examples(python): add langchain-krewe-inference-chatbot

[harrydev44]

Summary

Adds the Python sibling to the TypeScript example proposed in #1222: a LangChain chatbot that uses AgentKit's built-in x402_action_provider to call krewe — a decentralized AI inference network on Base mainnet — for four pay-per-call job kinds.

Same orchestrator, same pricing table, same handshake — just native Python idioms and the canonical langchain-cdp-chatbot example structure.

Why both PRs

The agent ecosystem splits roughly 50/50 between TypeScript (Vercel AI SDK, Mastra, agentkit-langchain TS) and Python (LangChain Python, autogen, CrewAI, pydantic-ai). Shipping only one half leaves the other audience locked out of the x402 action provider pattern. These two PRs together demonstrate the canonical "x402-paywalled inference as a first-class agent tool" recipe against a real production endpoint, on both sides of the monorepo.

What's in this PR

• python/examples/langchain-krewe-inference-chatbot/chatbot.py — entrypoint, mirrors langchain-cdp-chatbot/chatbot.py (chat + autonomous modes, identical CDP wallet wiring, persistent wallet file). Only the action-provider list and the system prompt differ.
• pyproject.toml — uv-managed, lifted from langchain-cdp-chatbot with only the package name + description changed. Same dep pins, same ruff config, same [tool.uv.sources] workspace deps.
• Makefile — identical install / run / lint / format targets.
• .env.local — adds NETWORK_ID, KREWE_PREDICT_URL, KREWE_MAX_PAYMENT_USDC (all with sensible defaults).
• README.md — prompts to try, prerequisites, env vars, run instructions, plus a section on the standalone krewe-x402 PyPI package for callers who don't need the full AgentKit stack.
• Root README.md — adds the new example to the Python directory tree (alphabetically between langchain-eth-account-chatbot and langchain-nillion-secretvault-chatbot).

Pricing surfaced in the system prompt

kind	Cost
text.structure	$0.005
text.embed	$0.01
web.scrape	$0.02
text.complete	$0.05

Quoted server-side per call; AgentKit's max_payment_usdc ceiling (default $0.10 in this example) protects against unexpected prices.

Production endpoint verified

The orchestrator at https://krewe-orchestrator-production.up.railway.app/v2/predict is live and returns the documented WWW-Authenticate: x402 scheme=\"x402-eip3009-usdc\" header on the 402 challenge — verified by hand with the existing x402_action_provider.

Test plan

• chatbot.py mirrors the proven langchain-cdp-chatbot template structure — wallet config, agent setup, chat/auto loops are identical patterns.
• x402 challenge against the production endpoint verified by hand: curl -X POST .../v2/predict returns 402 with the documented challenge JSON and WWW-Authenticate header.
• Tooling files (pyproject.toml, Makefile, ruff config) lifted verbatim from langchain-cdp-chatbot to match repo conventions.
• Root README.md Python examples tree updated alphabetically.

Happy to tweak the system prompt wording, the bundled action-provider list, or the env-var naming.

---

Companion PR: #1222 (TypeScript sibling).

Links:

• krewe site: https://www.krewe.world
• krewe repo: https://github.com/krewe-AI/krewe
• x402 protocol docs (krewe-side): https://github.com/krewe-AI/krewe/blob/main/docs/x402.md
• $KREW on Basescan: https://basescan.org/address/0xc411E6deE1F70F57F08714D3bd54b4F36f904B07
• Standalone Python SDK: https://pypi.org/project/krewe-x402/ (and source at https://github.com/krewe-AI/krewe/tree/main/clients/krewe-x402-python)

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 1 Sum 2