[Good First Issue] Port GPT-4 Turbo cost tracking fix to Python SDK

[nagasatish007]

Description

Port the longest-prefix model pricing fix from the TypeScript SDK (PR #2) to the Python SDK.

Context

In TypeScript, gpt-4-turbo-preview was matching gpt-4 pricing ($0.03/1K) instead of gpt-4-turbo pricing ($0.01/1K) — a 3x overcharge. The fix uses longest-prefix matching (sort by key length descending).

TypeScript PR: agentguard-ai/tealtiger-typescript-prod#2

What to change

1. Update get_model_pricing() in the Python SDK to use longest-prefix matching
2. Add enable_cost_tracking option to the Python OpenAI client
3. Expose response.cost as a numeric field
4. Add tests for GPT-4 Turbo preview and dated model aliases

Acceptance criteria

• get_model_pricing() uses longest-prefix match
• gpt-4-turbo-preview returns Turbo pricing, not GPT-4 pricing
• Tests pass: pytest tests/cost/
• Behavior matches TypeScript SDK exactly

Difficulty

🟢 Beginner — direct port from TypeScript, logic is straightforward

[resolvicomai]

I opened the Python SDK fix here: agentguard-ai/tealtiger-python-prod#4

It updates get_model_pricing() to use longest-prefix matching, adds enable_cost_tracking on the OpenAI wrapper, exposes numeric response.cost, and covers GPT-4 Turbo preview/datetime aliases with focused cost-tracking tests.

[nagasatish007]

Merged! Thanks @resolvicomai 🎉

This is the exact Python parity fix we needed — longest-prefix matching, numeric response.cost, and the bonus fix using response.model for cost calculation. Clean tests, matches TypeScript behavior.

You're now a Founding Contributor (#5) — welcome to the tiger pack! 🐯

If you found TealTiger useful, a ⭐ helps others discover it: https://github.com/agentguard-ai/tealtiger