refactor(py/samples): standardize naming taxonomy, consolidate shared logic, and close feature coverage gaps

[yesudeep]

Summary

Standardizes sample directory naming with a consistent taxonomy, extracts shared
flow logic and types into a common shared/ module, adds new framework feature
demos, and closes all feature coverage gaps in provider samples. Every sample now
follows a clear provider-*, framework-*, or web-* prefix convention.

Rationale

The samples directory had inconsistent naming (e.g., menu, flask-hello,
ollama-hello, short-n-long) making it hard to discover what a sample
demonstrates. Provider samples also had significant code duplication — the same
flow patterns (greeting, streaming, tools, vision, etc.) were reimplemented in
each sample. This PR:

1. Establishes three naming categories:
   • provider-* — model provider and plugin-specific samples
   • framework-* — Genkit framework feature demos (context, tools, RAG, etc.)
   • web-* — web framework integration samples (Flask, ASGI, multi-server)
2. Extracts shared types and flow logic into samples/shared/ so provider samples
   delegate to common implementations, reducing boilerplate
3. Adds a setup_sample() utility for one-call Rich traceback + structlog setup
4. Closes all ❌ cells in the feature coverage matrix

Changes

Sample renames (34 directories):

Old Name	New Name
amazon-bedrock-hello	provider-amazon-bedrock-hello
anthropic-hello	provider-anthropic-hello
cloudflare-workers-ai-hello	provider-cloudflare-workers-ai-hello
compat-oai-hello	provider-compat-oai-hello
deepseek-hello	provider-deepseek-hello
firestore-retreiver	provider-firestore-retriever (typo fix!)
google-genai-code-execution	provider-google-genai-code-execution
google-genai-context-caching	provider-google-genai-context-caching
google-genai-hello	provider-google-genai-hello
google-genai-vertexai-hello	provider-google-genai-vertexai-hello
google-genai-image / media-models-demo	provider-google-genai-media-models-demo
google-genai-vertexai-image	provider-google-genai-vertexai-image
huggingface-hello	provider-huggingface-hello
microsoft-foundry-hello	provider-microsoft-foundry-hello
mistral-hello	provider-mistral-hello
model-garden	provider-vertex-ai-model-garden
observability-hello	provider-observability-hello
ollama-hello	provider-ollama-hello
vertex-ai-vector-search-bigquery	provider-vertex-ai-vector-search-bigquery
vertex-ai-vector-search-firestore	provider-vertex-ai-vector-search-firestore
vertexai-rerank-eval	provider-vertex-ai-rerank-eval
xai-hello	provider-xai-hello
evaluator-demo	framework-evaluator-demo
format-demo	framework-format-demo
menu	framework-restaurant-demo
prompt-demo	framework-prompt-demo
realtime-tracing-demo	framework-realtime-tracing-demo
tool-interrupts	framework-tool-interrupts
flask-hello	web-flask-hello
multi-server	web-multi-server
short-n-long	web-short-n-long

New samples:

Sample	Description
framework-context-demo	Context propagation through flows, tools, and generate
framework-dynamic-tools-demo	Dynamic tool registration at runtime
framework-middleware-demo	Custom retry and logging middleware

Deleted samples:

Sample	Reason
ollama-simple-embed	Merged into provider-ollama-hello

Shared module consolidation (samples/shared/):

• shared/types.py — centralized Pydantic input/output models used across all provider samples (GreetingInput, StreamInput, CharacterInput, CodeInput, WeatherInput, ImageDescribeInput, ReasoningInput, etc.)
• shared/flows.py — reusable flow logic functions (generate_greeting_logic, generate_streaming_story_logic, describe_image_logic, solve_reasoning_problem_logic, etc.)
• shared/logging.py (new) — setup_sample() utility that configures Rich tracebacks + colored structlog/stdlib logging in one call
• shared/__init__.py — updated re-exports for all new types and functions

Provider sample enhancements:

• provider-ollama-hello: added Moondream v2 model for object detection (detect_objects flow with bounding boxes), explicit model parameters for tool-calling flows
• All provider samples refactored to use shared types and flow logic, removing duplicated boilerplate
• Feature coverage matrix now 100% clean (0 ❌ cells)

Other changes:

• Samples README rewritten with new taxonomy, architecture diagram, and feature coverage matrix
• pyproject.toml package names updated to match new directory names
• sample_parity_roadmap.md added, old sample_parity_analysis.md removed
• uv.lock regenerated

Testing

• bin/lint passes with 0 errors
• All pyproject.toml names match directory names (consistency check passes)
• All renames verified — no broken imports or missing files

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!