Use Cases

This document illustrates how a third-party vendor can integrate Infyn API in two different product scenarios:

1. Code-Review Bot for Software Teams
2. AI Customer-Support Assistant for e-commerce

---

​

1 · Code-Review Bot Vendor

​

Product vision

A SaaS company (e.g. CodeRabbit) offers an AI that reviews pull-requests, leaves inline comments, and chats with developers about code smells and style guidelines.

​

How the vendor uses Infyn

​

Example timeline

1. Developer Bob opens PR-58 on the analytics-service. Bob is already mapped to Infyn (user_id = "u-123").
   • User Facts pulled: preferred_language: rust, max_line_length: 120, review_tone: concise.
   • Memories pulled: the developer’s request in PR-44 to “stop flagging if-else statements” as style issues.
2. CodeRabbit sends the diff to /v2/chat:
   Copy

   {
     "user_id": "u-123",
     "message": "<git-diff>",
     "thread_id": "pr-58"
   }
   
3. Infyn replies with a structured review that clearly separates insights:
   • From User Facts → flags three lines over 120 characters and writes suggestions in a concise tone.
   • From Memories → skips flagging if-else statements, honoring Bob’s request captured in PR-44.
4. Bob replies: “Let’s cap my line length at 100 instead.”
   → Infyn overwrites the user fact max_line_length: 100 (facts are authoritative and mutable).
5. When the PR is merged Infyn stores its final inline comments as new memories so future reviews can reference how the SQL-injection risk was fixed here (memories accumulate, never overwritten).

​

Benefits

• Review comments respect Bob’s personal style prefs.
• Past advice is reused automatically; no re-inventing feedback.

---

​

2 · AI Customer-Support Assistant Vendor

​

Product vision

HelpDesk AI integrates into e-commerce sites, answering shoppers’ questions, tracking order issues, and learning each shopper’s preferences.

​

How the vendor uses Infyn

• Recent order conversations (memories)
• Stable facts (preferred payment method, usual sizes, allergies)

​

Example timeline

1. Shopper Emma starts a chat: “My espresso machine (order #9123) arrived with a cracked carafe. Also, please send future deliveries to my office.” (user_id = "e-999").
   • User Facts pulled: preferred_shipping: express, default_address: home, allergies: peanuts.
   • Memories pulled: last month’s conversation about a broken glass lid (order #9001).
2. HelpDesk AI forwards the message to /v2/chat with thread_id = "support-9123".
3. Infyn composes a reply by combining both data sources:
   • From Memories → skips redundant troubleshooting steps because it recalls Emma already tried them for order #9001 and directly offers a replacement.
   • From User Facts → proposes an express replacement shipment to her home address (current default).
4. Emma responds: “Ship the replacement to my office and switch me to standard shipping.”
   → Infyn overwrites default_address: office and preferred_shipping: standard inside user facts.
5. Infyn confirms the updates and logs this conversation as a new memory so future agents see the context, while the updated shipping preferences live in user facts and are applied automatically next time.

​

Benefits

• Assistant maintains consistent, up-to-date profile facts (shipping, allergies).
• Episodic memories (past orders) improve answer relevance without bloating the prompt.

---

​

Takeaways

• Structured user_facts keep the authoritative settings/minutiae small and fast.
• Unstructured memories capture the rich, ever-growing conversation and code history.
• Both vendors need only two core calls—/users and /v2/chat—to unlock personalised AI.