizer from bank SMS (structured outputs)
• Regional-language complaint summarizer (audio pipelines, Indic languages)
• Resume/JD matcher with explanations (shipped as a full recipe)
• College notes RAG bot (chunking, embeddings, citation)
• English-to-Hindi document simplifier (translation evals, streaming)
• Customer-support autoresponder with human handoff (guardrails, confidence thresholds)
• AI code-review GitHub Action on a budget (per-key spend caps)
• Voice IVR replacement for appointment booking (STT/TTS loop)
• Price-comparison shopping agent (tool use, caching)
• Exam MCQ generator and evaluator (shipped as a full recipe)
• Multi-model eval harness (shipped as a full recipe)

None of these need a second provider account. Swap model="openai/gpt-4o-mini" for anthropic/claude-3-5-haiku or deepseek/deepseek-chat and the same code keeps running.

Frequently Asked Questions

Do I need an AICredits account to use the cookbook?

Yes — each recipe reads an AICREDITS_API_KEY from a .env file. Sign up at aicredits.in, add wallet credits in INR, and create a key from the dashboard.

Is the cookbook free to use?

The code is free and open source (MIT license). Running a recipe costs whatever the underlying model call costs — each recipe README states a realistic estimate, typically under ₹1 for the included sample data.

Are TypeScript versions coming?

Python ships first. TypeScript versions of the three launch recipes are planned as a follow-up.

Can I contribute a recipe?

Yes. The repo's contribution note asks for small, runnable, beginner-friendly examples with real sample data and an honest cost estimate — open a PR.

Get started

• Star or clone the repo: github.com/chetanrakheja/aicredits-cookbook
• Get an API key and read the full integration docs: aicredits.in/docs/quickstart
• Related reading: Using the Anthropic SDK with AICredits