Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.narrativebanking.com/llms.txt

Use this file to discover all available pages before exploring further.

Coach is the in-product conversational assistant your users talk to when they need clarity on what is happening in their business and what to do next. In the current app experience, users ask free-form questions, get guided answers in chat, open linked detail views, and continue with follow-up prompts that turn insight into concrete next steps.

What’s new in v1.5.7

v1.5.7 makes Coach feel more like a clear business companion and less like a generic chatbot.
  • Responses use a cleaner full-width layout instead of a narrow answer bubble.
  • Daily briefing items appear as individual insight cards for easier scanning.
  • Answers avoid repetitive scope disclaimers when the user is already asking a business question.
  • Non-business answers are shorter and gently return the user to business topics.
  • Coach leads with evidence and context rather than sounding like it is making the decision for the user.
  • Coach is more careful around regulated financial, tax, and legal topics, where it should provide guidance and suggest professional support instead of advice.
  • Source sections are shown only when useful source links are available.
  • Coach should not claim it has created a project or completed an action unless that action is actually available.
  • Partner deployments can configure coaching tone and partner referral language.

Functional flow

Load current conversation context and prior history when the user opens Coach.
Render assistant output incrementally so users see progress instead of waiting for one final payload.
Handle text chunks, completion signals, and runtime errors as different event types in the UI.
After each completed answer, enable feedback capture and optional guided next-question suggestions.

Integration considerations

https://mintcdn.com/narrative-b13c445c/qi6sqKB_OLTe8ty2/SVG/frontend-shell.svg?fit=max&auto=format&n=qi6sqKB_OLTe8ty2&q=85&s=13c676d362a2a3a7a60bfa0fae245ad6

Session model

Keep conversation state bound to authenticated user context so resumes and restores remain coherent.
https://mintcdn.com/narrative-b13c445c/qi6sqKB_OLTe8ty2/SVG/coach-chat.svg?fit=max&auto=format&n=qi6sqKB_OLTe8ty2&q=85&s=48321b24059ca857f3f87b96dae8e1a6

Streaming lifecycle

Design the UI around in-progress, completed, and interrupted answer states.
https://mintcdn.com/narrative-b13c445c/qi6sqKB_OLTe8ty2/SVG/domain-services.svg?fit=max&auto=format&n=qi6sqKB_OLTe8ty2&q=85&s=4d68cb4946e69d67738f55ee16264633

Context binding

Attach conversations to relevant product entities so follow-up guidance stays grounded.
https://mintcdn.com/narrative-b13c445c/qi6sqKB_OLTe8ty2/SVG/refusal-behavior.svg?fit=max&auto=format&n=qi6sqKB_OLTe8ty2&q=85&s=7f2e9e92249aa3eb1c27b298f31fd56f

Feedback loop

Capture explicit user sentiment after messages complete to improve assistant relevance over time.

Common risk patterns

  • Session restore drift after account or context switches.
  • Duplicate client rendering when stream completion is not handled idempotently.
  • Feedback double-submission when retries are not deduplicated.