feat/masking - use normal middleware state

[Akatuoro]

Small update to #71 ...

Instead of manually integrating into langchains graph state, we can simply use the middleware state to persist our pseudonym data.

[hyperspace-pr-bot]

Summary

The following content is AI-generated and provides a summary of the pull request:

---

Title: Persist pseudonymization with middleware state

Category

Refactor

Summary

This PR simplifies pseudonymization state handling by using LangChain middleware state directly instead of manually integrating pseudonym data into graph checkpoint channels.

Changes

• Adds a middleware stateSchema for pseudonymization data:
  • per-run seed
  • hashToOriginal mapping
• Initializes PseudoSession from middleware state in beforeAgent.
• Persists updated pseudonymization state by returning a LangGraph Command from wrapToolCall.
• Removes custom pseudonymization checkpoint channel handling from the checkpoint saver.
• Simplifies PseudoSession by removing thread-based loading, caching, and graph-state lookup logic.
• Removes the custom pseudonymization thread-id helper, since state persistence is now handled by middleware.

Impact

Pseudonym data is now persisted through the standard LangChain middleware state mechanism, reducing custom checkpointing logic and keeping masking behavior aligned with the framework’s intended state handling.

Related: #71

Have you...

• Added relevant entry to the change log?

---

• 🔄 Regenerate and Update Summary
• ✏️ Insert as PR Description (deletes this comment)
• 🗑️ Delete comment

PR Bot Information

Version: 1.31.43

• Output Template: Repository PR Template
• File Content Strategy: Full file content
• Event Trigger: pull_request.opened
• LLM: gpt-5.5
• Correlation ID: 54fa8900-b5fb-11f1-9af4-14ce24118e2b
• Summary Prompt: Default Prompt
• GithubContextProvider: feat: masking of PII #71