Add README for prompting examples

Introduces a comprehensive README in examples/prompting outlining available notebooks, guides, and conventions for LLM prompting techniques. Provides quick start instructions, learning paths, and contribution guidelines to help users navigate and extend the collection.

Author: paytonison

examples/prompting/README.md

@@ -0,0 +1,102 @@
+Prompting Examples (LLMs)
+
+This folder gathers all prompting-related examples in one place so you can learn, compare, and reuse patterns quickly. It ranges from fundamentals (input formatting, caching) to advanced techniques (meta-prompting, guardrails, evaluation with a judge model), plus model-specific guides and applied recipes.
+
+⸻
+
+Quick start
+	1.	From repo root, install deps and set credentials as described in the main Cookbook README.
+	2.	Open any notebook in Jupyter / VS Code.
+	3.	If a notebook references a model you don’t have access to, swap in a compatible one (e.g., replace gpt-5 with gpt-4o, or o4-mini with gpt-4o-mini).
+	4.	For consistent comparisons, keep temperature low (e.g., 0–0.3) and set any random seed where applicable.
+
+⸻
+
+What’s inside
+
+Foundations & Building Blocks
+	•	How_to_format_inputs_to_ChatGPT_models.ipynb
+Practical patterns for structuring inputs (system/user messages, delimiters, schemas) to reduce ambiguity and boost reliability.
+	•	Prompt_Caching101.ipynb
+Shows how and when to cache stable prompt segments to reduce cost/latency on repeated operations.
+	•	prompt-engineering.txt
+Concise notes/checklist of core prompting heuristics.
+	•	prompts/ (directory)
+Reusable prompt templates/snippets you can copy into your own projects.
+
+Prompt Optimization & Patterns
+	•	Optimize_Prompts.ipynb
+Step-by-step refinements from vague → specific → structured; includes measurable checkpoints (accuracy, determinism, length).
+	•	Enhance_your_prompts_with_meta_prompting.ipynb
+“Prompt the prompter”: use meta-instructions and self-review to systematically improve outputs.
+	•	prompt-optimization-cookbook.ipynb
+A recipe-style tour of optimization tactics (role specification, constraints, formatting contracts, failure-mode nudges).
+	•	prompt-optimization-cookbook/ (directory)
+Companion assets and extended recipes referenced by the notebook.
+
+Guardrails, Evaluation & QA
+	•	Developing_hallucination_guardrails.ipynb
+Patterns to reduce unsupported claims: retrieval checks, cite-or-abstain, schema-validation, and refusal scaffolds.
+	•	Custom-LLM-as-a-Judge.ipynb
+Use an independent “judge” model to evaluate task outputs (rubrics, pairwise comparisons, error taxonomies).
+
+Model-Specific Prompting Guides
+	•	gpt-5_prompting_guide.ipynb
+Guidance for newest-gen models: leveraging extended context/tools while keeping prompts lean and structured. (Swap to your available model if needed.)
+	•	gpt4-1_prompting_guide.ipynb
+Tips tailored to the 4.1 series: instruction clarity, tool-use hooks, JSON contracts.
+	•	o3o4-mini_prompting_guide.ipynb
+Best practices for smaller/efficient models: tighter constraints, shorter exemplars, robust formatting to offset capacity limits.
+	•	Realtime_prompting_guide.ipynb
+Streaming + low-latency interaction patterns (turn-taking prompts, incremental planning, partial results).
+	•	Whisper_prompting_guide.ipynb
+Using prompts to bias/condition speech-to-text (domain terms, speaker/style context, punctuation/number handling).
+	•	Prompt_migration_guide.ipynb
+How to port prompts across model/API versions while preserving behavior (capabilities, defaults, temperature/top-p changes).
+
+Applied Examples
+	•	Unit_test_writing_using_a_multi-step_prompt.ipynb
+Multi-step prompting to draft and refine unit tests from natural-language specs with verification passes.
+	•	Unit_test_writing_using_a_multi-step_prompt_with_older_completions_API.ipynb
+Historical version of the above using the legacy Completions API for comparison/reference.
+
+⸻
+
+Suggested learning paths
+
+New to prompting?
+	1.	How_to_format_inputs… → 2) Optimize_Prompts → 3) Enhance_your_prompts_with_meta_prompting → 4) Prompt_Caching101
+
+Shipping something production-ish
+	1.	Developing_hallucination_guardrails → 2) Custom-LLM-as-a-Judge → 3) prompt-optimization-cookbook.ipynb → 4) your model-specific guide
+
+Speech / real-time track
+	1.	Whisper_prompting_guide → 2) Realtime_prompting_guide
+
+Legacy → modern migration
+	1.	Unit_test… (older completions) → 2) Unit_test… (multi-step) → 3) Prompt_migration_guide
+
+⸻
+
+Conventions for this folder
+	•	File naming: use snake_case and end with _guide.ipynb or a clear verb phrase (e.g., optimize_prompts.ipynb).
+	•	Comparisons: when demonstrating “bad vs. good,” keep identical tasks & evaluation across cells; only change the prompt.
+	•	Determinism: set low temperature and, where relevant, seeds to keep diffs meaningful.
+	•	Outputs: prefer structured outputs (JSON/Markdown tables) to ease automated checking.
+	•	Safety: when showing failure modes (e.g., hallucinations), add a final cell with a fix pattern (cite-or-abstain, schema checks, etc.).
+
+⸻
+
+Contributing
+	•	Add a short intro cell explaining the goal and what success looks like.
+	•	Include at least one automatable check (regex/schema/assert) to make improvements measurable.
+	•	If your example targets a specific model family, call it out up front and suggest fallback models.
+	•	Keep any shared assets in this folder (or its subfolders) and use relative paths.
+
+⸻
+
+Notes
+	•	Model names evolve. If a referenced model isn’t available in your account/region, substitute the closest family (e.g., gpt-4o / gpt-4o-mini).
+	•	Some notebooks intentionally show anti-patterns for teaching—read the headers to avoid copying them into production.
+
+If you spot missing patterns (retrieval-augmented prompts, evaluation rubrics for specific domains, code-gen chains, etc.), feel free to open a PR adding a focused notebook here.